• Medical Imaging
    Invitation
    Conferences
    Courses
    Precision Medicine Track
    Special Events
    PROSTATEx Challenge
    USCT Data Challenge
    Awards
    Sponsors
    Become a Sponsor
    Travel to Orlando
    Hotels - Orlando
    Onsite Services - Orlando
    Registration
    Proceedings
    For Authors and Presenters
    For Chairs and Committees
Renaissance Orlando at SeaWorld
Orlando, Florida, United States
11 - 16 February 2017
Print PageEmail Page

Hear the latest advances in Precision Medicine

Precision Medicine

The SPIE Medical Imaging Precision Medicine track highlights papers that showcase innovative ways to apply this multidimensional / multidisciplinary technology.

PRECISION MEDICINE:
An emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person is being highlighted in this year's program. Information regarding the Precision Medicine Initiative is available at: www.whitehouse.gov/precision-medicine.


Nanoparticle imaging probes for molecular imaging with computed tomography and application to cancer imaging
Paper 10132-32

Author(s):  Ryan K. Roeder, Univ. of Notre Dame (United States), et al.
Conference 10132: Physics of Medical Imaging
Session 7: Photon Counting I: Instrumentation
Date and Time: 2/14/2017 4:10:00 PM

Precision imaging is needed to realize precision medicine in cancer detection and treatment. Nanoparticle imaging probes and photon-counting detectors can act synergistically to enable quantitative molecular imaging with computed tomography (CT) in preclinical studies. Nanoparticle imaging probes have been designed for strong X-ray contrast, biostability, multimodal/multi-agent imaging, and targeted delivery. Examples will be presented to demonstrate quantitative molecular imaging of tumors, associated abnormalities (e.g., microcalcifications), multiple probe/tissue compositions, and cell populations overexpressing biomarkers using both CT and photon-counting spectral CT.


First experience with x-ray dark-field radiography for human chest imaging
Paper 10132-40

Author(s):  Peter B. Noel, Klinikum rechts der Isar der Technischen Univ. München (Germany), et al.
Conference 10132: Physics of Medical Imaging
Session 9: Phase Contrast Imaging
Date and Time: 2/15/2017 10:30:00 AM

To evaluate the performance of an experimental X-ray dark-field (XDF) radiography system for chest imaging in humans and to compare with conventional diagnostic CT imaging. A single human cadaver was imaged on the experimental XDF system. The XDF prototype operates at an acceleration voltage of up to 70 kVp and with a field-of-view large enough for clinical chest x-ray (>35x35cm2). It was feasible to extract a XDF signal of the whole human thorax. The performance of the experimental x-ray dark-field radiography setup offers, for the first time, obtaining multi-contrast chest x-ray images (attenuation and dark-field signal) from a human cadaver.


Affordable CZT SPECT with dose-time minimization
Paper 10132-55

Author(s):  James W. Hugg, Kromek (United States), et al.
Conference 10132: Physics of Medical Imaging
Session 11: Nuclear Medicine and Magnetic Resonance Imaging
Date and Time: 2/15/2017 5:10:00 PM

Pixelated CdZnTe (CZT) detector arrays are used in molecular imaging applications that can enable precision medicine, including small-animal SPECT, cardiac SPECT, molecular breast imaging (MBI), and general purpose SPECT. Advances have significantly improved efficiency of CZT-based molecular imaging systems and the cost has steadily declined. We have built a general purpose SPECT system using our 40 cm x 53 cm CZT gamma camera with 2 mm pixel pitch and characterized system performance. Compared to standard scintillator-based SPECT, CZT SPECT has many performance advantages. With CZT cost improving, affordable whole-body CZT general purpose SPECT is expected to enable precision medicine applications.


False dyssynchrony: problem with image-based cardiac functional analysis using x-ray computed tomography
Paper 10132-65

Author(s):  Katsuyuki Taguchi, The Johns Hopkins Univ. School of Medicine (United States), et al.
Conference 10132: Physics of Medical Imaging
Session 13: Modeling and Simulations I: CT
Date and Time: 2/16/2017 11:30:00 AM

We have developed a digitally synthesized patient which we call “Zach” (Zero millisecond Adjustable Clinical Heart) phantom, which allows for an access to the ground truth and assessment of image-based cardiac functional analysis (CFA) using CT images using clinically realistic settings. The study using Zach phantom revealed a major problem with image-based CFA: “False dyssynchrony.” Even though the true motion of wall segments is in synchrony, it may appear to be dyssynchrony with the reconstructed cardiac CT images. It is attributed to how cardiac images are reconstructed and how wall locations are updated over cardiac phases. The presence and the degree of false dyssynchrony may vary from scan-to-scan, which could degrade the accuracy and the repeatability (or precision) of image-based CT-CFA exams.


Detectability of artificial lesions in anthropomorphic virtual breast phantoms of variable glandular fraction
Paper 10132-68

Author(s):  Thomas J. Sauer, Carl E. Ravin Advanced Imaging Labs., Duke Univ. (United States), et al.
Conference 10132: Physics of Medical Imaging
Session 14: Modeling and Simulations II: Breast Imaging
Date and Time: 2/16/2017 1:40:00 PM

In previous work, we developed computational breast phantoms with normal anatomy. These include rule-based phantoms from FDA and patient-based phantoms from Duke. The goal of the current study is to develop multiple lesion models, demonstrate their insertion into the latest versions of these phantoms, and perform preliminary human and model observer studies. We insert simulated lesions into 10 of these phantoms each from Duke and FDA, representing a wide range of anatomical structures and content, with the intent of determining their detectability as affected by phantom type, lesion type, and local density.


A Patch-based CBCT scatter artifact correction using prior CT
Paper 10132-80

Author(s):  Xiaofeng Yang, Emory Univ. (United States), et al.
Conference 10132: Physics of Medical Imaging
Session PS1: Posters: Cone-Beam CT
Date and Time: 2/15/2017 5:30:00 PM

We have developed a novel patch-based cone beam CT (CBCT) artifact correction method based on prior CT images. The planning CT-based prior information was brought into the Bayesian deconvolution framework to perform the CBCT scatter artifact correction based on patch-wise nonlocal mean strategy. We evaluated the proposed correction method using a Catphan phantom with multiple inserts based on contrast-to-noise ratios (CNR) and signal-to-noise ratios (SNR), and the image spatial nonuniformity (ISN). All values of CNR, SNR and ISN in the corrected CBCT image were much closer to those in the planning CT images. The results demonstrated that the proposed CT-guided correction method could significantly reduce scatter artifacts and improve the image quality. This method has great potential to correct CBCT images allowing its use in adaptive radiotherapy.


Estimation of non-solid lung nodule volume with screening and sub-screening CT protocols: effect of reconstruction algorithm and measurement method
Paper 10132-97

Author(s):  Marios A. Gavrielides, U.S. Food and Drug Administration (United States), et al.
Conference 10132: Physics of Medical Imaging
Session PS2: Posters: CTI: New Technologies and Corrections
Date and Time: 2/15/2017 5:30:00 PM

Non-solid lung nodules are low density lesions that need to be monitored (typically with computed tomography, CT) for change in size. Volume measurements extracted with CT are being examined as a biomarker of lung nodule size; however, the effect of imaging protocols and measurement method on the volumetry of non-solid nodules has been under-investigated. In this study we examined the effect of different reconstruction algorithms and dose protocols on the volumetry of non-solid and solid nodules varying in size, and shape. The findings of the study can be valuable in developing standardized protocols and performance claims for non-solid nodules.


