Accurate estimation of relaxation time T₁ from MRI images is increasingly important for some clinical applications. Low noise, high resolution, fast and accurate T₁ maps from MRI images of the brain can be performed using a dual flip angle method. However, accuracy is limited by the scanners ability to deliver the prescribed flip angle due to the B₁ inhomogeneity, particularly at high field strengths (e.g. 3T). One of the most accurate methods to correct that inhomogeneity is to acquire a subject-specific B₁ map. However, since B₁ map acquisition takes up precious scanning time and most retrospective studies do not have B₁ map, it would be desirable to perform that correction from a template. For this work a dual repetition time method was used for B₁ map acquisition in five normal subjects. Inaccuracies due to misregistration of acquired T₁-weighted images were corrected by rigid registration, and the effects of misalignment were compared to those of B1 inhomogeneity. T₁-intensity histograms were produced and three-Gaussian curves were fitted for every fully-, partially- and non-corrected histogram in order to estimate and compare the white and gray matter peaks. In addition, in order to reduce the scanning time we designed a template based correction strategy. Images from different subjects were aligned using a twelve-parameter affine registration, and B₁ maps were aligned according to that transformation. Recomputed T₁ maps showed a significant improvement with respect to non-corrected ones. These results are very promising and have the potential for clinical application.

Purpose: Diffusion tensor imaging (DTI) is an inherently quantitative imaging technique that measures the diffusivities of water molecules in tissue. However, the accuracy of DTI measurements depends on many factors such S/N ratio and magnet field strength. Therefore, before quantitative assessment of tumor progression based on DTI metric changes can be made with confidence, one have to assess the accuracy or variance in the DTI metrics. This is especially important for multi-institutional clinical trials or for large institutions where patients may be imaged on multiple MR scanners at multiple times in follow up studies. In this presentation, we studied the feasibility of using CSF as an internal QC marker for data acquisition and processing qualities. Method: ADC and FA of CSF for brain tumor patients' DTI studies (total of 85 scans over three years) were analyzed. In addition, a phantom was used to check the inherent variations of the MR systems. Results: The results show that the coefficient of variations for ADC and FA are 8.4% and 13.2% in CSF among all patients. For all DTI scans done on 1.5 T scanners, they are 7.4% and 9.1%, while for 3T they are 9.8% and 18% respectively. Conclusion: CSF can be used as an internal QC measure of the DTI acquisition accuracy and consistency among longitude studies on patients, making it a potentially useful in multi-institutional trials.

Magnetic Resonance Imaging (MRI) techniques have achieved much importance in providing visual and quantitative information of human body. Diffusion MRI is the only non-invasive tool to obtain information of the neural fiber networks of the human brain. The traditional Diffusion Tensor Imaging (DTI) is only capable of characterizing Gaussian diffusion. High Angular Resolution Diffusion Imaging (HARDI) extends its ability to model more complex diffusion processes. Spherical harmonic series truncated to a certain degree is used in recent studies to describe the measured non-Gaussian Apparent Diffusion Coefficient (ADC) profile. In this study, we use the sampling theorem on band-limited spherical harmonics to choose a suitable degree to truncate the spherical harmonic series in the sense of Signal-to-Noise Ratio (SNR), and use Monte Carlo integration to compute the spherical harmonic transform of human brain data obtained from icosahedral schema.

According to Traditional Chinese Medicine (TCM) theory and certain clinical treatment reports, the sustained effects of acupuncture indeed exist, which may last several minutes or hours. Furthermore, increased attention has fallen on the sustained effects of acupuncture. Recently, it is reported that the sustained acupuncture effects may alter the default mode network (DMN). It raises interesting questions: whether the modulations of acupuncture effects to the DMN are still detected at other acupoints and whether the modulation patterns are different induced by different acupoints. In the present study, we wanted to investigate the questions. An experiment fMRI design was carried out on 36 subjects with the electroacupuncture stimulation (EAS) at the three acupoints: Guangming (GB37), Kunlun (BL60) and Jiaoxin (KI8) on the left leg. The data sets were analyzed by a data driven method named independent component analysis (ICA). The results indicated that the three acupoints stimulations may modulate the DMN. Moreover, the modulation patterns were distinct. We suggest the different modulation patterns on the DMN may attribute to the distinct functional effects of acupoints.

Purpose: To develop, test, and evaluate a novel unsupervised machine learning method for the analysis of multidimensional biomedical imaging data. Methods: The Exploration Machine (XOM) is introduced as a method for computing low-dimensional representations of high-dimensional observations. XOM systematically inverts functional and structural components of topology-preserving mappings. Thus, it can contribute to both structure-preserving visualization and data clustering. We applied XOM to the analysis of microarray imaging data of gene expression profiles in Saccharomyces cerevisiae, and to model-free analysis of functional brain MRI data by unsupervised clustering. For both applications, we performed quantitative comparisons to results obtained by established algorithms. Results: Genome data: Absolute (relative) Sammon error values were 2.21 · 10³ (1.00) for XOM, 2.45 · 10³ (1.11) for Sammon's mapping, 2.77 · 10³ (1.25) for Locally Linear Embedding (LLE), 2.82 · 10³ (1.28) for PCA, 3.36 · 10³ (1.52) for Isomap, and 10.19 · 10³(4.61) for Self-Organizing Map (SOM). - Functional MRI data: Areas under ROC curves for detection of task-related brain activation were 0.984 ± 0.03 for XOM, 0.983 ± 0.02 for Minimal-Free-Energy VQ, and 0.979 ± 0.02 for SOM. Conclusion: We introduce the Exploration Machine as a novel machine learning method for the analysis of multidimensional biomedical imaging data. XOM can be successfully applied to microarray gene expression analysis and to clustering of functional brain MR image time-series. By simultaneously contributing to dimensionality reduction and data clustering, XOM is a useful novel method for data analysis in biomedical imaging.

We are developing MRI-based attenuation correction methods for PET images. PET has high sensitivity but relatively low resolution and little anatomic details. MRI can provide excellent anatomical structures with high resolution and high soft tissue contrast. MRI can be used to delineate tumor boundaries and to provide an anatomic reference for PET, thereby improving quantitation of PET data. Combined PET/MRI can offer metabolic, functional and anatomic information and thus can provide a powerful tool to study the mechanism of a variety of diseases. Accurate attenuation correction represents an essential component for the reconstruction of artifact-free, quantitative PET images. Unfortunately, the present design of hybrid PET/MRI does not offer measured attenuation correction using a transmission scan. This problem may be solved by deriving attenuation maps from corresponding anatomic MR images. Our approach combines image registration, classification, and attenuation correction in a single scheme. MR images and the preliminary reconstruction of PET data are first registered using our automatic registration method. MRI images are then classified into different tissue types using our multiscale fuzzy C-mean classification method. The voxels of classified tissue types are assigned theoretical tissue-dependent attenuation coefficients to generate attenuation correction factors. Corrected PET emission data are then reconstructed using a threedimensional filtered back projection method and an order subset expectation maximization method. Results from simulated images and phantom data demonstrated that our attenuation correction method can improve PET data quantitation and it can be particularly useful for combined PET/MRI applications.

Ischemic stroke accounts for nearly 80% of all stroke cases. Although proton diffusion and perfusion magnetic resonance imaging (MRI) are the gold standards in ischemic stroke diagnostics, the use of hyperpolarized ¹²⁹Xe MRI has a potential role to contribute to the diagnostic picture. The highly lipophilic hyperpolarized ¹²⁹Xe can be non-invasively delivered via inhalation into the lungs where it is dissolved into the blood and delivered to other organs such as the brain. As such, we expect hyperpolarized ¹²⁹Xe to act as a perfusion tracer which will result in a signal deficit in areas of blood deprived tissue. In this work, we present imaging results from an animal model of transient ischemic stroke characterized through ¹²⁹Xe MRI. In this model, a suture is used to occlude the middle cerebral artery (MCA) in the rat brain, thus causing an ischemic event. After a period of MCA occlusion, the suture can then be removed to reperfuse the ischemic area. During the ischemic phase of the stroke, a signal void was observed in the MCA territory; which was subsequently restored by normal ¹²⁹Xe MRI signal once perfusion was reinstated. Further, a higher resolution one-dimensional chemical shift image shows a sharp signal drop in the area of ischemia. Validation of ischemic damage was shown through both proton diffusion-weighted MRI (DWI) and by 2,3,5-triphenyltetrazoliumchloride (TTC) staining. The results show the potential of ¹²⁹Xe to act as a perfusion tracer; information that may add to the diagnostic and prognostic utility of the clinical picture of stroke.

Hydrocephalus occurs due to a blockage in the transmission of cerebrospinal fluid (CSF) in either the ventricles or subarachnoid space. Characteristics of this condition include increased intracranial pressure, which can result in neurologic deterioration ^[1]. Magnetic resonance elastography (MRE) is an imaging technique that estimates the mechanical properties of tissue in vivo. While some investigations of brain tissue have been performed using MRE ^[2,3,4,5], the effects due to changes in interstitial pressure and fluid content on the mechanical properties of the brain remain unknown. The purpose of this work is to assess the potential of MRE to differentiate between the reconstructed properties of normal and hydrocephalic brains. MRE data was acquired in 18 female feline subjects, 12 of which received kaolin injections resulting in an acute form of hydrocephalus. In each animal, four MRE scans were performed during the process including one pre-injection and three post-injection scans. The elastic parameters were obtained using a subzone-based reconstruction algorithm that solves Navier's equations for linearly elastic materials ^[6]. The remaining cats were used as controls, injected with saline instead of kaolin. To determine the state of hydrocephalus, ventricular volume was estimated from segmenting anatomical images. The mean ventricular volume of hydrocephalic cats significantly increased (P ⪅ 0.0001) between the first and second scans. The mean volume was not observed to increase (P ⪆ 0.5) for the control cats. Also, there was an observable increase in the recorded elastic shear modulus of brain tissue in the normal and hydrocephalic acquisitions. Results suggest that MRE is able to detect changes in the mechanical properties of brain tissue resulting from kaolin-induced hydrocephalus, indicating the need for further study.

Mouse models are becoming instrumental for the study of lung disease. Due to its resolution and low cost, high resolution Computed Tomography (micro-CT) is a very adequate technology to visualize the mouse lungs in-vivo. Automatic segmentation and measurement of airways in micro-CT images of the lungs can be useful as a preliminary step prior other image analysis quantification tasks, as well as for the study of pathologies that alter the airways structure. In this paper, we present an efficient segmentation and reconstruction algorithm which simultaneously segments and reconstructs the bronchial tree, while providing the length and mean radius of each airway segment. A locally adaptive intensity threshold is used to account for the low signal to noise ratio and strong artifacts present in micro-CT images. We validate our method by comparing it with manual segmentations of 10 different scans, obtaining an average true positive volume fraction of 85.52% with a false positive volume fraction of 5.04%.

In this paper, a new nonrigid image registration method is presented to align two volumetric lung CT datasets with an application to estimate regional ventilation. Instead of the sum of squared intensity difference (SSD), we introduce the sum of squared tissue volume difference (SSTVD) as the similarity criterion to take into account the variation of intensity due to respiration. This new criterion aims to minimize the local difference of tissue volume inside the lungs between two images scanned in the same session or over short periods of time, thus preserving the tissue weight of the lungs. Our approach is tested using a pair of volumetric lung datasets acquired at 15% and 85% of vital capacity (VC) in a single scanning session. The results show that the new SSTVD predicts a smaller registration error and also yields a better alignment of structures within the lungs than the normal SSD similarity measure. In addition, the regional ventilation derived from the new method exhibits a much more improved physiological pattern than that of SSD.

The lungs undergo expansion and contraction during the respiratory cycle. Since many disease or injury conditions are associated with the biomechanical or material property changes that can alter lung function, there is a great interest in measuring regional lung ventilation and regional mechanical changes. We describe a technique that uses multiple respiratory-gated CT images and non-rigid 3D image registration to make local estimates of lung tissue expansion. The degree of regional lung expansion is measured using the Jacobian (a function of local partial derivatives) of the registration displacement field. We compare the ventral-dorsal patterns of lung expansion estimated in both retrospectively reconstructed dynamic scans and static breath-hold scans to a xenon CT based measure of specific ventilation and a semi-automatic reference standard in four anesthetized sheep studied in the supine orientation. The regional lung expansion estimated by 3D image registration of images acquired at 50% and 75% phase points of the inspiratory portion of the respiratory cycle and 20 cm H₂O and 25 cm H₂O airway pressures gave the best match between the average Jacobian and the xenon CT specific ventilation respectively (linear regression, average r² = 0.85 and r² = 0.84). The registration accuracy assessed by 200 semi-automatically matched landmarks in both the dynamic and static scans show landmark error on the order of 2 mm.

Blood-brain-barrier (BBB) breakdown is a hypothesized mechanism for hemorrhagic transformation in acute stroke. The Patlak analysis of a Perfusion Computed Tomography (PCT) scan measures the BBB permeability, but the method yields higher estimates when applied to the first pass of the contrast bolus compared to a delayed phase. We present a numerical phantom that simulates vascular and parenchymal time-attenuation curves to determine the validity of permeability measurements obtained with different acquisition protocols. A network of tubes represents the major cerebral arteries ipsi- and contralateral to an ischemic event. These tubes branch off into smaller segments that represent capillary beds. Blood flow in the phantom is freely defined and simulated as non-Newtonian tubular flow. Diffusion of contrast in the vessels and permeation through vessel walls is part of the simulation. The phantom allows us to compare the results of a permeability measurement to the simulated vessel wall status. A Patlak analysis reliably detects areas with BBB breakdown for acquisitions of 240s duration, whereas results obtained from the first pass are biased in areas of reduced blood flow. Compensating for differences in contrast arrival times reduces this bias and gives good estimates of BBB permeability for PCT acquisitions of 90-150s duration.

