The Key Technology Research On Mammograms Computer-aided Diagnosis

Mammography is especially valuable for early detection of breast cancer. This thesis summarizes the research on computer-aided diagnosis of breast masses, including breast boundary segmentation, image enhancement, and fast detection of breast mass regions.To separate breast region from background and decrease the impact of the background on our algorithm, we use double threshold and morphological method based on watershed transformation to extract breast border. Furthermore, we make use of different image enhancement methods to augment mammogram contrast. The enhancement can make the spicules detection more precise and make the abnormality regions more accessible to physician.Based on breast border extracting and image enhancement, we detect the possible centers of mass regions from the breast region. The characteristic of breast mass is that its gray level decreases from the center to outside, especially, the malignant mass has the spicules feature. We employed the multi-scale generic neighborhood operator to calculate the image gradient. Our algorithm is evaluated by comparing the detected mass centers with the mass centers from golden standard. The results demonstrate that within 2 false positive per image, our algorithm can detect more than 90% of masses. We develop the algorithm in our lab, employing the image database of DDSM. It will be better if it can be applied in clinical practice.