A new framework for early diagnosis of breast cancer using mammography images

Breast cancer is one of the most common types of cancer in women. This type of cancer can be detected and treated at an early stage, and the quality of life of sick individuals can significantly improve. In addition to radiology specialists, tools that can help these specialists are also needed in early diagnosis. The proposed study includes a tool with a new framework that can help diagnose with maximum accuracy. This study uses a multi-class MIAS data set of benign, malignant, and normal mammography images. First, only the breast region is determined from the images in this data set with the help of morphological operations. The bicubic interpolation-based super-resolution method enhances the details of the detected areas. In addition, the success of classification methods is directly related to the diversity of the data set. Therefore, the images in the data set are enriched with different transformations. Another vital point for classification methods is the extraction of the feature vector. A unique feature vector consisting of 11 features is obtained in the created framework. Then, these feature vectors are separated into three classes separately by seven different classifiers tested in the created framework: KNN, SVM, NB, LDA, DT, ANN, and CNN. The optimal parameters of these classifiers are obtained by the Grid Search method. Finally, the classification results are evaluated using accuracy, precision, sensitivity, specificity, F-score, and time metrics. The evaluation results prove that the created framework is a successful aid in diagnosing breast cancer.