While perfusion imaging is a well established diagnostic imaging technique, until now, it could not be performed using angiographic equipment. The ability to assess information about tissue perfusion in the angiographic suite should help to optimize management of patients with neurovascular diseases. We present a technique to measure cerebral blood volume (CBV) for the entire brain using an angiographic C-arm system. Combining a rotational acquisition protocol similar to that used for standard three-dimensional rotational angiography (3D DSA) in conjunction with a modified injection protocol providing a steady state of tissue contrast during the acquisition the data necessary to calculate CBV is acquired. The three-dimensional (3D) CBV maps are generated using a special reconstruction scheme which includes the automated detection of an arterial input function and several correction steps. For evaluation we compared this technique with standard perfusion CT (PCT) measurements in five healthy canines. Qualitative comparison of the CBV maps as well as quantitative comparison using 12 ROIs for each map showed a good correlation between the new technique and traditional PCT. In addition we evaluated the technique in a stroke model in canines. The presented technique provides the first step toward providing information about tissue perfusion available during the treatment of neurovascular diseases in the angiographic suite.

Knowledge of the hemodynamics in normal arterial trees of the brain is important to better understand the mechanisms responsible for the initiation and progression of cerebrovascular diseases. Information about the baseline values of hemodynamic variables such as velocity magnitudes, swirling flows, wall shear stress, pressure drops, vascular resistances, etc. is important for characterization of the normal hemodynamics and comparison with pathological states such as aneurysms and stenoses. This paper presents image-based computational hemodynamics models of cerebral arterial trees constructed from magnetic resonance angiography (MRA) images. The construction of large models of cerebral arterial trees is challenging because of the following main reasons: a) it is necessary to acquire high resolution angiographic images covering the entire brain, b) it is necessary to construct topologically correct and geometrically accurate watertight models of the vasculature, and c) the models typically result in large computational grids which make the calculations computationally demanding. This paper presents a methodology to model the hemodynamics in the brain arterial network that combines high resolution MRA at 3T, a vector representation of the vascular structures based on semi-manual segmentation, and a novel algorithm to solve the incompressible flow equations efficiently in tubular geometries. These techniques make the study of the hemodynamics in the cerebral arterial network practical.

tomographic angiography (CTA) being noninvasive, economical and informative, has become a common modality for monitoring disease status and treatment effects. Here, we present a new method for detecting and quantifying coronary arterial stenosis via CTA using fuzzy distance transform (FDT) approach. FDT computes local depth at each image point in the presence of partial voluming. Coronary arterial stenoses are detected and their severities are quantified by analyzing FDT values along the medial axis of an artery obtained by skeletonization. Also, we have developed a new skeletal pruning algorithm toward improving quality of medial axes and therefore, enhancing the accuracy of stenosis detection and quantification. The method is completed using the following steps - (1) fuzzy segmentation of coronary artery via CTA, (2) FDT computation of coronary arteries, (3) medial axis computation, (4) estimation of local diameter along arteries and (5) stenosis detection and quantification of arterial blockage. Performance of the method has been quantitatively evaluated on a realistic coronary artery phantom dataset with randomly simulated stenoses and the results are compared with a classical binary algorithm. The method has also been applied on a clinical CTA dataset from thirteen patients with 59 stenoses and the results are compared with an expert's quantitative assessment of stenoses. Results of the phantom experiment indicate that the new method is significantly more accurate as compared to the conventional binary method. Also, the results of the clinical study indicate that the computerized method is highly in agreement with the expert's assessments.

Purpose: ECG-gated abdominal CT angiography with reconstruction of multiple, temporally overlapping CT angiography datasets has been proposed for measuring aortic pulsatility. The purpose of this work is to develop algorithms to segment the aorta from surrounding structures from CTA datasets across cardiac phases, calculate registered centerlines and measurements of regional aortic pulsatility in patients with AAA, and to assess the reproducibility of these measurements. Methods: ECG-gated CTA was performed with a temporal resolution of 165 ms, reconstructed to 1 mm slices ranging at 14 cardiac phase points. Data sets were obtained from 17 patients on which two such scans were performed 6 to 12 months apart. Automated segmentation, centerline generation, and registration of centerlines between phases was performed, followed by calculation of cross-sectional areas and regional and local pulsatility. Results: Pulsatility calculations for the supraceliac region were very reproducible between earlier and later scans of the same patient, with average differences less than 1% for pulsatility values ranging from 2% to 13%. Local radial pulsatilities were also reproducible to within ~1%. Aneurysm volume changes between scans can also be quantified. Conclusion: Automated segmentation, centerline generation, and registration of temporally resolved CTA datasets permit measurements of regional changes in cross-sectional area over the course of the cardiac cycle (i.e., regional aortic pulsatility). These measurements are reproducible between scans 6-12 months apart, with differences in aortic areas reflecting both aneurysm remodeling and changes in blood pressure. Regional pulsatilities ranged from 2 to 13% but were reproducible at the 1% level.

Spine fractures are the most frequent complication of osteoporosis, a disease characterized by low bone mass and structural deterioration of bone tissue. In case of the spine, the trabecular network plays the main role in load carrying and distribution. A correct description of mechanical properties of this bone structure helps to differentiate between strong and weak bones and can be useful for fracture prediction and treatment monitoring. By means of the finite element method (FEM), applied to μCT images, we modelled biomechanical processes in probes during loading and correlated the estimated failure load with the maximum compressive strength (MCS), obtained in real biomechanical tests. We studied a sample of 151 specimens taken from the trabecular part of human vertebrae in vitro, visualised using μCT imaging at an isotropic resolution of 26μm and tested by uniaxial compression. Besides the standard way of estimating failure load, which takes into account only strong micro-fractures, we also included small micro-fractures, what improved the correlation with MCS (Pearson's correlation coefficient r=0.78 vs. r=0.58). This correlation coefficient was larger than that for both the standard morphometric parameters (r=0.73 for bone volume fraction) and for texture measures defined by the local (an-) isotropic scaling indices method (r=0.55) and Minkowski Functionals (r=0.61). However, the performance of the FEM was different for subsamples selected according to the MCS value. The correlation increased for strong specimens (r=0.88), slightly decreased for weak specimens (r=0.68) and markedly dropped for specimens with medium MCS, e.g. between 60<MCS<120, r=0.26.

Osteoporosis is bone disease which leads to low bone mass and the deterioration of the bone micro-architecture. Rarefied bone structures are more susceptible to fractures which are the worst complications of osteoporosis. Bone mineral density is considered to be the standard technique for predicting the bone strength and the effects of drug therapy. However, other properties of the bone like the trabecular structure and connectivity may also contribute. Here, we analyze μ-CT tomographic images for a sample of 151 specimens taken from human vertebrae in vitro. Using the local structural characterization of the bone trabecular network given by isotropic and anisotropic scaling indices, we generate structural decompositions of the μ-CT image and quantify the resulting patterns applying topological measures, namely the Minkowski Functionals (MF). The values of the MF are then used to assess the biomechanical properties of trabecular bone via a correlation analysis. Biomechanical properties were quantified by the maximum compressive strength calculated in an uniaxial compression test. We compare our results with those obtained using standard global histomorphometric parameters and the bone fraction BV/TV . Results obtained using structural decompositions obtained from anisotropic scaling indices were superior to those given by isotropic scaling indices. The highest correlation coefficient (r = 0.72) was better than those obtained for the standard global histomorphometric parameters and only comparable with the one given by BV/TV. Our results suggest that plate-like and dense column-like structures aligned along the direction of the external force play a relevant role for the prediction of bone strength.

Recently, micro-magnetic resonance imaging (μMRI) in conjunction with micro-finite element analysis has shown great potential in estimating mechanical properties - stiffness and elastic moduli - of bone in patients at risk of osteoporosis. Due to limited spatial resolution and signal-to-noise ratio achievable in vivo, the validity of estimated properties is often established by comparison to those derived from high-resolution micro-CT (μCT) images of cadaveric specimens. For accurate comparison of mechanical parameters derived from μMR and μCT images, analyzed 3D volumes have to be closely matched. The alignment of the micro structure (and the cortex) is often hampered by the fundamental differences of μMR and μCT images and variations in marrow content and cortical bone thickness. Here we present an intensity cross-correlation based registration algorithm coupled with segmentation for registering 3D tibial specimen images acquired by μMRI and μCT in the context of finite-element modeling to assess the bone's mechanical constants. The algorithm first generates three translational and three rotational parameters required to align segmented μMR and CT images from sub regions with high micro-structural similarities. These transformation parameters are then used to register the grayscale μMR and μCT images, which include both the cortex and trabecular bone. The intensity crosscorrelation maximization based registration algorithm described here is suitable for 3D rigid-body image registration applications where through-plane rotations are known to be relatively small. The close alignment of the resulting images is demonstrated quantitatively based on a voxel-overlap measure and qualitatively using visual inspection of the micro structure.

High-frequency (> 20 MHz) ultrasound images of preclinical tumor models are sensitive to changes in tissue microstructure that accompany tumor progression and treatment responses, but the relationships between tumor microanatomy and high-frequency ultrasound backscattering are incompletely understood. Computational models of tissue microstructure can be employed with ultrasound propagation simulators to investigate these relationships. This paper introduces a three-dimensional microanatomical model in which tissue is treated as a population of stochastically positioned spherical cells embedded in a homogeneous extracellular matrix, where each cell consists of a spherical nucleus surrounded by homogeneous cytoplasm. The model is used to represent the microstructure of both healthy mouse liver and experimental liver metastasis. Normal and cancerous tissue specimens stained with DAPI and H&E are digitized at 20× magnification and analyzed to specify values of the model parameters. Simulated healthy and tumor tissues are initialized based on the ratio of cell to nucleus diameter and the nuclear volume fraction and size distribution estimated by stereological analysis of the normal and cancerous liver specimens, respectively. For each simulated tissue, the spatial organization of cells is controlled by a Gibbs-Markov point process. The parameters of the Gibbs-Markov process are tuned to accurately reproduce the number density and distribution of center-to-center spacing of nuclei in the DAPI-stained slides of the corresponding experimental tissue specimen. The morphological variations that can be produced by changing the model parameters are expected to be sufficient to represent the microstructural changes during tumor progression that are the most significant determinants of high-frequency ultrasound backscattering.

As a non-invasive bladder tumor screening approach, magnetic resonance imaging (MRI)-based virtual cystoscopy (VCys) has received increasing attention for a better soft tissue contrast compared to computer tomography (CT)-based VCys. In this paper, some preliminary work on segmenting the inner boundary of bladder wall from both T1- and T2- weighted MR bladder images were presented. Via an iterative maximum a posteriori expectation-maximization (MAPEM) approach, the tissue mixture fractions inside each voxel were estimated. Considering the partial volume effect (PVE) that MR images suffer from, the advantages of such mixture-based segmentation approach are (1) statistics-based tissue mixture model that shapes each tissue type as a normal-distributed random variable, (2) closed-form mathematical MAP-EM iterative solution, and (3) capability and efficiency of the estimated tissue mixture fractions in reflecting PVE. Given the extracted inner bladder wall, manipulations could be further taken, for each individual voxel located on the inner bladder wall, to identify the outer bladder wall prior to the measurement of wall thickness. Not limited to geometrical analysis, the consideration of PVE in the study of early stage abnormality on the mucosa in the scope of VCys is believed to provide more textural information in distinguishing from neighboring artifacts about the surface deformations that is due to bladder tumors.

As a new mode of molecular imaging, bioluminescence tomography (BLT) will have significant effect on revealing the molecular and cellular information in vivo at the whole-body small animal level because of its high sensitive detection and facile operation. However, BLT is an ill-posed problem, it is necessary to incorporate a priori knowledge into the tomographic algorithm. In this paper, a novel Bayesian reconstruction algorithm for BLT is firstly proposed. In the algorithm, a priori permissible source region strategy is incorporated into the Bayesian network to reduce the ill-posedness of BLT. Then a generalized adaptive Gaussian Markov random field (GAGMRF) prior model for unknown source density estimation is developed to further reduce the ill-posedness of BLT on the basis of adaptive finite element analysis. Finally, the algorithm maximizes the log posterior probability with respect to a noise parameter and the unknown source density, the distribution of bioluminescent source can be reconstructed. In addition, the novel tomography algorithm based adaptive finite element makes the method more appropriate for complex phantom such as real mouse. In the numerical simulation, a heterogeneous phantom is used to evaluate the performance of the proposed algorithm with the Monte Carlo based synthetic data. The accurate localization of bioluminescent source and quantitative results show the effectiveness and potential of the tomographic algorithm for BLT.

The observed motion of subcellular particles in fluorescence microscopy image sequences of live cells is generally a superposition of the motion and deformation of the cell and the motion of the particles. Decoupling the two types of movements to enable accurate classification of the particle motion requires the application of registration algorithms. We have developed an intensity-based approach which is based on an optic-flow estimation algorithm for non-rigid registration of multi-channel microscopy image sequences of cell nuclei. First, based on 3D synthetic images we demonstrate that cell nucleus deformations change the observed motion types of particles and that our approach allows to recover the original motion. Second, we have successfully applied our approach to register 2D and 3D real microscopy image sequences. A quantitative comparison with a previous scheme has also been performed.

