A diagnostic system for classifying and segmenting breast cancer based on ultrasound images

Breast cancer is the most common type of cancer among women worldwide. Ultrasound is a type of imaging widely used in the diagnosis and examination of many soft tissues, including abnormalities of the breast because it offers the advantages of being real-time, portable, low-cost, and non-invasive. Ultrasounds suffer from high variability as well as speckle noise, which reduces image quality; therefore, it may be difficult for doctors to detect a cancerous cell. In this paper, we use image enhancement filters to improve image quality then we develop the U-Net model to segment ultrasound images. The accuracy reached was 0.97 and the dice coefficient was 0.95. Furthermore, we classify images as malignant or benign using various traditional techniques such as k nearest neighbors (KNN), Random Forest, Decision Tree and support vector machine (SVM). The highest rate is achieved by KNN. After the image enhancement, KNN obtained an accuracy of 0.86.