At present, as an important reference for clinical detection and radiotherapy guidance, medical image plays a key role in the development of medicine. Medical images are mainly include Computed Tomography (CT), Magnetic Resonance Imaging(MRI), X-ray, Ultrasound (US). Compared with the first three, ultrasound is cheaper, has better effect on soft tissue imaging and nearly no harm to human body, so it has been more and more widely used at this stage. Ultrasound image segmentation plays an important role in late image analysis, and it can provide accurate reference for clinical diagnosis and radiotherapy placement. In this paper, the traditional segmentation methods, deformable model-based methods and deep learning methods of ultrasonic images are reviewed.