As a promising optical molecular imaging modality, bioluminescence tomography (BLT) has attracted remarkable attention for its excellent performance and high cost-effectiveness, which can be employed to specifically and directly reveal physiological and pathological activities in vivo at molecular and cellular levels. The goal of BLT is to reconstruct the internal bioluminescent light source with surface measurements. Therefore, the calculation of surface light exitance plays an important role in the inverse source reconstruction, whereas photon propagation is complicated because of strongly scattering property of the biological tissue. In this contribution, a novel meshless local Petrov-Galerkin (MLPG) method based on diffusion approximation model is developed to avoid the complex and time-consuming mesh division in the conventional finite element method (FEM), and MLPG requires only a series of discretized nodes without consideration of element information and node connectivity. Compared with other meshless methods based on global weak-form, background cells used for Gauss quadrature are also omitted in the proposed method. In addition, the tissue optical parameters are incorporated as a priori knowledge in this algorithm. Finally, the performance of this method is valuated using two- and three-dimensional numerical simulation experiments. The results demonstrate the effectiveness and feasibility of the presented algorithm to predict boundary bioluminescent light power distribution.

The optic disc margin is of interest due to its use for detecting and managing glaucoma. We developed a method for segmenting the optic disc margin of the optic nerve head (ONH) in spectral-domain optical coherence tomography (OCT) images using a graph-theoretic approach. A small number of slices surrounding the Bruch's membrane opening (BMO) plane was taken and used for creating planar 2-D projection images. An edge-based cost function - more specifically, a signed edge-based term favoring a dark-to-bright transition in the vertical direction of polar projection images (corresponding to the radial direction in Cartesian coordinates) - was obtained. Information from the segmented vessels was used to suppress the vasculature influence by modifying the polar cost function and remedy the segmentation difficulty due to the presence of large vessels. The graph search was performed in the modified edge-based cost images. The algorithm was tested on 22 volumetric OCT scans. The segmentation results were compared with expert segmentations on corresponding stereo fundus disc photographs. We found a signed mean difference of 0.0058 ± 0.0706 mm and an unsigned mean difference of 0.1083 ± 0.0350 mm between the automatic and expert segmentations.

The optical projection tomography microscope (OPTM) is an optical microscope that acquires focus-invariant images from multiple views of single cells. Although the depth of field of the objective is short, it can be extended by scanning the objective's focal plane. This extended depth of field image is similar to a projection in conventional X-ray CT. Samples flow through a microcapillary tube filled with optical gel. Optical distortion is minimized by matching refractive index of optical gel and tube. Multiple projection images are taken by rotating the microcapillary tube with sub-micron mechanical precision. After these pseudoprojection images are further aligned, computed tomography methods are then applied to the images to create a 3D reconstruction with isometric resolution of 0.35 microns. Three-dimensional reconstructed images of fluorescent microspheres and cells are shown.

We developed a cryo-imaging system to provide single-cell detection of fluorescently labeled cells in mouse, with particular applicability to stem cells and metastatic cancer. The Case cryoimaging system consists of a fluorescence microscope, robotic imaging positioner, customized cryostat, PC-based control system, and visualization/analysis software. The system alternates between sectioning (10-40 μm) and imaging, collecting color brightfield and fluorescent blockface image volumes >60GB. In mouse experiments, we imaged quantum-dot labeled stem cells, GFP-labeled cancer and stem cells, and cell-size fluorescent microspheres. To remove subsurface fluorescence, we used a simplified model of light-tissue interaction whereby the next image was scaled, blurred, and subtracted from the current image. We estimated scaling and blurring parameters by minimizing entropy of subtracted images. Tissue specific attenuation parameters were found [uT : heart (267 ± 47.6 μm), liver (218 ± 27.1 μm), brain (161 ± 27.4 μm)] to be within the range of estimates in the literature. "Next image" processing removed subsurface fluorescence equally well across multiple tissues (brain, kidney, liver, adipose tissue, etc.), and analysis of 200 microsphere images in the brain gave 97±2% reduction of subsurface fluorescence. Fluorescent signals were determined to arise from single cells based upon geometric and integrated intensity measurements. Next image processing greatly improved axial resolution, enabled high quality 3D volume renderings, and improved enumeration of single cells with connected component analysis by up to 24%. Analysis of image volumes identified metastatic cancer sites, found homing of stem cells to injury sites, and showed microsphere distribution correlated with blood flow patterns. We developed and evaluated cryo-imaging to provide single-cell detection of fluorescently labeled cells in mouse. Our cryo-imaging system provides extreme (>60GB), micron-scale, fluorescence, and bright field image data. Here we describe our image preprocessing, analysis, and visualization techniques. Processing improves axial resolution, reduces subsurface fluorescence by 97%, and enables single cell detection and counting. High quality 3D volume renderings enable us to evaluate cell distribution patterns. Applications include the myriad of biomedical experiments using fluorescent reporter gene and exogenous fluorophore labeling of cells in applications such as stem cell regenerative medicine, cancer, tissue engineering, etc.

We have developed a rat model of single, sub-lethal thoracic irradiation. Our irradiation protocol is considered representative of exposures near the detonation site of a dirty bomb or small nuclear device. The model is being used to investigate techniques for identifying, triaging and treating possible victims. In addition to physiological markers of right ventricular hypertrophy, pulmonary vascular resistance, and arterial distensibility, we present two methods for quantifying microvascular density. We used methods including microfocal X-ray imaging to investigate changes in lung structure/function resulting from radiation exposure. Radiation pneumonitis is a complication in subjects receiving thoracic irradiation. A radiographic hallmark of acute radiation pneumonitis is a diffuse infiltrate corresponding to the radiation treatment field. We describe two methods for quantifying small artery dropout that occurs in the model at the same time-period. Rats were examined 3-days, 2-weeks, 1-month (m), 2-m, 5-m, and 12-m post-irradiation and compared with aged-matched controls. Right ventricular hypertrophy and increases in pulmonary vascular resistance were present during the pneumonitis phase. Vascular injury was dependent on dose and post-irradiation duration. Rats irradiated with 5 Gy had few detectable changes, whereas 10 Gy resulted in a significant decrease in both microvascular density and arterial distensibility around 2- m, the decrease in each lessening, but extending through 12-m. In conclusion, rats irradiated with a 10 Gy dose had changes in vascular structure concurrent with the onset of radiation pneumonitis that were detectable with our imaging techniques and these structural changes persist after resolution of the pneumonitis.

Biophotonic imaging is a novel, relatively low-cost method for in-vivo imaging of tumors in mouse models. This technique, utilizing luminescent cancer cells, can improve productivity for cancer investigators and reduce the number of mice needed to conduct an experiment by allowing longitudinal studies. However, many of the tools provided with these systems are intended for interactive use and are time consuming to use when large numbers of images are captured. Many studies require a specific determination of the location and tumor size, particularly relative to the anatomical details of the mouse; whether this is the entire mouse body, single organs, or custom, user defined regions. An automated method of registering mouse images to a whole body atlas mask with well defined anatomical details is presented. Bilinear scaling is used within the registration process and is shown to be successful since the trapezoidal shape chosen merges well with the natural shape of the mouse. After successful registration, quantification of the photon flux can be performed for the whole body and specific regions using a summation of intensity levels and photon flux per intensity level. Registration accuracy rates over 90% were achieved although results vary relative to the positioning of the mouse. This work provides a base to explore 3D and temporal registration techniques for such data sets.

Patients with chronic calorie insufficiency commonly suffer from upper gastrointestinal dysfunction and consequent dyspeptic symptoms, which may interfere with their nutritional rehabilitation. To investigate the relationship between gastric dysfunction and feeding behavior, we exposed mice to chronic caloric restriction and demonstrated gastric motor abnormalities in them. Gastric dysmotility is typically associated with dyspeptic symptoms but sensations cannot be directly assessed in animal models. Therefore, as an initial step toward establishing measurable correlates of postprandial symptoms in small animals, we have attempted to characterize central responses to food intake by positron emission tomography-computerized microtomography (PET-CT) in normal and calorically restricted mice. Animals consumed a standard test meal after an overnight fast before receiving 2-deoxy-2[¹⁸F]fluoro-D-glucose tracer. The same mice were also scanned in the fasting state on a separate day. We were able to bring the fed and fasting PET volume images into spatial registration with each other and with an MR-derived atlas of the mouse brain, so that the differences in uptake between the two states could be mapped quantitatively against the neuroanatomic regions of the atlas. Our approach is suitable for studying the effects of gastric dysmotilities on central responses to feeding.

Colon cancer is a leading cause of death in the US, even though many cases are preventable if tumors are detected early. One technique to promote screening is Computed Tomography Colonography (CTC). NM404 is a second generation phospholipid ether analogue which has demonstrated selective uptake and prolonged retention in 43/43 types of malignant tumors but not inflammatory sites or premalignant lesions. The purpose of this experiment was to evaluate (SWR x B6 )F1.Min mice as a preclinical model to test MicroPET/CT dual modality virtual colonoscopy. Each animal was given an IV injection of ¹²⁴I-NM404 (100 uCi) 24, 48 and 96 hours prior to scanning on a dedicated microPET/CT system. Forty million counts were histogrammed in 3D and reconstructed using an OSEM 2D algorithm. Immediately after PET acquisition, a 93 m volumetric CT was acquired at 80 kVp, 800 uA and 350 ms exposures. Following CT, the mouse was sacrificed. The entire intestinal tract was excised, washed, insufflated, and scanned ex vivo A total of eight tissue samples from the small intestine were harvested: 5 were benign adenomas, 2 were malignant adenocarcinomas, and 1 was a Peyer's patch (lymph tissue) . The sites of these samples were positioned on CT and PET images based on morphological cues and the distance from the anus. Only 1/8 samples showed tracer uptake. several hot spots in the microPET image were not chosen for histology. (SWR x B6)F1.Min mice develop benign and malignant tumors, making this animal model a strong candidate for future dual modality microPET/CT virtual colonography studies.

We are developing and evaluating choline molecular imaging with positron emission tomography (PET) for monitoring tumor response to photodynamic therapy (PDT) in animal models. Human prostate cancer (PC-3) was studied in athymic nude mice. A second-generation photosensitizer Pc 4 was used for PDT in tumor-bearing mice. MicroPET images with ¹¹C-choline were acquired before PDT and 48 h after PDT. Time-activity curves of ¹¹C-choline uptake were analyzed before and after PDT. For treated tumors, normalized choline uptake decreased significantly 48 h after PDT, compared to the same tumors pre-PDT (p ⪅ 0.001). However, for the control tumors, normalized choline uptake increased significantly (p ⪅ 0.001). PET imaging with ¹¹C-choline is sensitive to detect early tumor response to PDT in the animal model of human prostate cancer.

Understanding the geography of genetic expression in the mouse brain has opened previously unexplored avenues in neuroinformatics. The Allen Brain Atlas (www.brain-map.org) (ABA) provides genome-wide colorimetric in situ hybridization (ISH) gene expression images at high spatial resolution, all mapped to a common three-dimensional 200μm³ spatial framework defined by the Allen Reference Atlas (ARA) and is a unique data set for studying expression based structural and functional organization of the brain. The goal of this study was to facilitate an unbiased data-driven structural partitioning of the major structures in the mouse brain. We have developed an algorithm that uses nonnegative matrix factorization (NMF) to perform parts based analysis of ISH gene expression images. The standard NMF approach and its variants are limited in their ability to flexibly integrate prior knowledge, in the context of spatial data. In this paper, we introduce spatial connectivity as an additional regularization in NMF decomposition via the use of Markov Random Fields (mNMF). The mNMF algorithm alternates neighborhood updates with iterations of the standard NMF algorithm to exploit spatial correlations in the data. We present the algorithm and show the sub-divisions of hippocampus and somatosensory-cortex obtained via this approach. The results are compared with established neuroanatomic knowledge. We also highlight novel gene expression based sub divisions of the hippocampus identified by using the mNMF algorithm.

Registration of MRI to histopathological sections can enhance bioimaging validation for use in pathobiologic, diagnostic, and therapeutic evaluations. However, commonly used registration methods fall short of this goal due to tissue shrinkage and tearing after brain extraction and preparation. In attempts to overcome these limitations we developed a software toolbox using 3D blockface imaging as the common space of reference. This toolbox includes a semi-automatic brain extraction technique using constraint level sets (CLS), 3D reconstruction methods for the blockface and MR volume, and a 2D warping technique using thin-plate splines with landmark optimization. Using this toolbox, the rodent brain volume is first extracted from the whole head MRI using CLS. The blockface volume is reconstructed followed by 3D brain MRI registration to the blockface volume to correct the global deformations due to brain extraction and fixation. Finally, registered MRI and histological slices are warped to corresponding blockface images to correct slice specific deformations. The CLS brain extraction technique was validated by comparing manual results showing 94% overlap. The image warping technique was validated by calculating target registration error (TRE). Results showed a registration accuracy of a TRE < 1 pixel. Lastly, the registration method and the software tools developed were used to validate cell migration in murine human immunodeficiency virus type one encephalitis.

