Thyroid Ultrasound Image Segmentation: A Review

The incidence of thyroid nodule has increased enormously in recent years. Nowadays, computer-aided diagnosis (CAD) systems using ultrasound images have been widely developed in assisting radiologists to improve diagnosis accuracy. Image segmentation is the crucial key to the subsequent processing and determines the quality of final analysis in C AD. However ultrasound segmentation remains a challenging problem due to various disorders including inhomogeneity, artifact, speckle and weak boundary. A number of thyroid ultrasound segmentation techniques have been utilized in the last decades and reported to be accurate on private datasets. The improvement of the performance is still increasingly challenging. In this paper, we discuss the basic ideas, pros/cons of the approaches and group them into categories in which easier to understand. With more detail techniques presented, this review will be useful for one's involved in CAD systems development on thyroid ultrasound imaging.