Spectral CT applicability to bariatric patient size
Paper 10132-120

Author(s):  Yoad Yagil, Philips Medical Systems Technologies Ltd. (Israel), et al.
Conference 10132: Physics of Medical Imaging
Session PS4: Posters: Photon Counting: Spectral CT, Insturmentation, and Algorithms
Date and Time: 2/15/2017 5:30:00 PM

Achieving the benefits of dual energy CT (DECT) for large patients is challenging, especially for the very large bariatric patients, due to the high attenuation of low energy X-ray photons. We use an extra-large phantom configuration to study the spectral performance of the IQon dual-layer based DECT system. Scans were performed at 120 and 140 kVp using standard and very low (36%) X-ray dose levels. Results show good image quality of virtual mono-energetic images at 120 and 140 kVp scans at both standard and low dose levels. Contrast enhancement at low keV and reduced beam hardening artifacts at high keV are demonstrated, showing that low dose DECT scans are beneficial even for very large patients.


Lung nodule volume quantification and shape differentiation with an ultra-high resolution technique on a photon counting detector CT system
Paper 10132-137

Author(s):  Shuai Leng, Mayo Clinic (United States), et al.
Conference 10132: Physics of Medical Imaging
Session PS4: Posters: Photon Counting: Spectral CT, Insturmentation, and Algorithms
Date and Time: 2/15/2017 5:30:00 PM

A new ultra high-resolution (UHR) mode has been established on the whole body photon counting-detector (PCD) CT system. This mode could achieve the pixel size of 0.25 mm by 0.25 mm at the iso-center, while conventional Macro mode was limited by 0.5 mm by 0.5 mm. A set of synthetic lung nodules were scanned using both UHR and Macro modes with 2 reconstruction kernels (4 combinations in total). We demonstrated that the UHR mode with sharp kernel (S80f) reconstruction demonstrated advantages of detecting nodule volumes and shapes. Our results provided the potential and feasibility of this acquisition mode on the PCD CT as a quantitative imaging biomarker for future clinical applications, such as lung and musculoskeletal imaging.


Quantification of the uncertainty in coronary CTA plaque measurements using dynamic cardiac phantom and 3D-printed plaque models
Paper 10132-197

Author(s):  Taylor W. Richards, Carl E. Ravin Advanced Imaging Labs., Duke Univ. (United States), et al.
Conference 10132: Physics of Medical Imaging
Session PS10: Posters: Observers, Modeling, and Phantoms
Date and Time: 2/15/2017 5:30:00 PM

This study's purpose was to quantify measurement uncertainty of stenosis using newly developed physical coronary plaque models. Coronary plaque models were designed and 3D-printed with tissue equivalent materials. Realistic cardiac motion was achieved using left ventricle volume-time curves to create synchronized heart motion profiles. All scans were acquired using a retrospective gating technique on a dual-source CT system. Images were reconstructed, vessel centerlines were determined, enhanced lumens were segmented, and measurement uncertainties were calculated. Stenosis measurement uncertainty increased with increasing heart rate. These results show the utility of the model to ascertain and optimize metrics of coronary CT image quality.


First in-vivo x-ray dark-field chest radiography: a feasibility study in a living pig
Paper 10132-204

Author(s):  Peter B. Noel, Klinikum rechts der Isar der Technischen Univ. München (Germany), et al.
Conference 10132: Physics of Medical Imaging
Session PS11: Posters: Phase Contrast and Dark Field Imaging
Date and Time: 2/15/2017 5:30:00 PM

We present the first in-vivo x-ray dark-field (XDF) full-field chest radiography (35x35cm2) of a living pig at acquisition parameters suitable for in-vivo imaging and clinical demands (scan time of 40 s, estimated effective dose of 80 µSv). In order to acquire those images, we developed a novel high-energy XDF scanner, which overcomes the limitations of currently established setups. We consider this as a milestone in the bench-to-bedside translation of XDF imaging as we successfully demonstrated the translation of XDF radiography from mouse to human-sized animals. If the current pre-clinical benefits remain, XDF will be valuable diagnostic tool in clinical practice.


Fully automated lobe-based airway taper index calculation in a low dose MDCT CF study over 4 time-points
Paper 10133-30

Author(s):  Oliver Weinheimer, Ruprecht-Karls-Univ. Heidelberg (Germany), et al.
Conference 10133: Image Processing
Session 6: Quantitative Image Analysis
Date and Time: 2/13/2017 11:30:00 AM

Bronchiectasis is a lung disease where parts of the airways are permanently dilated. The development and the progression of bronchiectasis is not evenly distributed over the entire lungs. Accurate lobe-based detection and monitoring of the course of bronchiectasis is needed for individualized treatment. We developed a fully automated method for the precise calculation of lobe-based airway taper indices (ATI). ATI was applied to 144 volumetric inspiratory low-dose MDCT scans. The new method identified enlarged airway parts in the segmented airway trees. ATI can be an important tool for monitoring the progression and the individual treatment of patients with bronchiectasis.


Pseudo CT estimation from MRI using patch-based random forest
Paper 10133-68

Author(s):  Xiaofeng Yang, Emory Univ. (United States), et al.
Conference 10133: Image Processing
Session PS1: Posters
Date and Time: 2/13/2017 5:30:00 PM

We propose a pseudo CT estimation method from MR images, which is based on patch-based random forest. Patient-specific anatomical features are extracted from aligned training images. The well-trained random forest was used to predict the pseudo CT of the new patient. Our prediction technique was validated with human brain images. The prediction accuracy of our approach was assessed using the original CT images. Experimental results showed the proposed method could accurately predict CT images from MR images. We have developed a new pseudo CT prediction method based on patch-based random forest, demonstrated its clinical feasibility, and validated its prediction accuracy. This pseudo CT prediction technique could be a useful tool for MRI-based radiation treatment planning and attenuation correction in PET/MRI scanner.


Personalized design and virtual evaluation of physician-modified stent grafts for juxta-renal abdominal aortic aneurysms
Paper 10133-85

Author(s):  Prahlad G. Menon, Duquesne Univ. (United States), et al.
Conference 10133: Image Processing
Session PS1: Posters
Date and Time: 2/13/2017 5:30:00 PM

We present a computation suite to virtually design, deploy and virtually test endovascular aneurysm repair (EVAR) stent grafts in a juxtarenal aortic aneurysms (JAAs). We demonstrate its functionality in a parametrically designed JAA, which was modeled using a free-form mesh deformation tool upon a patient-specific aortic model with bilateral renal artery stenosis. Further, internal hemodynamics of the baseline abdominal aorta, its equivalent JAA model and its personalized post-EVAR geometry were compared using computational fluid dynamics (CFD), in terms of flow patterns, renal outflow rates, wall shear stress and energy dissipation, to illustrate viability of the virtually designed EVAR graft.