The efficiency of constructing an active appearance model (AAM) is limited by establishing the independent standard via time-consuming and tedious manual tracing. It is more challenging for 3D and 4D (3D+time) datasets as the smoothness of shape and motion is essential. In this paper, a three-stage pipeline is designed for efficient cross-modality model construction. It utilizes existing AAM and active shape model (ASM) of the left ventricle (LV) for magnetic resonance (MR) datasets to build 4D AAM of the LV for real-time 3D echocardiography (RT3DE) datasets. The first AAM fitting stage uses AAM for MR to fit valid shapes onto the intensity-transformed RT3DE data that resemble low-quality MR data. The fitting is implemented in a 3D phase-by-phase fashion to prevent introducing bias due to different motion patterns related to the two modalities and patient groups. The second global-scale editing stage adjusts fitted shapes by tuning modes of ASM for MR data. The third local-scale editing stage adjusts the fitted volumes at small local regions and produces the final accurate independent standard. By visual inspection, the AAM fitting stage successfully produces results that capture the LV motion - especially its base movement - within the cardiac cycle on 29 of the 32 RT3DE datasets tested. This multi-stage approach can reduce the human effort of the manual tracing by at least 50%. With the model built for a modality A available, this approach is generalizable to constructing the model of the same organ for any other modality B.

A realistic 3D coronary arterial tree (CAT) has been developed for the heart model of the computer generated 3D XCAT phantom. The CAT allows generation of a realistic model of the location, size and shape of the associated regional ischemia or infarction for a given coronary arterial stenosis or occlusion. This in turn can be used in medical imaging applications. An iterative rule-based generation method that systematically utilized anatomic, morphometric and physiologic knowledge was used to construct a detailed realistic 3D model of the CAT in the XCAT phantom. The anatomic details of the myocardial surfaces and large coronary arterial vessel segments were first extracted from cardiac CT images of a normal patient with right coronary dominance. Morphometric information derived from porcine data from the literature, after being adjusted by scaling laws, provided statistically nominal diameters, lengths, and connectivity probabilities of the generated coronary arterial segments in modeling the CAT of an average human. The largest six orders of the CAT were generated based on the physiologic constraints defined in the coronary generation algorithms. When combined with the heart model of the XCAT phantom, the realistic CAT provides a unique simulation tool for the generation of realistic regional myocardial ischemia and infraction. Together with the existing heart model, the new CAT provides an important improvement over the current 3D XCAT phantom in providing a more realistic model of the normal heart and the potential to simulate myocardial diseases in evaluation of medical imaging instrumentation, image reconstruction, and data processing methods.

Retinal vessel segmentation is a prerequisite for the analysis of vessel parameters such as tortuosity, variation of the vessel width along the vessel and the ratio between the venous and arterial vessel width. This analysis can provide indicators for the presence of a wide range of diseases. Different types of approaches have been proposed to segment the retinal vasculature and two important groups are vessel tracking and pixel processing based methods. An advantage of tracking based methods is the guaranteed connectedness of vessel segments, in pixel processing based methods connectedness is not guaranteed. In this work an automated vessel linking framework is presented. The framework links together separate pieces of the retinal vasculature into a connected vascular tree. To determine which vessel sections should be linked together the use of a supervised cost function is proposed. Evaluation is performed on the vessel centerlines. The results show that the vessel linking framework outperforms other automated vessel linking methods especially for the narrowest vessels.

A number of studies have documented that autism has a neurobiological basis, but the anatomical extent of these neurobiological abnormalities is largely unknown. In this study, we aimed at analyzing highly localized shape abnormalities of the corpus callosum in a homogeneous group of autism children. Thirty patients with essential autism and twenty-four controls participated in this study. 2D contours of the corpus callosum were extracted from MR images by a semiautomatic segmentation method, and the 3D model was constructed by stacking the contours. The resulting 3D model had two openings at the ends, thus a new conformal parameterization for high genus surfaces was applied in our shape analysis work, which mapped each surface onto a planar domain. Surface matching among different individual meshes was achieved by re-triangulating each mesh according to a template surface. Statistical shape analysis was used to compare the 3D shapes point by point between patients with autism and their controls. The results revealed significant abnormalities in the anterior most and anterior body in essential autism group.

Probabilistic maps are useful in functional neuroimaging research for anatomical labeling and for data analysis. The degree to which a probability map can accurately estimate the location of the structure of interest in a new individual depends on many factors, including the variability in the morphology of the structure of interest over subjects, the registration (normalization procedure and template) applied to align the brains among individuals and the registration used to map a new subject's dataset to the frame of the probabilistic map. Here, we take Heschl's gyrus (HG) as our structure of interest, and explore the impact of different registration methods on the accuracy with which a probabilistic map of HG can approximate HG in a new individual. We compare three registration procedures; high-dimensional (HAMMER); template-free B-spline-based groupwise; and segmentation-based (SPM5); to each other and to a previously published (affine) probabilistic map of HG.1 We quantitatively evaluate the accuracy of the resulting maps using evidence-based diagnostic measures within a leave-one-out cross-validation structure, to demonstrate that maps created using either HAMMER or SPM5 have relatively high sensitivity, specificity and positive predictive value, compared to a map created using the groupwise algorithm or compared to the published map.

Breast cancer is one of the most common causes of cancer death among women around the world. Researchers have found that a combination of imaging modalities (such as x-ray mammography, magnetic resonance, and ultrasound) leads to more effective diagnosis and management of breast cancers because each imaging modality displays different information about the breast tissues. In order to aid clinicians in interpreting the breast images from different modalities, we have developed a computational framework for generating individual-specific, 3D, finite element (FE) models of the breast. Medical images are embedded into this model, which is subsequently used to simulate the large deformations that the breasts undergo during different imaging procedures, thus warping the medical images to the deformed views of the breast in the different modalities. In this way, medical images of the breast taken in different geometric configurations (compression, gravity, etc.) can be aligned according to physically feasible transformations. In order to analyse the accuracy of the biomechanical model predictions, squared normalised cross correlation (NCC²) was used to provide both local and global comparisons of the model-warped images with clinical images of the breast subject to different gravity loaded states. The local comparison results were helpful in indicating the areas for improvement in the biomechanical model. To improve the modelling accuracy, we will need to investigate the incorporation of breast tissue heterogeneity into the model and altering the boundary conditions for the breast model. A biomechanical image registration tool of this kind will help radiologists to provide more reliable diagnosis and localisation of breast cancer.

Most conventional methods of breast cancer diagnosis such as X-ray, Ultrasound (US) and MRI have some issues ranging from weaknesses associated with tumour detection or classification to high cost. In this study, we propose a breast elastography technique based on 3D US. This technique is fast, expected to be cost effective and more sensitive and specific compared to US imaging. Unlike current elastography techniques that image relative elastic modulus, this technique is capable of imaging absolute Young's modulus (YM). In this technique, tissue displacements and surface forces used to mechanically stimulate the tissue are acquired and used as input to reconstruct the tissue YM distribution. For the displacements acquisition, we use a modified optical flow technique to estimate the displacement of each node from 3D US pre- and post-compression images. A force sensor is used to measure forces on the surface of the breast. These forces are input into an analytical model to estimate tissue stress distribution. By combining the stress field with the strain field calculated from the estimated displacements using Hooke's law, the YM can be reconstructed efficiently. Also, we adapted a micromechanics based model developed for strain distribution estimation in heterogeneous medium to update the reconstructed YM value of tumor more accurately.

A novel optical flow based technique is presented in this paper to measure the nodal displacements of soft tissue undergoing large deformations. In hyperelasticity imaging, soft tissues maybe compressed extensively [1] and the deformation may exceed the number of pixels ordinary optical flow approaches can detect. Furthermore in most biomedical applications there is a large amount of image information that represent the geometry of the tissue and the number of tissue types present in the organ of interest. Such information is often ignored in applications such as image registration. In this work we incorporate the information pertaining to soft tissue mechanical behavior (Neo-Hookean hyperelastic model is used here) in addition to the tissue geometry before compression into a hierarchical Horn-Schunck optical flow method to overcome this large deformation detection weakness. Applying the proposed method to a phantom using several compression levels proved that it yields reasonably accurate displacement fields. Estimated displacement results of this phantom study obtained for displacement fields of 85 pixels/frame and 127 pixels/frame are reported and discussed in this paper.

Understanding the complex relationship between the thermal contrasts on the breast surface and the underlying physiological and pathological factors is important for thermogram-based breast cancer detection. Our previous work introduced a combined thermal-elastic modeling method with improved ability to simultaneously characterize both elastic-deformation-induced and tumor-induced thermal contrasts on the breast. In this paper, the technique is further extended to investigate the dynamic behaviors of the breast thermal contrasts during cold stress and thermal recovery procedures in the practice of dynamic thermal imaging. A finite-element method (FEM) has been developed for dynamic thermal and elastic modeling. It is combined with a technique to address the nonlinear elasticity of breast tissues, as would arise in the large deformations caused by gravity. Our simulation results indicate that different sources of the thermal contrasts, such as the presence of a tumor, and elastic deformation, have different transient time courses in dynamic thermal imaging with cold-stress and thermal-recovery. Using appropriate quantifications of the thermal contrasts, we find that the tumor- and deformation-induced thermal contrasts show opposite changes in the initial period of the dynamic courses, whereas the global maxima of the contrast curves are reached at different time points during a cold-stress or thermal-recovery procedure. Moreover, deeper tumors generally lead to smaller peaks but have larger lags in the thermal contrast time course. These findings suggest that dynamic thermal imaging could be useful to differentiate the sources of the thermal contrast on breast surface and hence to enhance tumor detectability.

We present a model for soft tissue deformation derived empirically from 10 pig carcases. The carcasses are subjected to deformation from a known single source of pressure located at the skin surface, and the deformation is quantified by means of steel markers injected into the tissue. The steel markers are easy to distinguish from the surrounding soft tissue in 3D computed tomography images. By tracking corresponding markers using methods from point-based registration, we are able to accurately quantify the magnitude and propagation of the induced deformation. The deformation is parameterised by radial basis functions with compact support. The parameterisation yields an absolute error with mean 0.20 mm, median 0.13 mm and standard deviation 0.21 mm (not cross validated). We use the parameterisation to form a statistical deformation model applying principal component analysis on the estimated deformation parameters. The model is successfully validated using leave-one-out cross validation by subject, achieving a sufficient level of precision using only the first two principal modes; mean 1.22 mm, median 1.11 mm and standard deviation 0.67 mm.

Implementing constitutive relations that accurately describe the mechanical behavior of biological tissues in vivo is integral to the success of any model-based elastographic reconstruction technique, and the diagnostic value of the recovered images. Recently, poroelastic theory has been used to model tissue and other materials comprised of two distinct phases. Current linearly elastic techniques are not capable of fully describing the complex mechanical behavior of fluid-saturated tissues because they consider only a single solid phase, neglecting the influence of extracellular fluid. In an attempt to model the deformation of biological tissues more effectively in vivo by employing constitutive relations which are more representative of tissue structure and physiology, a three-dimensional (3D) finite element reconstruction algorithm has been developed based on the equations of dynamic poroelasticity. The algorithm operates on a single domain of O(10³) nodes. The performance of the algorithm was tested using simulated data. The results suggest that the technique is capable of recovering accurate distributions of the underlying mechanical properties of the solid matrix as well as the time-harmonic pressure field resulting from tissue vibration.

Cardiovascular disease results from pathological biomechanical conditions and fatigue of the vessel wall. Image-based computational modeling provides a physical and realistic insight into the patient-specific biomechanics and enables accurate predictive simulations of development, growth and failure of cardiovascular disease. An experimental validation is necessary for the evaluation and the clinical implementation of such computational models. In the present study, we have implemented dynamic Computed-Tomography (4D-CT) imaging and catheter-based in vivo measured pressures to numerically simulate and experimentally evaluate the biomechanics of the porcine aorta. The computations are based on the Finite Element Method (FEM) and simulate the arterial wall response to the transient pressure-based boundary condition. They are evaluated by comparing the numerically predicted wall deformation and that calculated from the acquired 4D-CT data. The dynamic motion of the vessel is quantified by means of the hydraulic diameter, analyzing sequences at 5% increments over the cardiac cycle. Our results show that accurate biomechanical modeling is possible using FEM-based simulations. The RMS error of the computed hydraulic diameter at five cross-sections of the aorta was 0.188, 0.252, 0.280, 0.237 and 0.204 mm, which is equivalent to 1.7%, 2.3%, 2.7%, 2.3% and 2.0%, respectively, when expressed as a function of the time-averaged hydraulic diameter measured from the CT images. The present investigation is a first attempt to simulate and validate vessel deformation based on realistic morphological data and boundary conditions. An experimentally validated system would help in evaluating individual therapies and optimal treatment strategies in the field of minimally invasive endovascular surgery.

Currently there is increased interest in the use of stents as flow diverters for the treatment of intracranial aneurysms, especially wide necked aneurysms that are difficult to treat by coil embolization or surgical clipping. This paper presents image-based patient-specific computational models of the hemodynamics in cerebral aneurysms before and after treatment with a stent alone, with the goal of better understanding the hemodynamic effects of these devices and their relation to the outcome of the procedures. Stenting of cerebral aneurysms is a feasible endovascular treatment option for aneurysms with wide necks that are difficult to treat with coils or by surgical clipping. However, this requires stents that are capable of substantially modifying the intra-aneurysmal flow pattern in order to cause thrombosis of the aneurysm. The results presented in this paper show that the studied stent was able to change significantly the hemodynamic characteristics of the aneurysm. In addition, it was shown that patient-specific computational models constructed from medical images are capable of realistically representing the in vivo hemodynamic characteristics observed during conventional angiography examinations before and after stenting. This indicates that these models can be used to better understand the effects of different stent designs and to predict the alteration in the hemodynamic pattern of a given aneurysm produced by a given flow diverter. This is important for improving current design of flow diverting devices and patient treatment plans.

