Abstract: In recent decades, the incidence of thyroid cancer has rapidly increased worldwide. Ultrasonography is the first choice of imaging modality for the diagnosis of thyroid nodules because of its low cost and easy availability. Conducting the examination is relatively easy, however, it requires physicians with a high level of expertise and rich experience in delivering imaging services. Ultrasonography is essential for the constant improvement of the diagnostic criteria of thyroid cancer. However, the promotion, popularization, and appropriate use of unified standards require a large amount of manpower and financial resources, thereby making the implementation process quite challenging. A comprehensive evaluation of the malignancy risk of thyroid nodules and the nature of cervical lymph nodes is essential for clinical decision-making, but this evaluation is arduous and complicated. Moreover, the demand for medical and health care resources in China is enormous, and the distribution of medical resources is inequitable. The rapid development of deep learning algorithms in the field of artificial intelligence has demonstrated their powerful performance in the field of medical image diagnosis. Big data combined with deep learning can effectively solve the current problems in clinical diagnosis and treatment. Deep learning has great advantages and application prospects in the field of medical image diagnosis and ultrasound diagnosis of thyroid nodules. The use of deep learning methods to analyze ultrasound images to construct an ultrasound automatic diagnostic system facilitates the ultrasound diagnosis of thyroid tumors and simplifies the work process of ultrasound doctors. Furthermore, it can help improve clinical practice efficiency. This review summarizes the research status and progress of deep learning in the field of ultrasound diagnosis of thyroid tumors and cervical lymph nodes in recent years.