- Biomedical Optics & Medical Imaging
- Defense & Security
- Electronic Imaging & Signal Processing
- Illumination & Displays
- Lasers & Sources
- Micro/Nano Lithography
- Optical Design & Engineering
- Optoelectronics & Communications
- Remote Sensing
- Sensing & Measurement
- Solar & Alternative Energy
- Sign up for Newsroom E-Alerts
- Information for:
Illumination & Displays
Interactive visualization tool augments pathology reporting
Radiologists can generate better reports by using interactive visualization tools, thereby improving physician efficiency and efficacy in diagnosing pathologies.
10 September 2006, SPIE Newsroom. DOI: 10.1117/2.1200608.0241
Advances in medical imaging have improved the diagnosis of various diseases. However, three-dimensional (3D) imaging modalities produce large amount of digital data that is difficult and tedious for physicians to interpret. Thus, computer-aided diagnosis (CAD) systems have started to play a critical role in interpreting these modalities. In turn, CAD helps physicians detect, isolate, and align pathologies and organs. Now, through picture archiving and communication systems (PACS), physicians may also report diagnoses through a computer interface. This is a great improvement over the current reporting system, in which the pathologies are manually recorded using static text.
Current 3D visualization technologies are used widely in viewing the data produced by the imaging equipment, and such capability is usually integrated into the same tool that controls the scan. However, in contrast to the abundant tools available for viewing, the reporting schemes remain static. The difficulty in presenting these useful images in reports is that the pathologies first need to be detected and isolated from the data before they can be presented in the compact format of a report.
To construct a compact, visual report, we start with current CAD functionality. For example, in Figure 1 we demonstrate a visualization tool for reporting lung nodules that shows a 3D, interactive summary of all the detected and segmented nodules. Then, using previously developed algorithms for automatic nodule segmentation and detection,1–3 we acquire the location, size, shape, and other properties of the nodules. The displayed results include the shapes and relative sizes of the nodules with segmented anatomic features serving as visual references. In this particular implementation, the two lung surfaces are rendered semi-transparent as the visual reference. Other references could be used, such as the thoracic cage, airways, or vessel trees.
Figure 1. (a) The scrollable main window displays the volume data in slice images. (b) The 3D window displays a summary of nodules, shown with their actual shape and relative sizes. The lung surfaces are shown in sampled dots as visual reference. Here, the user has clicked on the nodule indicated with arrow in the 3D window (b), triggering the main window (a) to display the corresponding slice image and highlighting the nodule in the volume data.
Because the global 3D summary is relatively compact, it may be used as an integral part of the reporting system for lung nodules. In addition to it being a dynamic and interactive environment that allows the user to navigate through the nodules—rotating and zooming—the display is spatially synchronized with the main window, so the user can always see the volume data. A click on a nodule in the 3D display updates the main display to the corresponding slice where the nodule was automatically detected, and the nodule location is outlined in the slice shown on the main window.
In the future of CAD, we anticipate a system able to detect and evaluate all types of pathologies and diseases. We call such system a combinatorial-CAD (C-CAD) system. A full report from the C-CAD system would also contain an evaluation of the general health condition of the patient as best can be deduced from the scanned body section. As with the CAD system, the C-CAD report would also be of a graphical and interactive nature.
The new report would allow physicians to examine radiological results with high efficacy and efficiency. Figure 2 shows an example of a proposed C-CAD visualization with a combination of important pathologies that allows the user to choose some of them to view in more detail. In addition to lung nodules, the system also has rendered the pathologies of the rib and vertebra both in volume and surface views, for whichever presents the abnormality best. The user would also be able to see more information in the main display window—such as the complete thoracic cage, vessel tree, airway tree, etc.—but in the limited reporting window, the user would see rendered only the most relevant information.
Figure 2. An example C-CAD report shows the abnormalities with irrelevant anatomies removed to prevent visual blocking. Brief annotations are shown on the display. When the user changes the view, the annotations automatically update to point to the correct location. In this particular view, three items are highlighted: a grown lung nodule, a metastasis tumor in the vertebra, and a deformation of one of the ribs. This display is linked to the main window displaying volume data (not shown).
Intelligent Vision and Reasoning, Department Siemens Corporate Research
Hong Shen received his Bachelor of Engineering degree from the Electronic Engineering Department of Tsinghua University in Beijing, China. His doctoral degree is from the Electrical, Computer, and System Engineering Department of Rensselaer Polytechnic Institute in Troy, NY. Currently, he is a project manager at Siemens Corporate Research in Princeton, NJ. His major interest is in the field of medical imaging and image analysis, especially computer-aided diagnosis. In addition, over the past several years, he has presented several papers at SPIE Medical Imaging Conferences.
ZOLL Medical Corporation
1. L. Fan, J. Qian, B. Odry, H. Shen, D. P. Naidich, G. Kohl, E. Klotz, Automatic Segmentation of Pulmonary Nodules by Using Dynamic 3D Cross-correlation for interactive CAD Systems,
Vol: 4684, 2002.
2. H. Shen, B. Goebel, B. Odry, A New Algorithm for Local Surface Smoothing with Application to Chest Wall Nodule Segmentation in Lung CT Data,
Vol: 5370, 2004.
3. J. Qian, L. Fan, G. Wei, C. L. Novak, B. Odry, H. Shen, L. Zhang, D. Naidich, J. Ko, A. Rubinowitz, G. Kohl, A Knowledge-based Automatic Detection of Multi-type Lung Nodules from Multi-detector CT Studies,
Vol: 4684, 2002.