Image-guided endovascular intervention (EIGI), using new flow modifying endovascular devices for intracranial aneurysm treatment is an active area of stroke research. The new polyurethane-asymmetric vascular stent (P-AVS), a vascular stent partially covered with a polyurethane-based patch, is used to cover the aneurysm neck, thus occluding flow into the aneurysm. This study involves angiographic imaging of partially covered aneurysm orifices. This particular situation could occur when the vascular geometry does not allow full aneurysm coverage. Four standard in-vivo rabbit-model aneurysms were investigated; two had stent patches placed over the distal region of the aneurysm orifice while the other two had stent patches placed over the proximal region of the aneurysm orifice. Angiographic analysis was used to evaluate aneurysm blood flow before and immediately after stenting and at four-week follow-up. The treatment results were also evaluated using histology on the aneurysm dome and electron microscopy on the aneurysm neck. Post-stenting angiographic flow analysis revealed aneurysmal flow reduction in all cases with faster flow in the distally-covered case and very slow flow and prolonged pooling for proximal-coverage. At follow-up, proximally-covered aneurysms showed full dome occlusion. The electron microscopy showed a remnant neck in both distally-placed stent cases but complete coverage in the proximally-placed stent cases. Thus, direct flow (impingement jet) removal from the aneurysm dome, as indicated by angiograms in the proximally-covered case, was sufficient to cause full aneurysm healing in four weeks; however, aneurysm healing was not complete for the distally-covered case. These results support further investigations into the treatment of aneurysms by flow-modification using partial aneurysm-orifice coverage.

Elastography is an emerging imaging modality that can visualize breast tumors via their tissue elasticity and strain properties. Computing strain elastograms by taking the first-order derivative of axial displacement images has a limitation, however; this method will amplify displacement measurement errors that, in turn, will degrade the quality of the strain elastograms. To overcome this limitation, we describe a novel spatial-filtering approach that involves fitting dimensionally increasing, multi-resolution radial basis functions (RBFs) to the displacement field. The results of experiments conducted on an elastically heterogeneous phantom revealed three important observations. Firstly, excessive filtering occurred when spatial filtering was performed using a firstorder radial basis function, to the extent that it was not always possible to discern high-contrast elastic inclusions in the resulting strain elastogram. Secondly, artifacts were incurred when spatial filtering was performed using the high-resolution first-order basis function. Thirdly, second-order basis functions improved our ability to discern lesions in strain elastograms, but this was highly dependent on the resolution of the basis function we employed. More specifically, we achieved better visibility using the low-resolution second-order basis functions, and lower visibility using the high-resolution basis functions. It was concluded that the radial-basis-function strain-estimation method performed sufficiently well to warrant more in-depth studies.

A novel cardiac chamber volume model is proposed for the assessment of left ventricular mechanical dyssynchrony. The tool is potentially useful for assessment of regional cardiac function and identification of mechanical dyssynchrony on MRI. Dyssynchrony results typically from a contraction delay between one or more individual left ventricular segments, which in turn leads to inefficient ventricular function and ultimately heart failure. Cardiac resynchronization therapy has emerged as an electrical treatment of choice for heart failure patients with dyssynchrony. Prior MRI techniques have relied on assessments of actual cardiac wall changes either using standard cine MR images or specialized pulse sequences. In this abstract, we detail a semi-automated method that evaluates dyssynchrony based on segmental volumetric analysis of the left ventricular (LV) chamber as illustrated on standard cine MR images. Twelve sectors each were chosen for the basal and mid-ventricular slices and 8 sectors were chosen for apical slices for a total of 32 sectors. For each slice (i.e. basal, mid and apical), a systolic dyssynchrony index (SDI) was measured. SDI, a parameter used for 3D echocardiographic analysis of dyssynchrony, was defined as the corrected standard deviation of the time at which minimal volume is reached in each sector. The SDI measurement of a healthy volunteer was 3.54%. In a patient with acute myocardial infarction, the SDI measurements 10.98%, 16.57% and 1.41% for basal, mid-ventricular and apical LV slices, respectively. Based on published 3D echocardiogram reference threshold values, the patient's SDI corresponds to moderate basal dysfunction, severe mid-ventricular dysfunction, and normal apical LV function, which were confirmed on echocardiography. The LV chamber segmental volume analysis model and SDI is feasible using standard cine MR data and may provide more reliable assessment of patients with dyssynchrony especially if the LV myocardium is thin or if the MR images have spatial resolution insufficient for proper resolution of wall thickness-features problematic for dyssynchrony assessment using existing MR techniques.

A major field in cognitive neuroscience investigates neuronal correlates of human decision-making processes [1, 2]. Is it possible to predict a decision before it is actually revealed by the volunteer? In the presented manuscript we use a standard paradigm from economic behavioral research that proved emotional influences on human decision making: the Ultimatum Game (UG). In the UG, two players have the opportunity to split a sum of money. One player is deemed the proposer and the other, the responder. The proposer makes an offer as to how this money should be split between the two. The second player can either accept or reject this offer. If it is accepted, the money is split as proposed. If rejected, then neither player receives anything. In the presented study a real-time fMRI system was used to derive the brain activation of the responder. Using a Relevance-Vector-Machine classifier it was possible to predict if the responder will accept or reject an offer. The classification result was presented to the operator 1-2 seconds before the volunteer pressed a button to convey his decision. The classification accuracy reached about 70% averaged over six subjects.

Foot ulcers affect millions of Americans annually. Areas that are likely to ulcerate have been associated with increased local skin temperatures due to inflammation and enzymatic autolysis of tissue. Conventional methods to assess skin, including inspection and palpation, may be valuable approaches, but usually they do not detect changes in skin integrity until an ulcer has already developed. Conversely, infrared imaging is a technology able to assess the integrity of the skin and its many layers, thus having the potential to index the cascade of physiological events in the prevention, assessment, and management of foot ulcers. In this paper, we propose a technique, asymmetry analysis, to automatically analyze the infrared images in order to detect inflammation. Preliminary results show that the proposed technique can be reliable and efficient to detect inflammation and, hence, predict potential ulceration.

We aim to capture and apply 3-dimensional bone fragility features for fracture risk estimation. Using inter-subject image registration, we constructed a hip QCT atlas comprising 37 patients with hip fractures and 38 age-matched controls. In the hip atlas space, we performed principal component analysis to identify the principal components (eigen images) that showed association with hip fracture. To develop and test a hip fracture risk model based on the principal components, we randomly divided the 75 QCT scans into two groups, one serving as the training set and the other as the test set. We applied this model to estimate a fracture risk index for each test subject, and used the fracture risk indices to discriminate the fracture patients and controls. To evaluate the fracture discrimination efficacy, we performed ROC analysis and calculated the AUC (area under curve). When using the first group as the training group and the second as the test group, the AUC was 0.880, compared to conventional fracture risk estimation methods based on bone densitometry, which had AUC values ranging between 0.782 and 0.871. When using the second group as the training group, the AUC was 0.839, compared to densitometric methods with AUC values ranging between 0.767 and 0.807. Our results demonstrate that principal components derived from hip QCT atlas are associated with hip fracture. Use of such features may provide new quantitative measures of interest to osteoporosis.

The purpose of this study was to evaluate whether contrast-enhanced C-arm CT (3D rotational angiography) can distinguish radiofrequency (RF) ablation lesions created in the left ventricle. Ablation lesions were created on the endocardial surface of the left ventricle of 6 swine using a 7 F RF ablation catheter with a 4 mm electrode. An ECGgated C-arm CT imaging protocol was used to acquire projection images during iodine contrast injection and every 5 min for up to 30 min, with no additional contrast. Reconstructed images were analyzed offline and the mean and standard deviation of the signal intensity of the ablation lesion, normal myocardium, and blood were measured. Eleven ablation lesions were visualized and the time-attenuation curve of the signal intensity was plotted. A mean signal intensity increase of 64.8 ±33.6 HU was measured in the late enhancement of seven lesions compared to normal myocardium. This is the first study to demonstrate RF ablation lesion enhancement patterns similar to those seen for MR imaging using C-arm CT, an imaging modality that can provide valuable feedback during cardiac interventional procedures.

Despite the increasing use of ¹¹C-PiB in research into Alzheimer's disease (AD), there are few standardized analysis procedures that have been reported or published. This is especially true with regards to partial volume effects (PVE) and partial volume correction. Due to the nature of PET physics and acquisition, PET images exhibit relatively low spatial resolution compared to other modalities, resulting in bias of quantitative results. Although previous studies have applied PVE correction techniques on ¹¹C-PiB data, the results have not been quantitatively evaluated and compared against uncorrected data. The aim of this study is threefold. Firstly, a realistic synthetic phantom was created to quantify PVE. Secondly, MRI partial volume estimate segmentations were used to improve voxel-based PVE correction instead of using hard segmentations. Thirdly, quantification of PVE correction was evaluated on 34 subjects (AD=10, Normal Controls (NC)=24), including 12 PiB positive NC. Regional analysis was performed using the Anatomical Automatic Labeling (AAL) template, which was registered to each patient. Regions of interest were restricted to the gray matter (GM) defined by the MR segmentation. Average normalized intensity of the neocortex and selected regions were used to evaluate the discrimination power between AD and NC both with and without PVE correction. Receiver Operating Characteristic (ROC) curves were computed for the binary discrimination task. The phantom study revealed signal losses due to PVE between 10 to 40 % which were mostly recovered to within 5% after correction. Better classification was achieved after PVE correction, resulting in higher areas under ROC curves.

Computational fluid dynamics (CFD) is a technique that is used increasingly in the biomedical field. Solving the flow equations numerically provides a convenient way to assess the efficiency of therapies and devices, ranging from cardiovascular stents and heart valves to hemodialysis workflows. Also in the respiratory field CFD has gained increasing interest, especially through the combination of three dimensional image reconstruction which results in highend patient-specific models. This paper provides an overview of clinical applications of CFD through image based modeling, resulting from recent studies performed in our center. We focused on two applications: assessment of the efficiency of inhalation medication and analysis of endobronchial valve placement. In the first application we assessed the mode of action of a novel bronchodilator in 10 treated patients and 4 controls. We assessed the local volume increase and resistance change based on the combination of imaging and CFD. We found a good correlation between the changes in volume and resistance coming from the CFD results and the clinical tests. In the second application we assessed the placement and effect of one way endobronchial valves on respiratory function in 6 patients. We found a strong patientspecific result of the therapy where in some patients the therapy resulted in complete atelectasis of the target lobe while in others the lobe remained inflated. We concluded from these applications that CFD can provide a better insight into clinically relevant therapies.

Using optimized voxel-based morphometry (VBM), this study systematically investigated the differences and similarities of brain structural changes during the early three developmental periods of human lives: childhood, adolescence and young adulthood. These brain changes were discussed in relationship to the corresponding cognitive function development during these three periods. Magnetic Resonance Imaging (MRI) data from 158 Chinese healthy children, adolescents and young adults, aged 7.26 to 22.80 years old, were included in this study. Using the customized brain template together with the gray matter/white matter/cerebrospinal fluid prior probability maps, we found that there were more age-related positive changes in the frontal lobe, less in hippocampus and amygdala during childhood, but more in bilateral hippocampus and amygdala and left fusiform gyrus during adolescence and young adulthood. There were more age-related negative changes near to central sulcus during childhood, but these changes extended to the frontal and parietal lobes, mainly in the parietal lobe, during adolescence and young adulthood, and more in the prefrontal lobe during young adulthood. So gray matter volume in the parietal lobe significantly decreased from childhood and continued to decrease till young adulthood. These findings may aid in understanding the age-related differences in cognitive function.

Dynamic Contrast Enhanced MRI (DCE-MRI) is one method for drug and tumor assessment. Selecting a consistent arterial input function (AIF) is necessary to calculate tissue and tumor pharmacokinetic parameters in DCE-MRI. This paper presents an automatic and robust method to select the AIF. The first stage is artery detection and segmentation, where knowledge about artery structure and dynamic signal intensity temporal properties of DCE-MRI is employed. The second stage is AIF model fitting and selection. A tri-exponential model is fitted for every candidate AIF using the Levenberg-Marquardt method, and the best fitted AIF is selected. Our method has been applied in DCE-MRIs of four different body parts: breast, brain, liver and prostate. The success rates in artery segmentation for 19 cases are 89.6%±15.9%. The pharmacokinetic parameters computed from the automatically selected AIFs are highly correlated with those from manually determined AIFs (R²=0.946, P(T<=t)=0.09). Our imaging-based tri-exponential AIF model demonstrated significant improvement over a previously proposed bi-exponential model.

Recently, evidences from fMRI studies have shown that there was decreased activity among the default-mode network in Alzheimer's disease (AD), and DTI researches also demonstrated that demyelinations exist in white matter of AD patients. Therefore, combining these two MRI methods may help to reveal the relationship between white matter damages and alterations of the resting state functional connectivity network. In the present study, we tried to address this issue by means of correlation analysis between DTI and resting state fMRI images. The default-mode networks of AD and normal control groups were compared to find the areas with significantly declined activity firstly. Then, the white matter regions whose fractional anisotropy (FA) value correlated with this decline were located through multiple regressions between the FA values and the BOLD response of the default networks. Among these correlating white matter regions, those whose FA values also declined were found by a group comparison between AD patients and healthy elderly control subjects. Our results showed that the areas with decreased activity among default-mode network included left posterior cingulated cortex (PCC), left medial temporal gyrus et al. And the damaged white matter areas correlated with the default-mode network alterations were located around left sub-gyral temporal lobe. These changes may relate to the decreased connectivity between PCC and medial temporal lobe (MTL), and thus correlate with the deficiency of default-mode network activity.

