There has been a huge focus in recent months on the rise of artificial intelligence (AI) in all aspects of modern life, and the head and neck clinic is no exception it appears. This paper builds on previous work to compare the diagnostic accuracy of an experienced head and neck radiologist looking at thyroid ultrasound scans with the AI of a computer-assisted design (CAD) system. The radiologist was blinded to any clinical information and classified nodules according to the American College of Radiology and American Thyroid Association guidelines. The CAD system reports on the probability that nodules are benign or malignant. The findings of this study are not that surprising, especially considering that the AI technology is only newly developed; the human performed better, and the CAD system had much lower specificity than the radiologist. What is interesting from this study is the high sensitivity and negative predictive value of the AI ultrasound system and this may have significant clinical potential for the future. The authors suggest a role for CAD in ruling out benign nodules, and saving radiologists time when reporting, and determining those patients who require fine-needle aspiration or surgery. The retrospective design of this study, and the bias imposed by all the scans being static, and involving patients who went on to have thyroidectomy, are recognised limitations of this paper. This is certainly an area which is likely to expand in the future as the technology develops.

Computer aided system for diagnosing thyroid nodules on ultrasound: a comparison with radiologist-based clinical assessments.
Gao L, Liu R, Jiang Y, et al.
Share This
Aileen Lambert

Great Ormond Street Hospital, London, UK.

View Full Profile