A decision support system based on pattern recognition for evaluating of mammographic images

Breast cancer is one of the leading causes of early mortality in women. Retrospective studies have shown that in current breast cancer screening between 10% to 25% of tumors are missed by radiologists. By the development of computer technology, today, computer aided systems are being used in order to give support to the experts in their province. Especially, medical science profits so much from these kinds of developments. Computer aid systems are becoming prevalent for doctors in diagnosis.A computer-aided detection system that prompts suspicious regions can draw attention of the radiologist to a tumor he might otherwise have overlooked. The use of computer aid diagnosis methods as a ?second opinion? or as a ?pre-reader? strategy has been proposed, may reduce the observer variation.Within the scope of this study, it is aimed to evaluate the segmentation of mammogram images with a pattern recognition system mainly of three operations; pre-treatment, feature extraction and classification. In order to this, in pre-treatment process, histogram based segmentation; in feature extraction process, wavelet transformation of segmented mammogram images and shannon entropy, and during classification process entropy feature by using backprobagation artificial neutral networks recognition of mammogram images is carried out. The use of developed decision support system is tried on 82 mammogram images. The test results have shown that the success of the proposed system is 86%.