Breast ultrasound image segmentation based on neutrosophic set and watershed method for classifying margin characteristics

Breast cancer is the leading cause of death in women worldwide. Ultrasonography (USG) is one of the imaging modalities which is widely used to detect and classify the mass abnormalities of the breast nodule. The use of image processing in the development a computer aided diagnosis (CADx) can assist the radiologists in analysing and interpreting the abnormalities of ultrasound nodules. This paper proposes an approach to classify the characteristics of breast nodule into circumscribed and not circumscribed classes. The proposed approach is implemented on 102 breast nodule images comprising of 57 circumscribed and 45 not circumscribed margins. Seven relevant features are extracted from nodule which is automatically segmented by combination neutrosophic set and watershed methods. The classification process based on multi-layer perceptron (MLP) classifier obtains the sensitivity of 96.49%, NPV of 95.35% and AUC of 0.972. These results indicate that the proposed approach successfully classify the margin characteristics of breast ultrasound nodule.