The integrated analysis of the Electroencephalography (EEG), Magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI) are instrumental for functional neuroimaging of the brain. A bottom-up integrated E/MEG and fMRI model based on physiology as well as a method for estimating its parameters are keys to the integrated analysis. We propose the variational Bayesian expectation maximization (VBEM) method to estimate parameters of our proposed integrated model. VBEM method iteratively optimizes a lower bound on the marginal likelihood. An iteration of the VBEM consists of two steps: a variational Bayesian expectation step implemented using the extended Kalman smoother (EKS) and the posterior probability of the parameters in the previous step, and a variational Bayesian maximization step to estimate the posterior distributions of the parameters. For a given external stimulus, a variety of multi-area models can be considered in which the number of areas and the configuration and strength of connections between the areas are different. The proposed VBEM method can be used to select an optimal model as well as estimate its parameters. The efficiency of the proposed VBEM method is illustrated using simulation and real datasets. The proposed VBEM method can be used to estimate parameters of other non-linear dynamical systems. This study proposes an effective method to integrate E/MEG and fMRI and plans to use these techniques in functional neuroimaging.

Although top-down perceptual process plays an important role in face processing, its neural substrate is still puzzling because the top-down stream is extracted difficultly from the activation pattern associated with contamination caused by bottom-up face perception input. In the present study, a novel paradigm of instructing participants to detect faces from pure noise images is employed, which could efficiently eliminate the interference of bottom-up face perception in topdown face processing. Analyzing the map of functional connectivity with right FFA analyzed by conventional Pearson's correlation, a possible face processing pattern induced by top-down perception can be obtained. Apart from the brain areas of bilateral fusiform gyrus (FG), left inferior occipital gyrus (IOG) and left superior temporal sulcus (STS), which are consistent with a core system in the distributed cortical network for face perception, activation induced by top-down face processing is also found in these regions that include the anterior cingulate gyrus (ACC), right oribitofrontal cortex (OFC), left precuneus, right parahippocampal cortex, left dorsolateral prefrontal cortex (DLPFC), right frontal pole, bilateral premotor cortex, left inferior parietal cortex and bilateral thalamus. The results indicate that making-decision, attention, episodic memory retrieving and contextual associative processing network cooperate with general face processing regions to process face information under top-down perception.

Functional magnetic resonance imaging (fMRI) has complementary spatiotemporal resolution compared to Electroencephalography (EEG) as well as Magnetoencephalography (MEG). Thus, their integrated analysis should improve the overall resolution. To integrate analysis of E/MEG and fMRI, we extend our previously proposed integrated E/MEG and fMRI neural mass model to a multi-area model by defining two types of connections: the Short-Range Connections (SRCs) between minicolumns within the areas and Long-Range Connections (LRCs) between inter-areas minicolumns. The nonlinear input/output relationship in the proposed model is derived from the state space representation of the multi-area model. The E/MEG signals are originated from the overall synaptic activities of the pyramidal cells of all minicolumns and can be calculated using the lead field matrix (i.e., forward electromagnetic model). The fMRI signal is extracted from the proposed integrated model by calculating the overall neural activities in the areas and using it as the input of the extended balloon model (EBM). Using the simulation results, the capabilities of the proposed model to generate E/MEG and fMRI signals is shown. In addition, changes in the dynamics of the model to variations of its parameters were evaluated and lead us to the appropriate ranges for the parameters. In conclusion, this work proposes an effective method to integrate E/MEG and fMRI for the more effective use of these techniques in functional neuroimaging.

This work proposed to use the linear Gaussian Bayesian network (BN) to construct the effective connectivity model of the brain's default mode network (DMN), a set of regions characterized by more increased neural activity during rest-state than most goal-oriented tasks. In a complete unsupervised data-driven manner, Bayesian information criterion (BIC) based learning approach was utilized to identify a highest scored network whose nodes (brain regions) were selected based on the result from the group independent component analysis (Group ICA) examining the DMN. We put forward to adopt the statistical significance testing method for regression coefficients used in stepwise regression analysis to further refine the network identified by BIC. The final established BN, learned from the functional magnetic resonance imaging (fMRI) data acquired from 12 healthy young subjects during rest-state, revealed that the hippocampus (HC) was the most influential brain region that affected activities in all other regions included in the BN. In contrast, the posterior cingulate cortex (PCC) was influenced by other regions, but had no reciprocal effects on any other region. Overall, the configuration of our BN illustrated that a prominent connection from HC to PCC existed in the DMN.

Functional network connectivity (FNC) measures the temporal dependency among the time courses of functional networks. However, the marginal correlation between two networks used in the classic FNC analysis approach doesn't separate the FNC from the direct/indirect effects of other networks. In this study, we proposed an alternative approach based on partial correlation to evaluate the FNC, since partial correlation based FNC can reveal the direct interaction between a pair of networks, removing dependencies or influences from others. Previous studies have demonstrated less task-specific activation and less rest-state activity in Alzheimer's disease (AD). We applied present approach to contrast FNC differences of resting state network (RSN) between AD and normal controls (NC). The fMRI data under resting condition were collected from 15 AD and 16 NC. FNC was calculated for each pair of six RSNs identified using Group ICA, thus resulting in 15 (2 out of 6) pairs for each subject. Partial correlation based FNC analysis indicated 6 pairs significant differences between groups, while marginal correlation only revealed 2 pairs (involved in the partial correlation results). Additionally, patients showed lower correlation than controls among most of the FNC differences. Our results provide new evidences for the disconnection hypothesis in AD.

Conventional voxel-based group analysis of functional magnetic resonance imaging (fMRI) data typically requires warping each subject's brain images onto a common template to create an assumed voxel correspondence. The implicit assumption is that aligning the anatomical structures would correspondingly align the functional regions of the subjects. However, due to anatomical and functional inter-subject variability, mis-registration often occurs. Moreover, wholebrain warping is likely to distort the spatial patterns of activation, which have been shown to be important markers of task-related activation. To reduce the amount of mis-registration and distortions, warping at the brain region level has recently been proposed. In this paper, we investigate the effects of both whole-brain and region-level warping on the spatial patterns of activation statistics within certain regions of interests (ROIs). We have chosen to examine the bilateral thalami and cerebellar hemispheres during a bulb-squeezing experiment, as these regions are expected to incur taskrelated activation changes. Furthermore, the appreciable size difference between the thalamus and cerebellum allows for exploring the effects of warping on various ROI sizes. By applying our recently proposed 3D moment-based invariant spatial features to characterize the spatial pattern of fMRI activation statistics, we demonstrate that whole-brain warping generally reduced discriminability of task-related activation differences. Applying the same spatial analysis to ROIs warped at the region level showed some improvements over whole-brain warping, but warp-free analysis resulted in the best performance. We hence suggest that spatial analysis of fMRI data that includes spatial warping to a common space must be interpreted with caution.

Studying interactions within the brain leads to an emerging field: functional connectivity. Functional connectivity between two brain units (neuron columns, recording sites, regions) can be defined as the temporal correlation between their time courses. Correlation between time courses of brain units are measured in different ways, e.g., Data-driven and/or Model-based approaches. This paper focuses on the former. The commonly used measures in Data-driven approach include, but are not limited to Coherence, Synchronization, Mutual Information, Nonlinear correlation coefficient, and Phase-Locking Values. We first describe the underlying reasons why these measures originated from distinctive fields of science and engineering can be applied to assess functional connectivity; then give the quantitative evaluation to each measure and indicate what are the limitations and conditions when they are applied, finally demonstrate the relations between these measures that may provide a basis for consistent assessments and interpretations on functional connectivity under investigation.

Magnetoencephalography (MEG) is a neuroimaging technique for brain activation detection. This technique does not provide a unique solution due to ill-posedness of its inverse solution. Several methods are proposed to improve the MEG inverse solution. Minimum Norm (MN) is a simple method whose solution is distributed and biased toward the superficial sources. In addition, its solution is sensitive to the noise. Several methods are proposed to improve performance of the MN method. In this paper, we propose a method whose solution is less sensitive to the noise and spatially unbiased toward the superficial sources. Control of focal solution properties is achieved by specifying a parameter in the proposed method. Performance of the proposed method is compared to others using simulation studies consisting of single and multiple dipole sources as well as an extended source model. Proposed method has superior performance compared to non-iterative methods. Its performance is similar to the iterative methods but its computational load is lower.

A hybrid blood input function (BIF) model that incorporates region of interests (ROIs) based peak estimation and a two exponential tail model was proposed to describe the blood input function. The hybrid BIF model was applied to the single-input-multiple-output (SIMO) optimization based approach for BIF estimation using time activity curves (TACs) obtained from ROIs defined at left ventricle (LV) blood pool and myocardium regions of dynamic PET images. The proposed BIF estimation method was applied with 0, 1 and 2 blood samples as constraints for BIF estimation using simulated small animal PET data. Relative percentage difference of the area-under-curve (AUC) measurement between the estimated BIF and the true BIF was calculated to evaluate the BIF estimation accuracy. SIMO based BIF estimation using Feng's input function model was also applied for comparison. The hybrid method provided improved BIF estimation in terms of both mean accuracy and variability compared to Feng's model based BIF estimation in our simulation study. When two blood samples were used as constraints, the percentage BIF estimation error was 0.82 ± 4.32% for the hybrid approach and 4.63 ± 10.67% for the Feng's model based approach. Using hybrid BIF, improved kinetic parameter estimation was also obtained.

The statistic of RV-coefficient is a good substitute for the Pearson correlation coefficient to measure the temporal similarity of two local brain regions. However, the hypothesis test of RV-coefficient is a hard problem which limits its application. This paper discussed the problem in details. Since the distribution of RV-coefficient is unknown, we do not know a critical p-value to statistically test its significance. We proposed a new strategy to test the significance of RV calculated from fMRI. In order to approximate the p-value, we elicited the first two moments of the population permutation distribution of RV; we then derived a formula to more closely approximate the normal distribution with these transformed statistics. These transformations of statistics are suggested for a precise approximation to the permutational p-value even under large number of observations. This strategy of test can greatly reduce the computational complexity and avoid "calculation catastrophe", we then use the statistic of RV to extract the map of functional connectivity from fMRI and test its significance with the strategy proposed here.

Magnetoencephalography (MEG) is a multi-channel, functional imaging technique. It measures the magnetic field produced by the primary electric currents inside the brain via a sensor array composed of a large number of superconducting quantum interference devices. The measurements are then used to estimate the locations, strengths, and orientations of these electric currents. This magnetic source imaging technique encompasses a great variety of signal processing and modeling techniques which include Inverse problem, MUltiple SIgnal Classification (MUSIC), Beamforming (BF), and Independent Component Analysis (ICA) method. A key problem with Inverse problem, MUSIC and ICA methods is that the number of sources must be detected a priori. Although BF method scans the source space on a point-to-point basis, the selection of peaks as sources, however, is finally made by subjective thresholding. In practice expert data analysts often select results based on physiological plausibility. This paper presents an eigenstructure approach for the source number detection in MEG neuroimaging. By sorting eigenvalues of the estimated covariance matrix of the acquired MEG data, the measured data space is partitioned into the signal and noise subspaces. The partition is implemented by utilizing information theoretic criteria. The order of the signal subspace gives an estimate of the number of sources. The approach does not refer to any model or hypothesis, hence, is an entirely data-led operation. It possesses clear physical interpretation and efficient computation procedure. The theoretical derivation of this method and the results obtained by using the real MEG data are included to demonstrates their agreement and the promise of the proposed approach.

Detailed knowledge of the hemodynamic conditions in normal cerebral arteries is important for a better understanding of the underlying mechanisms leading to the initiation and progression of cerebrovascular diseases. The goal of our research is to characterize the hemodynamic patterns in the major cerebral arteries of normal subjects using 4D phase-contrast magnetic resonance imaging (PC-MR) and image-based computational fluid dynamics (CFD), and to assess the consistency of the flow patterns determined by these two techniques. Time-resolved 4D PC-MR images of the cerebral arteries at the level of the Circle of Willis were acquired on three normal subjects and corresponding subject-specific CFD models were constructed. Visualizations of the flow fields show that qualitatively, the major flow structures, swirling flows, flow directions in communicating arteries, etc. observed in the PC-MR images and the CFD calculations are consistent. However, each technique has limitations that introduce differences between the corresponding flow fields. This paper discusses these differences in order to better interpret the results obtained with each technique, and to be aware of the regions along the arteries where each technique is expected to over simplify the velocity patterns or yield under or over estimations of the velocity and wall shear stress magnitudes.

This study addresses whether the vasculature grows in proportion to the myocardium as the rat heart develops. The volume of myocardium and coronary vessels were estimated from micro-CT images of the hearts injected with Microfil(R) contrast agent. Young (n=5) and adult (n=5) hearts were scanned, resulting in 3D images comprised of 20μm on-a-side cubic voxels. The myocardial muscle and vessel lumen volumes were measured for all vessels 40 to 320μm in diameter by an erosion and dilation method applied to the binary images in which the contrast in the vessels were assigned "1" and all non-opacified entities were assigned "0". The average total muscle volume increases by 50%, 129.4 to 237.4mm³, from young to adult rats, while the luminal volume increases by 10%, 16.6 to 18.6mm³. The vessel volume is 12% of the total muscle volume in young and 8% in adults. For a given vessel volume, the muscle volume in the young is 82% of the muscle volume in adults. We conclude that as the heart matures, the myocardium grows more rapidly than the vasculature. This may result in greater angles of separation between vessel branches, and the increase in myocardial coronary volume. The ratio suggests either higher blood flow velocity or a lower metabolic rate in adults.

