Advances in Artificial Intelligence for the Ultrasound Diagnosis of Thyroid Nodules

The increasing detection of thyroid nodules and thyroid cancer has made timely malignancy detection and the avoidance of unnecessary biopsy, surgery, and overtreatment important clinical priorities. Ultrasound is a convenient, real-time, inexpensive, easy-to-obtain and repeated non-invasive imaging method for screening and risk assessment of thyroid nodules. The diagnostic results of human operators and equipment settings can affect the sensitivity of image interpretation. Recently, with the development of artificial intelligence (AI), machine learning and deep learning have been applied to thyroid disease diagnosis. Segmentation of images, feature extraction, differentiation of benign and malignant nodules, risk stratification, biopsy decision support, cytological examination, contrast-enhanced ultrasound video analysis, and prediction of lymph node metastasis are all included. Research has shown that, at present, AI support can improve the precision and uniformity of diagnosis independently of operator experience. It will also be convenient for new doctors and health workers in primary care to use. Review of the technical foundations, current applications, advantages and disadvantages, and future directions of artificial intelligence in ultrasound diagnosis of benign and malignant thyroid tumors. To provide information on the clinical application and use of these technologies in personalised diagnosis and treatment.