Automatic polyp detection in colonoscopy videos
Paper 10133-94

Author(s):  Zijie Yuan, Arizona State Univ. (United States), et al.
Conference 10133: Image Processing
Session PS1: Posters
Date and Time: 2/13/2017 5:30:00 PM

Colon cancer is the second cancer killer in the US. Colonoscopy is the primary method for screening and prevention of colon cancer, but during colonoscopy, a significant number (25%) of polyps (precancerous abnormal growths inside of the colon) are missed; therefore, the goal of our research is to reduce the polyp miss-rate of colonoscopy. This paper presents a method to detect polyp automatically in a colonoscopy video. Our proposed system has two stages: Candidate generation that produces candidate patches, and candidate classification that sorts out candidates into polyp or non-polyp via convolutional neural networks (CNNs).


Computer-aided detection of bladder masses in CT urography
Paper 10134-2

Author(s):  Kenny H. Cha, Univ. of Michigan (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 1: Pelvis
Date and Time: 2/13/2017 8:20:00 AM

We are developing a computer-aided detection system for bladder cancer in CTU. The bladder was segmented using a deep-learning convolutional neural network with level sets. The bladder wall was extracted and transformed into a straightened bladder wall profile, which was analyzed to identify lesion candidates. The candidates were automatically segmented and morphological features were extracted from each candidate. LDA classifier was trained using leave-one-out method to differentiate between true and false lesions. The system achieved a sensitivity of 86.5% at 3.3 FPs/case for the training set, and 84.6% at 3.8 FPs/case for the test set.


Bladder cancer treatment response assessment using deep learning in CT with transfer learning
Paper 10134-3

Author(s):  Kenny H. Cha, Univ. of Michigan (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 1: Pelvis
Date and Time: 2/13/2017 8:40:00 AM

We are developing a CAD system for bladder cancer treatment response assessment in CT. We compared the performance of deep-learning convolution neural networks (DL-CNN) of different network sizes, and with and without transfer learning using natural scenes images or ROIs inside and outside the bladder. With 104 temporal lesion pairs from 87 lesions in 82 patients, two-fold cross-validation was performed to train the DL-CNN to differentiate responders and non-responders. Our study demonstrated the feasibility of using DL-CNN for treatment response assessment using different network structures, and that transfer learning did not improve the accuracy of assessment.


The effects of slice thickness and radiation dose level variations on computer-aided diagnosis (CAD) nodule detection performance in pediatric chest CT scans
Paper 10134-10

Author(s):  Nastaran Emaminejad, Univ. of California, Los Angeles (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 2: Lung I
Date and Time: 2/13/2017 11:30:00 AM

Performing the lowest possible dose CT scan while still accomplishing the desired clinical task is an important goal and is especially important in pediatric populations. In this study our purpose was to investigate the effects of dose reduction and different slice thicknesses on a CAD system’s lung nodule detection performance for Pediatric Chest CT scans. Reduced-dose raw data was generated from raw data of 58 CT scans by a noise addition simulator. CAD tool performed nodule detection task on reconstructed image data. Reduced-dose data had relatively similar detection rate, and thinner slice thickness was more influenced by dose level variations.


Deep ensemble learning of virtual endoluminal views for polyp detection in CT colonography
Paper 10134-15

Author(s):  Hiroyuki Yoshida, Massachusetts General Hospital (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 3: Colon and G.I.
Date and Time: 2/13/2017 2:20:00 PM

We developed and evaluated an ensemble deep convolutional neural network (DCNN) scheme (EDN) for learning of virtual endoluminal (VE) images to improve the performance of computer-aided detection (CADe) of polyps in CT colonography (CTC). Nine different types of renderings were generated for the VE images of polyp candidates detected by a conventional CADe system. We re-trained eleven DCNNs representing three types of publically available DCNN models to identify polyps from the VE images. An EDN was developed by combining these DCNNs using a random forest classifier as the meta-classifier. Evaluation results on 154 clinical CTC cases showed that the EDN was effective in improving the detection performance of CADe in CTC, especially in small polyps.


Computer-assisted optical biopsy for colorectal polyps
Paper 10134-18

Author(s):  Fernando Navarro, Technische Univ. München (Germany), et al.
Conference 10134: Computer-Aided Diagnosis
Session 3: Colon and G.I.
Date and Time: 2/13/2017 3:20:00 PM

We propose a method for computer-assisted optical biopsy for colorectal polyps, with the final goal of assisting the medical expert during the colonoscopy. Our approach is based on recent advancements in convolutional neural networks for image representation. We use features obtained from a pre-trained CNN and a random forest classier. For our study 776 polyp images were acquired and histologically analysed after polyp resection. We report a performance increase w.r.t. conventional engineered features and comparable results to a state-of-the art method using 3D shape features.


Computer-aided theragnosis using quantitative ultrasound methods and convolutional neural networks
Paper 10134-29

Author(s):  Mehrdad J. Gangeh, Univ. of Toronto (Canada), et al.
Conference 10134: Computer-Aided Diagnosis
Session 5: Breast I
Date and Time: 2/14/2017 11:50:00 AM

A non-invasive computer-aided-theragnosis (CAT) system was developed for the early assessment of responses to neoadjuvant chemotherapy in patients with locally advanced breast cancer (LABC). The CAT system was based on quantitative ultrasound spectroscopy methods in conjunction with deep learning techniques. This study, for the first time, applied convolutional neural networks (CNNs) in the application of therapeutic cancer response assessment in order to monitor the effectiveness of neoadjuvant chemotherapy early during the course of treatment administration. The results of the classification using the developed CAT system based on deep learning indicated an improvement of performance up to 10% compared to a CAT system with handcrafted features using the LBPs.


Quantification of CT images for the classification of high- and low-risk pancreatic cysts
Paper 10134-31

Author(s):  Jayasree Chakraborty, Memorial Sloan-Kettering Cancer Ctr. (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 6: Liver and Abdomen
Date and Time: 2/14/2017 1:40:00 PM

Pancreatic cancer is the most lethal cancer. Early detection at a pre-cancerous stage is the best tool for preventing this disease. Intraductal papillary mucinous neoplasms represent the only radiographically identifiable precursor lesion of pancreatic cancer. Standard imaging criteria determine risk groups: low-risk patients are regularly observed, while high-risk patients undergo resection. We propose objective classification of risk groups based on quantitative imaging features. We apply new features representing the solid component (i.e. areas of high intensity) and extract standard texture features. Solid component and texture features achieve AUC of 0.71 with accuracy of 70.8% and 0.73 and 77.3%, respectively.


Preoperative assessment of microvascular invasion in hepatocellular carcinoma
Paper 10134-34

Author(s):  Jayasree Chakraborty, Memorial Sloan-Kettering Cancer Ctr. (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 6: Liver and Abdomen
Date and Time: 2/14/2017 2:40:00 PM

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related death worldwide. Microvascular invasion (MVI) is the most important predictor of early recurrence. There are no preoperative tests for MVI. Preoperative identification of MVI would optimally select patients for resection or liver transplant, and potentially expand the eligibility of patients for transplant. This paper presents quantitative analysis of CT texture to identify MVI. We describe a two-stage classification of HCCs into uniform and heterogeneous groups followed by MVI detection with accuracy of 76.7% and 74.06% and area under the ROC curve of 0.76 and 0.79, respectively.


