Thyroid nodules are more common. Ultrasound examination is relatively common，but the accuracy of diagnosis is closely related to the examination skills，clinical experience and analytical ability of sonographers. As the incidence of thyroid cancer increases，improving the accuracy of ultrasound is a problem to solve. As an extension of artificial intelligence，deep learning is becoming a research hotspot in the diagnosis of benign and malignant thyroid tumors by ultrasound. This paper introduces the research progress and application value of deep learning in ultrasound diagnosis of thyroid nodules in order to provide greater clinical value.