چكيده به لاتين
The constant prevalence of breast cancer around the world is a major cause of death in women. Early diagnosis of this cancer in women without symptoms can greatly reduce their mortality rate. Screening programs provide many mammograms that should be carefully reviewed by radiologists, since early symptoms of breast cancer are very Mild, screening is difficult for radiologists, and there is some evidence that screening programs, Radiologists have not diagnosed about 25 percent of the cancers that could be seen in the review, and, on the other hand, independent observation by two radiologists significantly increases the sensitivity of screening. This, of course, raises costs and the work pressure of Radiologists increases. Computer Aided Design (CAD) systems can as an auxiliary systems for radiologists, increase the accuracy of diagnosis and improve the classification of cancerous and non-cancerous cases. In this study, using image processing and data mining techniques, mamograms from the DDSM database containing additional lable, were improved, then using the SSD algorithm, suspicious cancerous abnormalities were extracted and in the form of Attribute matrices were stored, then by applying the Generalized LASSO Regression model on the characteristic matrix, suspected cases were classified into three groups including normal, benign and malignant. The way the proposed system works is that it can them into three classes (normal, benign, malignant) with an accuracy of 93.85%, and is to collect the rest of cases, which are likely to be found in three mentioned classes, into a separate class and recommend as a class that need more test to radiologist to be more careful.