Validation of an image registration and segmentation method on ECG-gated CT data of a physical dynamic stent graft model
Paper 10134-42

Author(s):  Maaike Koenrades, Univ. of Twente (Netherlands), et al.
Conference 10134: Computer-Aided Diagnosis
Session 9: Vessels
Date and Time: 2/15/2017 10:10:00 AM

Knowledge on dynamics of implanted stent grafts is sparse yet of vital importance in the prediction of failure and to improvement of stent grafts. We have evaluated the accuracy of a previously proposed segmentation and registration algorithm to detect motion of endovascular stent grafts on ECG-gated 4D-CT data using a physical dynamic stent graft model. Motion patterns obtained from the CT data were validated against motion patterns obtained from camera recordings. We show that the combined algorithm performs sufficiently well for the detection of aortic stent graft motion which could aid in the prediction and prevention of stent graft failure.


IDH mutant status assessment of glioma using three-dimensional texture features of multimodal MR images
Paper 10134-61

Author(s):  Xi Zhang, Fourth Military Medical Univ. (China), et al.
Conference 10134: Computer-Aided Diagnosis
Session 13: Brain
Date and Time: 2/16/2017 10:50:00 AM

IDH mutant status is critical to treatment strategy and prognosis prediction of patients with glioma. We aimed to verify the hypothesis that appropriate texture features derived from multimodal MR images may be used for noninvasive and well-repeatable detection of IDH mutant status. In this study, 410 3D GLCM and GLGCM textural features extracted from ten multimodal MRI can reflect the difference between gliomas with and without IDH mutation. With sample augmentation and feature selection, an optimal subset of 42 features was selected and its AUC for mutation detection reached to 0.9416, indicated that it may be a potential imaging biomarkers for predicting IDH mutant status.


Radiogenomic analysis of lower grade glioma: a pilot multi-institutional study shows an association between quantitative image features and tumor genomics
Paper 10134-62

Author(s):  Ashirbani Saha, Duke Univ. (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 13: Brain
Date and Time: 2/16/2017 11:10:00 AM

Recent studies showed the effectiveness of genomic analysis of lower grade gliomas (LGG) for stratification of patients into groups with different prognosis and proposed specific genomic classifications. We studied the association of one of those genomic classifications with imaging features on a multi-institutional dataset. Our novel radiogenomic analysis using 2D and 3D shape features of tumor showed that there is a strong association between the tumor cluster-of-clusters subtype and two imaging features. This result shows high promise for the potential use of imaging as a surrogate measure for genomics in the decision process regarding treatment of lower grade glioma patients.


Radiogenomic analysis of hypoxia pathway reveals computerized MRI descriptors predictive of overall survival in glioblastoma
Paper 10134-63

Author(s):  Niha G. Beig, Case Western Reserve Univ. (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 13: Brain
Date and Time: 2/16/2017 11:30:00 AM

Glioblastoma Multiforme (GBM) is a highly aggressive brain tumor with a median survival of 14 months. Hypoxia is a predominant feature in GBM, and is known to be associated with tumor growth, and resistance to conventional therapy. In this study, we hypothesized that radiomic descriptors can capture molecular variations of tumor hypoxia on routine MRI that are otherwise not appreciable; and can discriminate patients with short-term survival (STS, overall survival (OS) < 7 months), mid-term (MTS) (7 months< OS16 months).


Radiomics biomarkers for accurate tumor progression prediction of oropharyngeal cancer
Paper 10134-68

Author(s):  Lubomir M. Hadjiiski, Univ. of Michigan Health System (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 14: Head and Neck
Date and Time: 2/16/2017 2:20:00 PM

We are developing a decision support system that merges available radiomics, histopathologic and molecular biomarkers in a predictive model based on statistical outcomes of previous cases and machine learning to assist clinicians in assessing oropharyngeal tumor progression. Preliminary test AUCs of 0.87, 0.74, and 0.71 were achieved with the prediction models in the individual domains of radiomics, histopathologic, and molecular biomarkers, respectively. Combining predictions in all 3 domains increased the test AUC to 0.94. The integrated multi-domain approach shows promise for assisting in tumor progression prediction.


Quantitative analysis of CT attenuation distribution patterns of nodule components for pathologic categorization of lung nodules
Paper 10134-71

Author(s):  Chuan Zhou, Univ. of Michigan Health System (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session 15: Lung II
Date and Time: 2/16/2017 3:50:00 PM

A quantitative analysis method was developed to characterize the CT attenuation distribution patterns of nodule components with radiomics features for the classification of different pathological categories of nodules. One hundred and three cases from the NLST database were used in this study. Three logistic regression models were built using leave-one-case-out resampling and receiver operating characteristic (ROC) analysis to differentiate invasive nodules from (1) pre-invasive nodules, (2) benign nodules, and (3) the group of pre-invasive and benign nodules. The test area under the ROC curve (AUC) achieved 0.877±0.036, 0.811±0.041 and 0.833±0.037, respectively, demonstrating the feasibility of pathologic categorization using radiomics features.


Identifying prognostic imaging biomarkers by extracting high-risk volumes of glioblastoma on multi-parametric MRI
Paper 10134-84

Author(s):  Ruijiang Li, Stanford Univ. School of Medicine (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS1: Posters: Brain
Date and Time: 2/15/2017 5:30:00 PM


Improving utility of brain tumor confocal laser endomicroscopy: objective value assessment and diagnostic frame detection with convolutional neural networks
Paper 10134-85

Author(s):  Mohammadhassan IzadyYazdanabadi, Barrow Neurological Institute, St. Joseph's Hospital and Medical Ctr. (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS1: Posters: Brain
Date and Time: 2/15/2017 5:30:00 PM

Handheld, portable Confocal Laser Endomicroscopy (CLE) is being explored for neurosurgery of brain tumors because of its ability to image histopathological features of the tissue in real time during surgery. We used 16795 CLE images in a 4-fold cross validation manner to see if AlexNet can predict diagnostic value of CLE images and detect diagnostic ones. Our results show that a deeply trained AlexNet network can predict the diagnostic value of new images with 91% accuracy. Our method is also fast enough to be smoothly integrated with the speed and precision of CLE imaging application in precision brain tumor surgery.


Automatic classification of cardioembolic and arteriosclerotic ischemic strokes from apparent diffusion coefficient datasets using texture analysis and deep learning
Paper 10134-86

Author(s):  Javier Villafruela, Univ. of Calgary (Canada), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS1: Posters: Brain
Date and Time: 2/15/2017 5:30:00 PM

Stroke is a leading cause of death and disability in the western hemisphere. The ability to determine the exact phenotype of an acute ischemic stroke is highly relevant for optimal treatment decision and preventing recurrent events. However, the differentiation of atherosclerotic and cardioembolic phenotypes can be especially challenging due to similar appearance and symptoms. The aim of this study was to develop and evaluate the feasibility of an image-based deep learning approach for discriminating between arteriosclerotic and cardioembolic acute ischemic strokes using ADC datasets from acute stroke patients.


