Advances in Breast Ultrasound Segmentation and Classification

Breast cancer is one of the most prevalent cancers affecting women worldwide. Ultrasound is extensively utilized for clinical screening and diagnosis due to its affordability, absence of radiation, and rapid imaging capability. To enhance diagnostic accuracy, computer-aided diagnosis (CAD) systems have been developed, with segmentation and classification being key techniques. This review systematically examines 62 recent studies on breast ultrasound segmentation and classification, covering various imaging techniques such as B-mode, elastography, 3D ultrasound, contrast-enhanced ultrasound (CEUS), and color Doppler. specifically, we detail the challenges and deep-learning-based methods associated with these modalities. Comparative analysis reveals that current deep learning approaches typically achieve Dice coefficients ranging from 0.79 to 0.91 for segmentation and classification accuracies exceeding 88.2% in multimodal settings. Finally, this article identifies critical research gaps, including data scarcity and model interpretability, and discusses future directions such as multimodal fusion and explainable AI (XAI) to further improve clinical applicability.