Abstract

Prevalence of thyroid diseases is increasing globally. Detection of thyroid nodules using diagnostic imaging relies heavily on physicians’ expertise. Development of artificial intelligence (AI) approaches has led to significant advancement in visual identification. Machine learning and radiomic are approaches of artificial intelligence that have the potential to improve clinical diagnosis. AI approaches can be used to detect biological anomalies, diagnose neoplasms, and predict response to therapy. However, diagnostic accuracy of these approaches is still a point of contention. Aim of this article is to give a general review of aspects, limits, and key challenges in use of artificial intelligence for thyroid imaging. Core principles and process parameters of learning algorithms, cavernous learning, and technological frontier as well as data processing criteria, the distinction between AI approaches, and their constraints are discussed in this article.