Does the prediction of breast cancer improve using a combination of mammographic density measures compared to individual measures alone?
Paper 10134-93

Author(s):  Susan M. Astley, The Univ. of Manchester (United Kingdom), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS2: Posters: Breast
Date and Time: 2/15/2017 5:30:00 PM

We investigate the relationship between breast density and breast cancer risk by combining three methods of measuring breast density. Methods agreed 44% of the time, while 8.6% had a combination of extremes or a differ category for each density method. Logistic regression showed that visual assessment by visual analogue scales (VAS) was better at predicting breast cancer risk than automated methods, and that this was only marginally improved by using a combination of density measures.


Using estimated weight to predict breast cancer risk
Paper 10134-101

Author(s):  Susan M. Astley, Ctr. for Imaging Science, The Univ. of Manchester (United Kingdom), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS2: Posters: Breast
Date and Time: 2/15/2017 5:30:00 PM

Personalised screening requires assessment of individual cancer risk, and hence weight. Breast fat volume was measured in 40431 women for whom 10-year Tyrer-Cuzick risk using self-reported weight was known. Each woman’s weight was also estimated from the relationship between self-reported weight and fat volume in the cohort; estimated weights were used to re-calculate risk. Women were assigned to risk categories (low<2%, population 2-3.49%, moderate 3.5-4.9%, high>5%) by both methods. 1852 women (4.6%) moved one category when using estimated weight. 72.8% of changes were between low and population risk, indicating that estimated weight may be suitable for personalized screening risk assessment.


Development and validation of a radiomics nomogram for progression-free survival prediction in stage IV EGFR-mutant non-small cell lung cancer
Paper 10134-133

Author(s):  Jie Tian, Key Lab. of Molecular Imaging, Institute of Automation, Chinese Academy of Sciences (China), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS8: Posters: Lung
Date and Time: 2/15/2017 5:30:00 PM

As reported by the clinical practice guideline of American Society of Clinical Oncology (ASCO), TKIs are increasingly recommended for stage IV EGFR-mutant NSCLC. Unfortunately, most patients eventually develop therapeutic resistance, but how to assess the resistance of TKIs for individual remains a challenge. In this study, we developed a noninvasive individualized nomogram integrated a CT phenotypic classifier and three clinicopathological risk factors for clinical prediction of progression-free survival, which presented excellent agreement in the different datasets. The proposed nomogram could be further served in clinic aided decision to integrate a standardized, instant and global manner decision-support system for the clinic.


Volume calculation of CT lung lesions based on Halton low-discrepancy sequences
Paper 10134-137

Author(s):  Liansheng Wang, Xiamen Univ. (China), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS8: Posters: Lung
Date and Time: 2/15/2017 5:30:00 PM

In this paper, Halton low-discrepancy sequence is applied to calculate the volume of CT lung lesions. Compared with LSTK, the proposed method can calculate volume without reconstructing the 3D model of lung lesions and surface triangulating. Our method is particularly suitable for medical applications with a large number of two-dimensional radiographic images. The proposed method can generate more uniform random points compared with MC method. The experimental results are demonstrated that our proposed method can achieve more accurate results compared with LSTK and MC method. Clinically, our proposed method can be easily generalized to calculate lesion volumes of other organs.


Applying radiomics approach for predicting tumor response to chemotherapy at early stage: an initial study for ovarian cancer patients
Paper 10134-151

Author(s):  Yuchen Qiu, The Univ. of Oklahoma (United States), et al.
Conference 10134: Computer-Aided Diagnosis
Session PS10: Posters: Pelvis
Date and Time: 2/15/2017 5:30:00 PM


Towards quantitative quasi-static elastography with a gravity-induced deformation source
Paper 10135-1

Author(s):  Rebekah H. Griesenauer, Vanderbilt Univ. (United States), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session 1: Modeling Tissue Deformation
Date and Time: 2/14/2017 8:00:00 AM

Precise applications of stress-strain relationships of tissue within the breast can improve the accuracy of biomechanical models that attempt to simulate breast deformations. Quantitative determination of patient specific breast tissue mechanical properties is highly desirable for applications in image registration and image guided therapies. In this study, we developed a stiffness estimation method that is performed using deformations representative of supine therapeutic procedures. The method proposed is workflow friendly, quantitative, and uses a non-contact, gravity-induced deformation source. In a clinical context, the elastography method was shown to be promising for use in biomechanical model assisted supine procedures.


3D surface estimation from sparse electrophysiological measurements and integration with preoperative MRI in deep brain stimulation surgery
Paper 10135-16

Author(s):  Andreas Husch, Ctr. Hospitalier de Luxembourg (Luxembourg), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session 3: Neurosurgical Procedures
Date and Time: 2/14/2017 2:40:00 PM


Image-guided smart laser system for precision implantation of cells in cartilage
Paper 10135-30

Author(s):  Nitesh Katta, The Univ. of Texas at Austin (United States), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session 7: Optical Sensing
Date and Time: 2/15/2017 1:20:00 PM

We describe a “smart laser knife” that combines optical coherence tomography (OCT) with a 100 nanosecond-pulsed thulium (Tm) laser for precise seeding of cells into diseased cartilage. OCT identifies damaged areas with micrometer resolution and provides real-time feedback of the cartilage tissue. The co-axial Tm laser ablates strategically-placed microwells. A hydrogel seeded with human mesenchymal stem cells is then deposited into the microwells with an image-guided injection needle. A bench-top system demonstrates operation of the smart laser knife and associated procedures.


Concentric agonist-antagonist robots for minimally invasive surgeries
Paper 10135-36

Author(s):  Kaitlin P. Oliver Butler, The Univ. of Tennessee Knoxville (United States), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session 8: Novel Robots and Robotic Procedures
Date and Time: 2/15/2017 3:50:00 PM

We present a design concept for a Concentric Agonist-Antagonist Robot, a parallel robot comprised of two concentric tubes with selectively removed material that is actuated with push-pull, agonist-antagonist motion. It avoids the instability issues that arise from actuating precurved, concentric tubes while maintaining the open lumen and simple construction of traditional concentric-tube continuum robots. Highly adaptable, this design attains a large range of motion and can be modified to achieve various workspaces and end-effector orientations. Paired with additive manufacturing, it has the potential to offer cost-effective customization for patients in the fields of minimally invasive surgery and precision medicine.


Patient-specific indirectly 3D printed mitral valves for pre-operative surgical modelling
Paper 10135-42

Author(s):  Olivia Ginty, Robarts Research Institute (Canada), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session 9: Cardiac Procedures
Date and Time: 2/16/2017 8:20:00 AM

Significant mitral valve regurgitation affects approximately two percent of the population [1]. Approximately five percent of mitral valve repair surgeries have non-trivial mitral regurgitation (MR grade 2 or higher) post-surgery. In this work, we propose the use of diagnostic trans-esophageal 3D image data and 3D printing technology to create patient specific mitral valve models. These models are used in a dynamic, beating heart simulator to replicate patient pathology. Our goal is to use models to assess different repair options prior to surgery, in the hope of optimizing patient outcomes. The current work reports preliminary results comparing systolic and diastolic modelling.


