Ultrasound image segmentation methods: A review

Abstract

Breast cancer is one of the leading causes of death in México and among the world. This is mainly due to late diagnosis and the price of cancer treatment. Ultrasound (US) is one of the most used tools for image-based assessment of this disease, since it can help discriminate solid vs. cystic masses, as well as between benign or malignant masses. However, ultrasound imaging depends largely on the radiologists experience. A detection not depending on such experience should produce better diagnosis; therefore, it is necessary to develop automatic detection systems for US images. These systems are based on the image segmentation, which is an image processing technique used to analyze and group pixels by their features. US image segmentation represents important challenges due to the complicated appearance of healthy and tumoral tissue in the ultrasound image that includes speckle pattern, low contrast, blurred boundaries, etc. In this work, we provide a review of tests and comparisons of various segmentation methods that had helped to detect lesions in US images, we present some visual examples to compare this methods and we evaluate their performance in order to develop computer-aided diagnostic systems that help radiologists to do better diagnoses.