A telediagnostic system for automatic detection of lesions in digital mammograms

Breast cancer is the second type of cancer that affects women in the world, losing only for non melanoma skin cancer. The mammography is the most accurate exam for the diagnostic of this disease, allowing to discovery in its initial stage and increasing the chance of prompt treatment. This paper presents the modeling and implementation of Web based telediagnostic system for automated detection and register of lesions in mammographic images, based on Independent Component Analysis (ICA) and Support Vector Machine (SVM). The system analyses digital mammography images submited for Internet providing an image diagnosis and indicating the presence of suspicious regions, which can be evaluated and treated by an specialist. Also presents the methodology for development of proposed system. A system's prototype was developed to run tests that could measure its efficiency. Mini-MIAS was the training database used to test the algorithms. We also used SMV to classify image interest regions as normal or suspicious. Test results showed that 89.13% of normal samples were correctly classified and 92.17% of suspicious samples were also classified apropiately. We note that the proposed method provides support to detect suspicious regions in mammography images, incorporated in a automated CAD system.