Patient specific atrial models for pre-procedure surgical planning
Paper 10135-45

Author(s):  Justin Laing, Western Univ. (Canada), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session 9: Cardiac Procedures
Date and Time: 2/16/2017 9:20:00 AM

Minimally invasive cardiac procedures requiring a trans-septal puncture such as atrial ablation and MitraClip© mitral valve repair are becoming increasingly common. Such beating heart therapies rely heavily on ultrasound and fluoroscopic imaging. However, in many cases with complex or abnormal patient morphology, image guidance can be time-consuming and difficult, leading to excessive x-ray and contrast exposure for the patient. We have developed a workflow for creating patient-specific models for left and right atria using realistic tissue mimicking materials. These models can be used for both training and practice for complex cases.


Physiology informed virtual surgical planning: a case study with a virtual airway surgical planner and BioGears
Paper 10135-64

Author(s):  Lucas N. Potter, Old Dominion Univ. (United States), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session PS1: Posters
Date and Time: 2/15/2017 5:30:00 PM

Upper airway stenosis is a common condition whose diagnosis and treatment are often confounded by institutional factors such as clinicians experience or preference as opposed to objective metrics. The Virtual Pediatric Airway Workbench (VPAW) –a surgical planner for tracheal stenosis, addresses this by providing fluid simulation and tracheal geometry related objective measures. VPAW however lacks physiological information that could change the outcomes of a surgical plan. In this work we performed a case study by integrating a human physiological engine BioGears with VPAW. This offered additional information in order to plan patient-specific surgery modeling in a holistic sense.


Patient identification using a near-infrared laser scanner
Paper 10135-92

Author(s):  Jirapong Manit, Univ. zu Lübeck (Germany), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session PS1: Posters
Date and Time: 2/15/2017 5:30:00 PM

We propose to use a novel near-infrared (NIR) laser-based head tracking system to determine the tissue thickness of a person's forehead and use it as a biometric feature. Based on spatial and tissue thickness measurements from MRI, we created 4D reference point clouds for 30 subjects. The registration of NIR measurements with these point clouds yielded spatial and tissue thickness error features. In a 2D feature space, the lowest feature distances were, with 100% accuracy, the ones from same subject measurement-point cloud registrations. This new approach could simplify patient verification and lays the foundation for a future human identification technique.


Real- time liver tumour tracking with low contrast radiopaque markers
Paper 10135-101

Author(s):  Sankar Arumugam, Liverpool and Macarthur Cancer Therapy Ctrs. (Australia), et al.
Conference 10135: Image-Guided Procedures, Robotic Interventions, and Modeling
Session PS1: Posters
Date and Time: 2/15/2017 5:30:00 PM

Accurate positioning of the target volume is paramount for Stereotactic Body Radio Therapy (SBRT). In liver SBRT planning the CT dataset of patient implanted with radio opaque markers are used to define the internal target volume (ITV) in radiotherapy planning and CBCT based pre-treatment position verification before the treatment. An improved auto segmentation methodology that will identify the position of a low contrast radiopaque fiducial marker has been investigated in this study. The developed methodology has been implemented in an in-house developed real-time 3D position monitoring system, SeedTracker, for liver SBRT .The overall performance of SeedTracker software was evaluated using a retrospective analysis of CBCT images, acquired with the Elekta-XVI system, of five liver cancer patients implanted with embolization coils. The system showed an auto seed segmentation True Positive Rate (TPR) of 96%. The 3D seed and plan isocentre positions estimated by SeedTracker agreed with the XVI 4D reconstructed data with a mean (σ) difference of 0.2(1.8) mm. The developed system has a potential application of monitoring target position during treatment delivery in liver SBRT.


Can BI-RADS features on mammography be used as a surrogate for expensive genomic testing in breast cancer patients?
Paper 10136-57

Author(s):  Michael R. Harowicz, Duke Univ. School of Medicine (United States), et al.
Conference 10136: Image Perception, Observer Performance, and Technology Assessment
Session PS5: Posters: Image Perception and Technology Assessment in Breast Imaging
Date and Time: 2/13/2017 5:30:00 PM

Molecular subtype and the Oncotype Dx recurrence score (ODRS) are used to properly assign patients to treatment. Radiogenomics may offer a less expensive and less invasive surrogate for these genomic tests through the use of Breast imaging-reporting and data system (BI-RADS) features on mammography. This retrospective study analyzes the association between BI-RADS shape and margin features on mammography with molecular subtype and ODRS. Our results show a statistically significant association between BI-RADS shape features and subtype (p=0.017), while finding that BI-RADS features were not significantly associated with ODRS.


Use of patient specific 3D printed neurovascular phantoms to evaluate the clinical utility of a high resolution x-ray imager.
Paper 10137-17

Author(s):  Swetadri Vasan Setlur Nagesh, Toshiba Stroke and Vascular Research Ctr. (United States), et al.
Conference 10137: Biomedical Applications in Molecular, Structural, and Functional Imaging
Session 5: Fluid and Cardiovascular
Date and Time: 2/13/2017 8:00:00 AM

Modern 3D printing technology can create vascular phantoms accurately based on actual human patient anatomies facilitating a realistic simulation environment for interventions. We present two experimental setups combining 3D printed patient-specific neurovasculature with anthropomorphic head phantoms, and with human skull and acrylic plates to simulate different patient pathologies. These setups are used to perform and compare various interventional treatments under x-ray guidance using a high resolution Complementary-Metal-Oxide-Semiconductor-based Micro-Angiographic-Fluoroscope (MAF-CMOS) and a state-of-the-art Flat Panel Detector (FPD). The images from these will be scored by blinded observers to evaluate the clinical utility of the high resolution MAF-CMOS detector.


Phenotypic feature quantification of patient derived 3D cancer spheroids in fluorescence microscopy image
Paper 10137-29

Author(s):  Mi-Sun Kang, Ewha Womans Univ. (Korea, Republic of), et al.
Conference 10137: Biomedical Applications in Molecular, Structural, and Functional Imaging
Session 7: Innovations in Image Processing I
Date and Time: 2/13/2017 1:40:00 PM

We present a cell image quantification method for image-based drug response prediction from patient-derived glioblastoma cells. To this end, we first performed image stitching to create an image of each well of the plate with the same environment. To automatically detect the colonies, we used a learning based classification algorithm. The nuclear intensity and morphological characteristics were used for individual nuclei segmentation. Next, we calculated the location correlation of each cell that is appeal of the cell density in the well environment. Finally, we compared the results for drug-treated and untreated cells. This technique could potentially be applied for drug screening and quantification of the effects of the drugs.


