The Implementating And Optimizing Of Tumor Detection By Mammo Computer-aided Diagnosis(CAD)

Breast cancer is harmful to women’s lives and health ,and it is becoming one of today’s major killers. Incidence of breast cancer showed a rising trend year by year in our country. Therefore, a large number of patients had breast DR photography to troubleshoot the disease. These results must be taken individually reviewed by a physician, and it brings the workload of doctors in film-reading increased substantially, increasing the difficulty of lesions detection timely and accurately. In this case, we bring the computer-aided diagnosis (CAD) .With this technology, computers used to do the pre-diagnosis of focus which have significant or typical features, and verify lesions typical of pre-and screen to get rid of the negative region, so that doctors can focus on more difficult lesions of the investigation, or the result of automatic detection Assessment, so that to some extent reduce the doctor’s work intensity and improve the quality and efficiency of film-reading.In this paper, based on the characteristics of breast lesions, by induction, sum up and experimental methods, using MATLAB simulation tools, implementation, and optimization of breast CAD algorithm. In the implementation process, the mass of the testing process includes the following stages: MG image preprocessing, image processing (compression, region of interest extraction), lesion area separation. In image processing and lesion area between the separation stage, joined the support vector machine (SVM) mechanism, which allows repeated training, to improve the classification accuracy, thus improving the separation process in the lesion area screen out false positive capacity . The main characteristics of breast lesions, using the histogram, Otsu, support vector machine (SVM) and other classical algorithm, attributes and characteristics of these algorithms analysis, on which the parameters, the algorithm structure adjustment, after the experiment, to be completed Extraction of the classification of breast lesions.In order to improve detection efficiency and accuracy of key technology and testing process in CAD described in detail the role and comparative discussion of the results obtained by experiment, to analyze the optimization effect, most of them end up solving the problem way to achieve the optimization purposes, first of all, for before compression Threshold Value to determine the problem, the use of Otsu method for processing, and with parallel processing to increase the extraction speed; for the test sample lesions property and more traditional methods should not judge the situation, the use of SVM Support Vector Machine for compression and noise reduction to the samples after the lesion screening, resulting in lesions of the extraction results, the same time, through repeated training of the algorithm can effectively improve the accuracy of CAD detection algorithm.