This paper combines different parallelization strategies for speeding up motion and deformation computation by non-rigid registration of a sequence of images. The registration is performed in a two-level acceleration approach: (1) parallelization of each registration process using MPI and/or threads, and (2) distribution of the sequential registrations over a cluster. On a 24-node double quad-core Intel Xeon (2.66 GHz CPU, 16 GB RAM) cluster, the method is demonstrated to efficiently compute the deformation of a cardiac sequence reducing the computation time from more than 3 hours to a couple of minutes (for low downsampled images). It is shown that the distribution of the sequential registrations over the cluster together with the parallelization of each pairwise registration by multithreading lowers the computation time towards values compatible with clinical requirements (a few minutes per patient). The combination of MPI and multithreading is only advantageous for large input data sizes. Performances are assessed for the specific scenario of aligning cardiac sequences of taggedMagnetic Resonance (tMR) images, with the aim of comparing strain in healthy subjects and hypertrophic cardiomyopathy (HCM) patients. In particular, we compared the distribution of systolic strain in both populations. On average, HCM patients showed lower average values of strain with larger deviation due to the coexistence of regions with impaired deformation and regions with normal deformation.

Gated single photon emission tomography (gSPECT) is a well-established technique used routinely in clinical practice. It can be employed to evaluate global left ventricular (LV) function of a patient. The purpose of this study is to assess LV systolic and diastolic function from gSPECT datasets in comparison with cardiac magnetic resonance imaging (CMR) measurements. This is achieved by applying our recently implemented 3D active shape model (3D-ASM) segmentation approach for gSPECT studies. This methodology allows for generation of 3D LV meshes for all cardiac phases, providing volume time curves and filling rate curves. Both systolic and diastolic functional parameters can be derived from these curves for an assessment of patient condition even at early stages of LV dysfunction. Agreement of functional parameters, with respect to CMR measurements, were analyzed by means of Bland-Altman plots. The analysis included subjects presenting either LV hypertrophy, dilation or myocardial infarction.

With the increasing popularity of cardiac ablation therapy, studies of the procedural effects on left atrial and pulmonary vein morphology are becoming more important. Of particular interest is evaluation of atrial and pulmonary vein remodeling following ablation therapy using structural imaging. One challenge that arises when comparing pulmonary vein morphology across subjects is defining the ostial location. Strategies for defining this important anatomical location include volume renderings from multiple angles, or drawing lines in cross-sectional images. Drawbacks of these techniques include subjectivity between raters as well as limited use of three dimensional volumetric information. In this work, we describe a method for automatically identifying the pulmonary vein ostia from CT images using a single user selected seedpoint. The technique makes use of the full three dimensional volumetric information, by computing a centerline along each pulmonary vein and defining the ostium using oblique cross-sectional image planes along the curve axis. The ostium is defined as the point at which there is a spike in the oblique cross-sectional area. The method is demonstrated on each of the four pulmonary veins in four patient datasets, for a total of sixteen applications of the algorithm. The results are compared against manual delineations of the pulmonary vein ostia, with overall mean distances ranging from approximately 1.5 to 5.0 mm. In conclusion, although the pulmonary veins exhibit variable anatomic shapes and orientations across different patient datasets, our proposed automated method produces results comparable to manual delineation of the ostia.

Imaging techniques try to identify patients who may respond to cardiac resynchronization therapy (CRT). However, it may be clinically more useful to identify patients for whom CRT would not be beneficial as the procedure would not be indicated for this group. We developed a novel, clinically feasible and technically-simple echocardiographic dyssynchrony index and tested its negative predictive value. Subjects with standard indications for CRT had echo preand post-device implantation. Atrial-ventricular dyssynchrony was defined as a left ventricular (LV) filling time of <40% of the cardiac cycle. Intra-ventricular dyssynchrony was quantified as the magnitude of LV apical rocking. The apical rocking was measured using tissue displacement estimates from echo data. In a 4-chamber view, a region of interest was positioned within the apical end of the middle segment within each wall. Tissue displacement curves were analyzed with custom software in MATLAB. Rocking was quantified as a percentage of the cardiac cycle over which the displacement curves showed discordant behavior and classified as non-significant for values <35%. Validation in 50 patients showed that absence of significant LV apical rocking or atrial-ventricular dyssynchrony was associated with non-response to CRT. This measure may therefore be useful in screening to avoid non-therapeutic CRT procedures.

Computed tomography (CT) enables high resolution, whole-body imaging with excellent depth penetration. The development of new targeted radiopaque CT contrast agents can provide the required sensitivity and localization for the successful detection and diagnosis of smaller lesions representing earlier disease. Nanoscale, perfluorooctylbromide (C8F17Br, PFOB) droplets have previously been used as untargeted contrast agents in X-ray imaging, and form the basis of a promising new group of agents that can be developed for targeted CT imaging. For successful targeting to disease sites, new PFOB droplet formulations tailored for ideal in vivo performance (e.g., biodistribution, toxicity, and pharmacokinetics) must be developed. However, the direct assessment of PFOB agents in biological environments early in their development is difficult using CT, as its sensitivity is not adequate for identification of single probes in vitro or in vivo. In order to allow single droplet interactions with cells to be directly assessed using standard cellular imaging tools, we integrate an optical marker within the PFOB agent. In this work, a new method to label a PFOB agent with fluorescent quantum dot (QD) nanoparticles is presented. These composite PFOB-QD droplets loaded into macrophage cells result in fluorescence on a cellular level that correlates well to the strong CT contrast exhibited in corresponding tissue-mimicking cell pellets. QD loading within the PFOB droplet core allows optical labeling without influencing the surface-dependent properties of the PFOB droplets in vivo, and may be used to follow PFOB localization from in vitro cell studies to histopathology.

As a promising tool for in-vivo molecular imaging of small animals, Bioluminescence Tomography (BLT) aims at the quantitative reconstruction of the bioluminescent source distribution from the detected optical signals on the body surface. Mathematically, BLT is a highly ill-posed inverse problem per se. Most existing works are based on Tikhonov regularization in which the selection of the proper regular parameter is quite difficult. In this paper, two direct regularization methods, truncated singular value decomposition (TSVD) and truncated total least squares (TTLS), as well as two iterative regularization methods, conjugate gradient least squares (CGLS) and least squares QR decomposition (LSQR), are applied to the inverse problem in BLT, with the finite element method solving the diffusion equation. In the numerical simulation, a heterogeneous phantom is designed to compare and evaluate the four methods. The results show that all the four methods can reconstruct the position of bioluminescence sources accurately and are more convenient in the determination of regularization parameter than Tikhonov method. In addition, with a priori knowledge of the source permissible region employed in the reconstruction, the iterative methods are faster than the two direct methods. Among the four methods, LSQR performs quite stably when both model noise and measure noise are considered.

Bioluminescence tomography (BLT) is a novel in vivo technique in small animal studies, which can reveal the molecular and cellular information at the whole-body small animal level. At present, there is an increasing interest in multispectral bioluminescence tomography, since multispectral data acquisition could improve the BLT performance significantly. In view to the ill-posedness of BLT problem, we develop an optimal permissible source region strategy to constrain the possible solution of the source by utilizing spectrum character of bioluminescent source. Then a linear system to link the measured data with the unknown light source variables is established by utilizing the optimal permissible region strategy based on adaptive finite element analysis. Furthermore, singular value decomposition analysis is used for data dimensionality reduction and improving computational efficiency in multispectral case. The reconstructed speed and stability benefit from adaptive finite element, the permissible region strategy and singular value decomposition. In the numerical simulation, the heterogeneous phantom experiment has been used to evaluate the performance of the proposed algorithm with the Monte Carlo based synthetic data. The reconstruction results demonstrate the merits and potential of our methodology for localizing bioluminescent source.

Glaucoma is a group of diseases which can cause vision loss and blindness due to gradual damage to the optic nerve. The ratio of the optic disc cup to the optic disc is an important structural indicator for assessing the presence of glaucoma. The purpose of this study is to develop and evaluate a method which can segment the optic disc cup and neuroretinal rim in spectral-domain OCT scans centered on the optic nerve head. Our method starts by segmenting 3 intraretinal surfaces using a fast multiscale 3-D graph search method. Based on one of the segmented surfaces, the retina of the OCT volume is flattened to have a consistent shape across scans and patients. Selected features derived from OCT voxel intensities and intraretinal surfaces were used to train a k-NN classifier that can determine which A-scans in the OCT volume belong to the background, optic disc cup and neuroretinal rim. Through 3-fold cross validation with a training set of 20 optic nerve head-centered OCT scans (10 right eye scans and 10 left eye scans from 10 glaucoma patients) and a testing set of 10 OCT scans (5 right eye scans and 5 left eye scans from 5 different glaucoma patients), segmentation results of the optic disc cup and rim for all 30 OCT scans were obtained. The average unsigned errors of the optic disc cup and rim were 1.155 ± 1.391 pixels (0.035 ± 0.042 mm) and 1.295 ± 0.816 pixels (0.039 ± 0.024 mm), respectively.

In optical imaging, although the standard P1 diffusion theory is widely used, its angular flux at boundary is discontinuous, and this model is not incapable of exactly modeling light transport in biological tissue with partially-reflective boundary. In this work, we present a hybrid P1-DP0 (P1 spherical harmonics-double DP0 spherical harmonics) diffusion theory in 3D environment, which effectively interpolates between the P1 and DP0 approximation by a space-dependent weight factor α(r) that controls the local angular approximation. Comparing to the P1 model, the solutions of our model are consistently accurate over a broad range of optical properties. Moreover, with the same reduced scattering and absorption properties, the hybrid model for high anisotropic scattering which is the common case for mammal tissue is more accurate than the low one. Finally, this theory is validated by Monte Carlo simulations.

Clinically-translatable redox imaging methods developed in the Chance laboratory have been used for imaging mitochondrial metabolic states in tissues. The fluorescence of reduced pyridine nucleotide (PN or NADH) and oxidized flavoproteins (Fp) in the respiratory chain is sensitive to intracellular redox states. The redox ratios, i.e., Fp/(Fp+NADH) and NADH/(Fp+NADH) provide important metabolic information in living tissues. Usually the higher the metabolic flux, the less NADH, the more oxidized Fp, and the higher Fp redox ratio. Snap-freezing tissue samples under the liquid nitrogen condition preserves the tissue metabolic state in vivo. Here we report our work on the calibration of a homebuilt Charged Coupled Device (CCD) cryogenic redox imager using a series of snap-frozen solution standards of NADH and Fp. The NADH concentration ranged from 0-1318 μM and Fp from 0-719 μM. The sensitivity ratio of NADH and Fp channels was determined from the slope ratio of the two calibration curves and was used to correct the redox ratio of a human melanoma mouse xenograft. The NADH and Fp reference standards were placed adjacent to the tissue samples and their emission intensities were used to quantitatively determine the concentrations of NADH and Fp in a mouse xenograft of a human breast cancer line. Our method of imaging tissue samples along with reference NADH and Fp standards should facilitate the comparison of redox images obtained at different times or with different instrument parameters.

Measurement of oxygen saturation has proved to give important information about the eye health and the onset of eye pathologies such as Diabetic Retinopathy. Recently, we have presented a multi-aperture system enabling snapshot acquisition of human fundus images at six different wavelengths. In our setup a commercial fundus ophthalmoscope was interfaced with the multi-aperture system to acquire spectroscopic sensitive images of the retina vessel, thus enabling assessment of the oxygen saturation in the retina. Snapshot spectroscopic acquisition is meant to minimize the effects of eye movements. Higher measurement accuracy can be achieved by increasing the number of wavelengths at which the fundus images are taken. In this study we present an improvement of our setup by introducing an other multi-aperture camera that enables us to take snapshot images of the fundus at nine different wavelengths. Careful consideration is taken to improve image transfer by measuring the optical properties of the fundus camera used in the setup and modeling the optical train in Zemax.

A robust pattern classifier algorithm for the variable symmetric plane model, where the driving noise is a mixture of a Gaussian and an outlier process, is developed. The veracity and high-speed performance of the pattern recognition algorithm is proved. Bioluminescence tomography (BLT) has recently gained wide acceptance in the field of in vivo small animal molecular imaging. So that it is very important for BLT to how to acquire the highprecision region of interest in a bioluminescence image (BLI) in order to decrease loss of the customers because of inaccuracy in quantitative analysis. An algorithm in the mode is developed to improve operation speed, which estimates parameters and original image intensity simultaneously from the noise corrupted image derived from the BLT optical hardware system. The focus pixel value is obtained from the symmetric plane according to a more realistic assumption for the noise sequence in the restored image. The size of neighborhood is adaptive and small. What's more, the classifier function is base on the statistic features. If the qualifications for the classifier are satisfied, the focus pixel intensity is setup as the largest value in the neighborhood.Otherwise, it will be zeros.Finally,pseudo-color is added up to the result of the bioluminescence segmented image. The whole process has been implemented in our 2D BLT optical system platform and the model is proved.