Disease quantification on PET/CT images without object delineation
Paper 10137-30

Author(s):  Yubing Tong, Univ. of Pennsylvania School of Medicine (United States), et al.
Conference 10137: Biomedical Applications in Molecular, Structural, and Functional Imaging
Session 7: Innovations in Image Processing I
Date and Time: 2/13/2017 2:00:00 PM

This paper presents a novel approach to disease quantification (DQ) via PET/CT images without performing explicit image segmentation. The concept of an object-dependent disease map is introduced to express disease severity without performing explicit delineation of either the object or the lesions and partial volume correction. Our results based on 20 each of PET/CT NEMA phantom data, liver lesions, and lung lesions demonstrate better than 2% agreement on phantoms and better than 8% overall agreement with ground truth measurements on patient images. We are currently exploring extensions of this approach to body-wide applications.


Tracking cancer-specific T-cells in vivo with gold nanoparticles and CT imaging
Paper 10137-44

Author(s):  Rinat Meir, Bar-Ilan Univ. (Israel), et al.
Conference 10137: Biomedical Applications in Molecular, Structural, and Functional Imaging
Session 9: Novel Imaging Methods
Date and Time: 2/14/2017 11:30:00 AM

Application of immune cell-based cancer therapy in routine clinical practice is challenging, due to the poorly-understood mechanisms underlying success or failure of treatment. A novel method was designed for longitudinal and quantitative in vivo cell tracking, based on the superior visualization abilities of classical X-ray CT, combined with state-of-the-art nanotechnology. This new method for cell tracking offers a valuable tool for studying the fate of immune cells in cancer immunotherapy. We were able to track cancer-specific T-cells labeled with gold nanoparticles and to examine the distribution, migration and kinetics of T-cells as they accumulated at the tumor site.


Nonrigid 2D registration of whole fluoroscopic coronary artery image sequence with periodic deformation field
Paper 10137-67

Author(s):  Taewoo Park, Hankuk Univ. of Foreign Studies (Korea, Republic of), et al.
Conference 10137: Biomedical Applications in Molecular, Structural, and Functional Imaging
Session PS4: Posters: Fluid and Cardiovascular
Date and Time: 2/13/2017 5:30:00 PM


Photothermal characterization of gold nanorods for superficial breast cancer therapy in an engineered 3 dimensional human decellularized adipose tissue platform
Paper 10137-73

Author(s):  Ki-Hwan Nam, Korea Basic Science Institute (Korea, Republic of), et al.
Conference 10137: Biomedical Applications in Molecular, Structural, and Functional Imaging
Session PS6: Posters: Innovations in Image Processing
Date and Time: 2/13/2017 5:30:00 PM

An engineered 3D human decellularized adipose tissue (hDAT) platform was fabricated to characterize a heating method and to control the generated heat in the tissue used for the clinical treatment of superficial breast cancer therapy using gold nanorods (GNRs, citrate-coated) irradiated near-infrared laser. The hDAT was designed to mimic a localized tumor in the human breast tissue and to measure temperature and heat distribution of the GNRs according to the variables (concentration, laser power density, and tissue depth) on the cross-section of the tissue. The hDAT platform will provide an optimized condition for clinical treatment of photothermal cancer therapy.


Early classification of Alzheimer's disease using hippocampal texture from structural MRI
Paper 10137-85

Author(s):  Yong Liu, Institute of Automation (China), et al.
Conference 10137: Biomedical Applications in Molecular, Structural, and Functional Imaging
Session PS7: Posters: Machine Learning
Date and Time: 2/13/2017 5:30:00 PM

In this study, we specifically to investigate if the proposed hippocampal textures (intensity, shape, texture) can be served as an MRI biomarker of AD. MRI data was obtained from 48 AD and 38 normal samples. All features of the training data are ranked using t-test ranking method; and Fisher criterion between AD and NC were introduced for selecting optimal parameters for classification based on support vector machine(SVM). The result highlights the presence of hippocampal texture abnormalities in AD, and the possibility that texture may serve as a neuroimaging biomarker for AD.


Phenotype analysis of early risk factors from electronic medical records improves image-derived diagnostic classifiers for optic nerve pathology
Paper 10138-13

Author(s):  Shikha Chaganti, Vanderbilt Univ. (United States), et al.
Conference 10138: Imaging Informatics for Healthcare, Research, and Applications
Session 3: Precision Medicine
Date and Time: 2/15/2017 1:40:00 PM

We examine CT imaging and EMR data of 392 subjects with disorders of the optic nerve. We developed an automated image-processing pipeline that identifies the orbital structures within the human eye, and calculates descriptive measurements. We customized the PheWAS study to derive diagnostic EMR phenotypes. We used random forest classifiers to evaluate the predictive power of image-derived markers, EMR phenotypes, and clinical visual assessments in identifying disease cohorts from a control group. The addition of EMR phenotypes to the imaging markers improves the classification accuracy against controls. This study illustrates the importance of diagnostic context for interpretation of image-derived markers.


The development and implementation of MOSAIQ Integration Platform (MIP) based on the radiotherapy workflow
Paper 10138-14

Author(s):  Xin Yang, Sun Yat-Sen Univ. Cancer Ctr. (China), et al.
Conference 10138: Imaging Informatics for Healthcare, Research, and Applications
Session 3: Precision Medicine
Date and Time: 2/15/2017 2:00:00 PM

To meet the special demands in China and the particular needs for the radiotherapy department, a MOSAIQ Integration Platform CHN (MIP) based on the workflow of radiation therapy (RT) has been developed. It is a key to facilitate the information sharing and department management.


Feasibility of fabricating personalized 3D-printed bone grafts guided by high-resolution imaging
Paper 10138-22

Author(s):  Abigail Hong, Univ. of Pennsylvania (United States), et al.
Conference 10138: Imaging Informatics for Healthcare, Research, and Applications
Session 6: Novel Applications in 3D Printing
Date and Time: 2/16/2017 10:10:00 AM

We propose a synthetic, 3D-printed alternative to the standard autograft or allograft method to treat critical size bone defects, specifically in reference to the proximal femur. The basis of our patient-specific graft will be clinically obtained, high-resolution images that will be transformed into a format suitable for 3D printing. The model will possess osteoinductive, osteoconductive, and osteogenic bone properties as well as mechanical stability and provide a proper scaffolding for human mesenchymal stem cells to promote further growth.


3D printed abdominal aortic aneurysm phantom for image guided surgical planning with a patient specific fenestrated endovascular graft system
Paper 10138-23

Author(s):  Karen M. Meess, The Jacobs Institute, Inc. (United States), et al.
Conference 10138: Imaging Informatics for Healthcare, Research, and Applications
Session 6: Novel Applications in 3D Printing
Date and Time: 2/16/2017 10:30:00 AM

Following the new trends in precision medicine, Abdominal Aortic Aneurysm (AAA) treatment using patient-specific fenestrated endovascular grafts is an approach which enables endovascular treatment for juxtarenal AAAs. The X-ray guided procedure requires precise orientation of radiopaque markers of multiple endografts within the arteries. To avoid periprocedural complications and improve training, patient-specific 3D printed phantoms were fabricated to familiarize physicians with complex image-guided procedures and new devices in a risk-free simulation environment.We demonstrated how patient-specific devices and patient-specific 3D printed phantoms can help to improve clinical interventional procedures using surgical planning.


