A Study On Breast Mass Classification In Mammogram Based On Multi-agent Algorithm

Breast cancer is one of the main causes of death among the women and early diagnosis is an important means to reduce the mortality rate of breast cancer. Computer aided diagnosis (CAD) system that can assist the radiologists as a second opinion to improve the detection and diagnosis efficiency in mammogram reading has been extensively studied.In past years, many classification algorithms have been proposed for mass classification in mammographic CAD system. However, there is few research has been done on the classification with classifier fusion method, which can has been used in many research fields to improve the performance of classification. In this thesis, we propose to use multi-agent method for multiple classifiers fusion, and applied it to the masses classification. The main contributions and innovations are listed as follows:1) An analysis was taken on the application of Multi-Agent algorithm for multi-classification fusion in mass diagnosis.2) Two new features were designed for mass classification. The experiment results demonstrated their efficiency.3) A mass classification model based on Multi-Agent-fusion algorithm was constructed. Compared with the traditional fusion methods, Multi-Agent based Algorithm showed advantages both in the accuracy and stability of classification.4) The Multi-Agent based fusion algorithm was applied to the classification of the bi-view masses. A preliminary study was taken to explore its significance in bi-view information integration for diagnosis.