Colorectal carcinoma is the third most commonly diagnosed cancer and the second leading cause of death from cancer in the United States. Bladder cancer is the fifth cause of cancer deaths in the United States. Often the cancers are diagnosed at an advanced stage after the patients have developed symptoms, explaining their high mortality rates. Since most colon cancers arise from polyps over a 5 to 15 year period of malignant transformation, screening programs to detect small polyps less than 1 cm in diameter have been advocated. Unfortunately most people do not follow this recommendation. A similar situation exists for bladder tumor staging. The health relatedness of this project is to dramatically increase the number of people willing to participate in screening programs by offering a convenient, nearly risk-free procedure. Virtual colonoscopy (VCon) and virtual cystoscopy (VCys), as new procedures in which computed tomographic (CT) or magnetic resonance (MR) images of the patient's abdomen are taken and a computer visualization system is used to virtually navigate within a constructed 3D model of the colon or bladder, has demonstrated the potential for colon cancer screening and bladder tumor evaluation. We have been contributing the VCon and VCys development for several years. The broad, long-term objective of this project is to develop 3D texture-based computer aided detection (txCAD) techniques to facilitate VCon and VCys as accurate, cost-effective, non-invasive, comfortable techniques to screen large segments of the population. The Phase I specific aims are: (1) To develop tissue-mixture image segmentation mitigating the partial volume effect. (2) To extract the mucosa layer and cleanse the lumen space of the hollow organs from the segmentation. (3) To identify suspicious patches in the mucosa layer with high sensitivity and reasonable false negatives. (4) To obtain the entire volume of each suspicious patch. The Phase II aims are: (5) To extract 3D geometrical, morphological and texture information from each suspicious volume. (6) To classify the extracted features using learning machine to eliminate false positives. (7) To evaluate the performance of the proposed txCAD on a patient database of over 200 cases. It is hypothesized that the proposed txCAD will significantly improve the detection performance compared to previous geometry-based CAD (sfCAD), and dramatically reduces the physician's interaction time with our developed VCon and VCys systems.
Showing the most recent 10 out of 12 publications