Design optimization for accurate flow simulations in 3D printed vascular phantoms derived from computed tomography angiography
Paper 10138-25

Author(s):  Kelsey N. Sommer, Toshiba Stroke and Vascular Research Ctr. (United States), et al.
Conference 10138: Imaging Informatics for Healthcare, Research, and Applications
Session 6: Novel Applications in 3D Printing
Date and Time: 2/16/2017 11:10:00 AM

We developed a design to manage the distal arterial flow resistance and pressure creating physiologically and geometrically accurate phantoms. Patient specific CT data were imported into a Vital Imaging workstation, segmented, and exported as STL files. Using a mesh-manipulation program, we created flow models of the coronary tree and the Circle of Willis. Distal arteries were connected to a compliance chamber. The phantom was then printed using a multi-material printer: the vessel in Tango+ and the fluid flow simulation chamber in VeroBlue. The model was connected to a programmable pump and pressure sensors measured flow characteristics through the phantoms.


Initial retrospective FFR investigation using flow measurements in patient specific 3D printed coronary phantoms
Paper 10138-26

Author(s):  Lauren Shepard, Univ. at Buffalo (United States), et al.
Conference 10138: Imaging Informatics for Healthcare, Research, and Applications
Session 6: Novel Applications in 3D Printing
Date and Time: 2/16/2017 11:30:00 AM

Accurate patient specific phantoms for device testing or endovascular treatment planning can be 3D printed. We expand the applicability of this approach for diagnosed cardiovascular disease, in particular, for CT-geometry derived benchtop measurements of Fractional Flow Reserve, the reference standard for determination of significant individual coronary artery atherosclerotic lesions. 3D model based FFR measurements correlated well with stenosis severity. FFR measurements for each stenosis grade were: 0.4 severe (n=1), 0.55 moderate (n=1), and 0.75 mild (n=2). This approach can be used for accurate FFR measurement, as a future diagnostic tool, or for testing CT image based FFR methods.


Multi-frequency accelerating strategy for the contrast source inversion method of ultrasound waveform tomography using pulse data
Paper 10139-7

Author(s):  Hongxiang Lin, The Univ. of Tokyo (Japan), et al.
Conference 10139: Ultrasonic Imaging and Tomography
Session 2: Ultrasound Tomography and Photoacoustics
Date and Time: 2/15/2017 10:30:00 AM

This work constructs a multi-frequency accelerating strategy for the contrast source inversion (CSI) method using pulse data in the time domain. We introduce the CSI method of ultrasound waveform tomography in the frequency domain. The idea of frequency hopping helps recursively calculate CSI in the current frequency using the result obtained from the former frequency reconstruction. In the numerical test, the data were generated by the K-wave simulator and have been processed to suit the computation of the CSI method. We investigate the performance of the multi-frequency and single-frequency reconstructions and conclude that the multi-frequency accelerating strategy significantly reduces the computational cost.


Quantitative 3D high resolution transmission ultrasound tomography: creating clinically relevant images
Paper 10139-33

Author(s):  James Wiskin, QT Ultrasound LLC (United States), et al.
Conference 10139: Ultrasonic Imaging and Tomography
Session 7: New Applications of Ultrasound in Medicine and Biology
Date and Time: 2/16/2017 2:40:00 PM

We show the results of applying a 3D nonlinear inverse scattering algorithm to 3D data in a clinically useful time frame. This method yields Quantitative Transmission (QT) ultrasound images with high spatial and contrast resolution. We reconstruct various 2D and 3D phantoms and verify speed of sound values with independent measurements. We show that 2D reconstruction algorithms can introduce artifacts into the QT breast image and show high resolution microscopic anatomic correlations comparing cadaveric breast QT images with MRI to establish imaging capability and accuracy. We show reconstructions from volunteers, and an objective visual grading analysis to confirm clinical imaging capability and accuracy


Automatic computational labeling of glomerular textural boundaries
Paper 10140-15

Author(s):  Pinaki Sarder, Univ. at Buffalo (United States), et al.
Conference 10140: Digital Pathology
Session 3: Precision Medicine and Grading
Date and Time: 2/13/2017 9:00:00 AM

Diagnosing renal disease requires time consuming manual inspection of needle biopsies. Automated quantification of the same characteristics of biopsies by a digital method greatly decreases the burden on pathologists and improves the reproducibility of the biopsy process. The current barrier to the automated quantification of renal injury in proteinuria is the digital identification of the glomerulus. We have developed an integrated method, based on Gabor filter bank based textural segmentation, statistical F-testing, and distance transform, for segmenting glomerular boundaries from renal biopsies. Our method outperforms sole Gabor filter bank based method, and is able to operate on multiple histological stains.


Convolutional neural networks for prostate cancer recurrence prediction
Paper 10140-16

Author(s):  Neeraj Kumar, Indian Institute of Technology Guwahati (India), et al.
Conference 10140: Digital Pathology
Session 3: Precision Medicine and Grading
Date and Time: 2/13/2017 9:20:00 AM

Precise prediction of treatment outcome is important for effective cancer treatment planning. We present an approach to predict the risk of prostate cancer (PCa) recurrence, for patients undergoing radical prostatectomy (RP) for whom follow up information (recurrent vs. non-recurrent) was available, through automated analysis of H&E stained tumor slides acquired at the time of treatment. Current PCa grading using Gleason score, which is based on gland shapes alone, often fail to successfully predict the recurrence. Our PCa recurrence prediction approach uses two separate convolutional neural network (CNNs). The first CNN detects nuclear centers, including separate ones for touching or overlapping nuclei, by processing small patches (51 x 51) and estimating the distance of their central pixel from the nearest nucleus boundary (distance transform). The second CNN works with larger patches (101 x 101) centered at the detected nuclear centers and estimates the probability of it belonging to a tissue sample of a patient for whom biochemical recurrence was observed after RP. The estimate of recurrence probability for a patient is based on aggregation of recurrence probabilities of nuclei detected in its tissue images. The proposed approach gave 0.81 AUC for a sample of 30 recurrent cases and 30 non-recurrent controls, after being trained on another set of 80 cases and control patients each. If validated further, such an approach might help in choosing between a combination of treatment options such as active surveillance, radical prostatectomy, radiation and hormone therapy. It can also generalize to prediction of outcomes in other cancers.


Leveraging unsupervised training sets for multi-scale compartmentalization in renal pathology
Paper 10140-17

Author(s):  Pinaki Sarder, Univ. at Buffalo (United States), et al.
Conference 10140: Digital Pathology
Session 4: Keynote and Trends
Date and Time: 2/13/2017 10:10:00 AM

We have developed an unsupervised image segmentation method to reveal pertinent structures simultaneously in multi-scale with unparalleled computational speed. We model an image as a graph, where nodes are pixels and edges are the Euclidean distances between the nodes. We partition the graph by minimizing a Potts model Hamiltonian. Computational speed is optimized by selectively choosing sparse graph of the input image using a Cantor pairing function. The unsupervised output is used to train a multiclass support vector machine to optimally segment the complete image. We apply this method in renal pathology, demonstrating the capability to segment biologically relevant structures.


Sign up for event e-alerts

 Subscribe


Browse Defense, Security, and Sensing 2011 papers


Journal of Medical Imaging

Get email updates for the Journal of Medical Imaging.