Breast Computer-aided Detection Algorithm On Fuzzy Classification

Breast cancer is one of the most frequently maligant diseases. It has the highest prevalence among all cancers in the population. Early detection of breast cancer is of utmost importance. Because of the high spatial resolution,simple operation, low cost, mammography has become the best available imaging examination for the detection of early signs of breast cancer at present. It can reveal evidence of the lesion of the breast, such as masses and calcifications. Mammographic screening has been shown to be effective in reducing breast cancer mortality rates. But it is difficult for experienced doctors to find the clustered micro-calcifications in the mammogram in time. In addition, the doctor must read a lot of theX-films every day, so, it is easy to misgiagnosis. As a means of reducing errors, computer-aided detection (CAD) techniques could offer a cost-effective alternative for double reading. A CAD system could act as a second reader, prompting the radiologist to review areas in a mammogram deemed to be suspicious by specialized computer algorithms.The performance of classification is the key point for a CAD system, which relies on the database of feature vectors. In traditional CAD system, training dataset always come from DDSM and MIAS which are built from the samples of American and European. Therefore, the performance is often not good for Chinese. To address this problem, we first established a database belongs to Chinese. The database consists of human information as well as image information. We firstly add the human information into image feature vectors. Then, according to the multi-source prosperities of feature vectors, we employ Multiple Kernel Support Vector Machine which separated the feature vectors into different groups to map to different dimension spaces. At last, classification model is trained based on the feature vectors. The experimental results show that the detection performance has been improved slightly compared with Support Vector Machine. This thesis firstly introduces fuzzy theory into the detection of micro-calcification. The classified samples are fuzzed by membership, and the value of membership function for each test sample is calculated, which makes the results more objective and quantified. In this way, more information is provided to a doctor for further diagnosis.A lot of experimental results show that the proposed system, including computer-aided detection of breast database, anthropology, the image feature extraction, feature classification and other functions, is very practicability and stable. As a tool of breast calcification detection, it can improve the efficiency and accuracy of clinical examination.