Research Project
2715054
There are several evidences to show that the early detection of breast cancer can reduce mortality from the disease. Also, X-ray film screen mammography has shown more sensitivity and specificity to detect tumors than any other non-invasive diagnostic techniques that are currently used. Hence, in this study digital mammograms are considered. The digital mammography also facilitates the use of computer aided detection and diagnosis that are the focus of this study. In 90% of in situ carcinomas, calcifications are the dominant abnormality. Therefore, as far as the early stage cancers are concerned, the proportion of lesions detected on the basis of microcalcifications are becomingly increasingly greater. Further, about 80% of the breast cancers are found to contain microcalcifications on histology examination. Therefore, the accurate detection of microcalcifications should facilitate in early detection of breast cancer and saving human lives. It is possible to detect these microcalcifications by using detection algorithms since they exhibit difference in density, size and shape from normal anatomic structures. Hence, the focus of this study is to develop and evaluate detection/segmentation algorithms based on 2D wavelets and probabilistic networks. The microcalcifications classified as suspicious in the segmented image (mammogram) will be highlighted for close examination by a radiologist. The technology that results from this study can be considered as a tool that helps a radiologist in detecting and diagnosing breast cancer in an efficient way. The main advantages of the proposed approach are: (1) noise reduction, (2) image enhancement and (3) good segmentation. Since in general, mammograms are poor quality images with low contrast due to the small differences in X-ray attenuation between breast tissues and noisy, the detection techniques that reduce noise and enhance images would improve the detection performance. PROPOSED COMMERCIAL APPLICATIONS: The proposed Phase I study will result in a novel methodology that (a) enhances the image quality of mammogram, (b) accurately segments and detects microcalcifications, (c) classifies microcalcifications into suspicious or not, and (d) highlights the suspicious microcalcifications for close examination by a radiologist. Such a methodology will lead to a commercial application of computer aided detection and diagnostic tool in digital mammography. Such a technology should help radiologists in early detection of breast cancer and thus save human lives.
