DETECTION AND CLASSIFICATION OF MICROCALCIFICATION IN MAMMOGRAMS

Abstract

Breast cancer is becoming a major cause of death among women. As its cause is still unknown, early detection of breast cancer has become important to improve breast cancer treatment. Mammography plays an important part in early detection of breast cancers. Mammography is the X-ray imaging that uses a low-dose x-ray system for examining the breasts. Among all the mammogram abnormalities, microcalcications are the most difficult type of tumor to detect. Area of thesis is microcalcification detection in mam-mograms and classification on the basis of diagnostic feature information extracted form mammograms. Thesis work proposed a computer aided diagnostic system for detection and clas-sification of microcalcification in mammograms. CAD system consist of four stages - Enhancement, Segmentation, Feature extraction, and Classification of microcalcification into benign and malignant case. Enhancement algorithm developed is local contrast enhancement based on morpho-logical operation. For the evaluation of the performance of the enhancement algorithm, contrast improvement index and detail variance to background variance ratio are used. After enhancement, algorithm for segmentation of microcalcification based on edge detection and morphological operation is implemented. After segmentation of each mi-croclacification from mammogram, diagnostic features are calculated. shape based and texture based features are extracted. Total 28 features are calculated. Feature set extracted is used for classification of microcalcification between benign and malignant case. Feature set is divided into two groups - training data set and testing data set. SVM with radial basis function is trained first using training data set and then tested using testing data set.