As a novel optical molecular imaging technique, bioluminescence tomography (BLT) can be used to monitor the biological activities non-invasively at the cellular and molecular levels. In most of known BLT studies, however, the time variation of the bioluminescent source is neglected. It gives rise to the inconsistent views during the multiview continuous wave measurement. In other words, the real measured data from different measured views come from 'different' bioluminescent sources. It could bring large errors in bioluminescence reconstruction. In this paper, a posteriori correction strategy for adaptive FEM-based reconstruction is proposed and developed. The method helps to improve the source localization considering the bioluminescent energy variance during the multiview measurement. In the method, the correction for boundary signals by means of a posteriori correction strategy, which adopts the energy ratio of measured data in the overlapping domains between the adjacent measurements as the correcting factor, can eliminate the effect of the inconsistent views. Then the adaptive mesh refinement with a posteriori error estimation helps to improve the precision and efficiency of BLT reconstruction. In addition, a priori permissible source region selection based on the surface measured data further reduces the ill-posedness of BLT and enhances numerical stability. Finally, three-dimension numerical simulations using the heterogeneous phantom are performed. The numerically measured data is generated by Monte Carlo (MC) method which is known as the Gold standard and can avoid the inverse crime. The reconstructed result with correction shows more accuracy compared to that without correction.

Computed tomography (CT) examination is often used to quantify the relation between lung function and airway remodeling in chronic obstructive pulmonary disease (COPD). In this preliminary study, we examined the association between lung function and airway wall computed attenuation ("density") in 200 COPD screening subjects. Percent predicted FVC (FVC%), percent predicted FEV1 (FEV1%), and the ratio of FEV1 to FVC as a percentage (FEV1/FVC%) were measured post-bronchodilator. The apical bronchus of the right upper lobe was manually selected from CT examinations for evaluation. Total airway area, lumen area, wall area, lumen perimeter and wall area as fraction of the total airway area were computed. Mean HU (meanHU) and maximum HU (maxHU) values were computed across pixels assigned membership in the wall and with a HU value greater than -550. The Pearson correlation coefficients (PCC) between FVC%, FEV1%, and FEV1/FVC% and meanHU were -0.221 (p = 0.002), -0.175 (p = 0.014), and -0.110 (p = 0.123), respectively. The PCCs for maxHU were only significant for FVC%. The correlations between lung function and the airway morphometry parameters were slightly stronger compared to airway wall density. MeanHU was significantly correlated with wall area (PCC = 0.720), airway area (0.498) and wall area percent (0.611). This preliminary work demonstrates that airway wall density is associated with lung function. Although the correlations in our study were weaker than a recent study, airway wall density initially appears to be an important parameter in quantitative CT analysis of COPD.

Sponges consist of a tissue skeleton that provides structure for its elaborate pore system of canals and chambers. Sponges have been noted for their remarkable ability to support cellular life within these pores. For that reason, their structure is of great interest to us since our goal is to create a scaffold that supports cell vitality beyond diffusion depth from the scaffold surface. In the sponge this is achieved by convective transport of nutrients through the pore system. Hence, understanding of the architecture of sea sponges has the potential to aid in the production of better design of porous, cell-populated, synthetic tissue scaffolds. Pore geometry affects depth and distribution of the solute transport needed to sustain the cells lining the pores. To explore this aspect we need accurate 3D measurements of pore architecture and interconnectivity. Three-dimensional micro-CT imaging can be used to characterize the desired microarchitecture labyrinthine pore structure of a sea sponge. The sea sponge was collected, dried, and then rotated in small angular measurements inside the scanner as an x-ray image was obtained at each angle of view. Reconstructed cross-sectional images of the sponge consisted of up to 10⁷ cubic voxels, 20μm on a side. After reconstruction, Analyze 8.1 was used to display and generate measurements of the sponge's pores. The pores were subjected to a sequence of morphological erosion and dilation operations, each of which either removed or added one layer of voxels from the outer surface of the segmented pore. Hence, each erosion removed 40μm from the diameter of a pore. Progressive erosions were used to calculate pore volume and to disconnect pores from adjacent pores, thereby identifying connecting throats(s) as well as their diameter. Along with diameter, individual pore volume and surface area were also computed. Results show that the throats were predominately 264±129μm in diameter. Preliminary data show complex pore structures can be analyzed with morphological erosion and dilation image analysis techniques to provide significant quantitative data. Such data has provided information about throat identification and diameter, as well as pore volume and surface area. Ground work has also been laid for computing flow path of least resistance through the pore labyrinth from any point in the labyrinth.

Microwave imaging for breast cancer detection is becoming a promising alternative technique to current imaging modalities. The significant contrast between dielectric properties of normal and malignant breast tissues makes microwave imaging a useful technique to provide important functional information for diagnoses. However, one of its limitations is that it intrinsically cannot produce high resolution images as other conventional techniques such as MRI or X-ray CT do. Those modalities are capable of producing high quality anatomical images, but unlike microwave imaging, they often cannot provide the necessary functional information about tissue health. In order to refine the resolution of the microwave images while also preserving the functional information, we have recently developed a new strategy, called soft prior regularization. In this new approach, the prior anatomical information of the tissue from either x-ray, MR or other sources is incorporated into our microwave imaging reconstruction algorithm through the following steps: First, the anatomical information is used to create a reconstruction mesh which defines the boundaries of different internal regions. Second, based on location of each mesh node, an associated weighting matrix is defined, such that all nodes within each region are grouped with each other. Finally, the soft prior matrix is used as a regularizing term for our original Gauss- Newton reconstruction algorithm. Results from initial phantom experiments show a significant improvement in the recovered dielectric properties.

We propose a new registration method that can do the alignment of two medical images using simultaneously the ordinary Procrustes analysis as well as a maximum likelihood framework with the EM algorithm. In an initial registration, we first extract the feature points representing the shape information from the boundary of the segmented object, and then we apply the ordinary Procrustes analysis to register exactly two sets of extracted feature points. For the final registration, we define a new alignment measure with the log-likelihood function derived by the Bayes theory and the maximum likelihood method with EM algorithm. In the E-step, we compute the posterior distribution of label variable by taking expectation for the log-likelihood function. And in the M-step, we derive the estimators for all parameters by maximizing the log-likelihood function. Then, we can optimize the transformation parameters for the final image registration by applying iteratively this measure. Finally, we conduct the various experiments to analyze the accuracy and precision of the proposed method. The experimental results show that our method has great potential power to register various images given by multimodality instruments.

Structural features extraction is essential in various molecular biology applications such as functional classification or binding site prediction for molecular docking. In the literature, methods to study the topology and the accessibility of molecule surfaces exist. Some of them are based on the 3D Delaunay triangulation of the set of points formed by the atoms center. In this paper, we propose to investigate the spectral properties of this triangulation by computing and analyzing the first eigenvector of its adjacency matrix. This technique is already used in graph theory to extract core features and to compare networks, 3D meshes, or any set of points and edges. Tests were performed, providing two promising results. First, the correlation between eigenvectors computed from a molecular complex and one of its component is much higher than between structure independent molecules. It allows to find common sub-structures between molecules even after small conformation changes, because no distance is considered, but only the adjacency of the Delaunay triangulation. Second, the value of the eigenvector at indexes corresponding to binding site atoms is higher than for other surface atoms. As this feature is correlated with no other important geometric or physicochemical binding site properties (curvature, depth, hydrogen bonds capacity, ...), it can be integrated in a larger process aiming to localize binding sites.

CT colonography is a minimally invasive technique that can be used to find polyps and malignant tumors in the colon. However, if a polyp or malignant tumor is found, a colonoscopy is then required to further investigate and remove it. One major problem in relaying the location of a polyp between radiologists and colonoscopists is the ambiguity of the divisions between various colon segments. Because there exists no concrete separator between segments, miscommunication of polyp locations can result. In an effort to minimize such miscommunications, an automated labeling program has been created. This program reads in CT images and returns physical coordinates of the divisions between segments. Such a system would allow for a more universally accepted method for communication of polyp location between radiologists and colonoscopists, and hopefully increase the speed and ease with which such polyp location can be reported. The purpose of this study was to validate the automated method of labeling by comparing physical coordinates of region dividers found using the program with those manually determined by a radiologist. The segments were defined with a modified version of a procedure developed by Taylor et al (Radiology 229:99-108, 2003). A set of 30 scans was used to train the system and then a test set of 216 cases was used to validate the system. The system reported locations that averaged 1-3 cm different than manually reported locations. The errors are on the order of the diameters of the colonic segments and are in the clinically acceptable range.

The nanoparticle magnetization induced by a sinusoidal field is a distorted sinusoid and the amount of the distortion is determined by the mobility of the magnetizations. The distortion is higher for nanoparticle magnetizations with limited mobility. The mobility of the magnetizations is influenced by a variety of factors that are important in biomedical applications: temperature, viscosity and binding are perhaps the most obvious. The distortion can be quantified by a ratio of the Fourier coefficients, which is independent of concentration and can be measured in vivo. For example, we have introduced a method of measuring the temperature in vivo and the method can be extended to measure the other factors that influence the mobility of the magnetization.

The general linear model (GLM) has been extensively applied to fMRI data in the time domain. However, traditionally time series data can be analyzed in the Fourier domain where the assumptions made as to the noise in the signal can be less restrictive and statistical tests are mathematically more rigorous. A complex form of the GLM in the Fourier domain has been applied to the analysis of fMRI (BOLD) data. This methodology has a number of advantages over temporal methods: 1. Noise in the fMRI data is modeled more generally and closer to that actually seen in the data. 2. Any input function is allowed regardless of the timing. 3. Non-parametric estimation of the transfer functions at each voxel are possible. 4. Rigorous statistical inference of single subjects is possible. This is demonstrated in the analysis of an experimental design with random exponentially distributed stimulus inputs (a two way ANOVA design with input stimuli images of alcohol, non-alcohol beverage and positive or negative images) sampled at 400 milliseconds. This methodology applied to a pair of subjects showed precise and interesting results (e.g. alcoholic beverage images attenuate the response of negative images in an alcoholic as compared to a control subject).

Probabilistic atlases of anatomical organs, especially the brain and the heart, have become popular in medical image analysis. We propose the construction of probabilistic atlases which retain structural variability by using a size-preserving modified affine registration. The organ positions are modeled in the physical space by normalizing the physical organ locations to an anatomical landmark. In this paper, a liver probabilistic atlas is constructed and exploited to automatically segment liver volumes from abdominal CT data. The atlas is aligned with the patient data through a succession of affine and non-linear registrations. The overlap and correlation with manual segmentations are 0.91 (0.93 DICE coefficient) and 0.99 respectively. Little work has taken place on the integration of volumetric measures of liver abnormality to clinical evaluations, which rely on linear estimates of liver height. Our application measures the liver height at the mid-hepatic line (0.94 correlation with manual measurements) and indicates that its combination with volumetric estimates could assist the development of a noninvasive tool to assess hepatomegaly.

This paper describes a method for estimating, from dynamic contrast-enhanced MRI raw k-space data of the breast, parameter maps that model tissue properties associated with a compartmental model of contrast exchange. The contrast agent kinetics, as represented by these parameter maps, are important in distinguishing benign and malignant tumors. The proposed model-based reconstruction algorithm estimates tissue parameter maps directly from MRI k-space data, thereby allowing a new and improved set of spatiotemporal resolution and noise tradeoffs. Realistic noise levels and an undersampling factor of R=4 appeared to provide reasonable accuracy for the kinetic parameters of interest.

This study was undertaken to register 3D parametric breast images derived from Gd-DTPA MR and F-18-FDG PET/CT dynamic image series. Nonlinear curve fitting (Levenburg-Marquardt algorithm) based on realistic two-compartment models was performed voxel-by-voxel separately for MR (Brix) and PET (Patlak). PET dynamic series consists of 50 frames of 1-minute duration. Each consecutive PET image was nonrigidly registered to the first frame using a finite element method and fiducial skin markers. The 12 post-contrast MR images were nonrigidly registered to the precontrast frame using a free-form deformation (FFD) method. Parametric MR images were registered to parametric PET images via CT using FFD because the first PET time frame was acquired immediately after the CT image on a PET/CT scanner and is considered registered to the CT image. We conclude that nonrigid registration of PET and MR parametric images using CT data acquired during PET/CT scan and the FFD method resulted in their improved spatial coregistration. The success of this procedure was limited due to relatively large target registration error, TRE = 15.1±7.7 mm, as compared to spatial resolution of PET (6-7 mm), and swirling image artifacts created in MR parametric images by the FFD. Further refinement of nonrigid registration of PET and MR parametric images is necessary to enhance visualization and integration of complex diagnostic information provided by both modalities that will lead to improved diagnostic performance.

We are developing enhanced volume rendering techniques for color image data. One target application is cryo-imaging, which provides whole-mouse, micron-scale, anatomical color, and molecular fluorescence image volumes by alternatively sectioning and imaging the frozen tissue block face. With the rich color images provided by cryo-imaging, we use true-color volume rendering and visually enhance anatomical regions by proper selection of voxel opacity. To compute opacity, we use color and/or gradient feature detection followed by suitable opacity transfer functions (OTF). An interactive user interface allows one to select from among multiple color and gradient feature detectors, OTF's, and their associated parameters, and to compute in live time new volume visualizations from within the Amira platform. We are also developing multi-resolution volume rendering techniques to accommodate extremely large (⪆60GB) cryo-image data sets. Together, these enhancements enable us to interactively interrogate cryo-image volume data and create useful renderings with "implicit segmentation" of organs.