Study: AI outperforms radiologists in managing thyroid nodules

A deep learning algorithm determined whether a thyroid nodule required a biopsy with a level of accuracy equal to or, in some cases, better than that of trained radiologists, according to a study published July 9 in Radiology.

In the study, led by scientists at the Duke University School of Medicine in Durham, N.C., the algorithm and a group of radiologists — three experts in thyroid nodule analysis and nine other radiologists who regularly perform the analysis in clinical practice — were each tasked with analyzing a set of 99 thyroid nodules. Their results were compared to previously made post-biopsy diagnoses for each nodule, then to each other's assessments.

On average, the artificial intelligence and the three expert radiologists recommended biopsies with similar sensitivity and specificity. However, the algorithm achieved higher sensitivity than five of the nine other radiologists, and higher specificity than seven of the nine.

The researchers concluded that, if implemented into clinical practice, the AI, which consistently returned the same result when shown the same image, could eliminate the variability often found between the analyses of different radiologists. Additionally, the algorithm "could reduce the time required for interpretation of thyroid nodules, which puts some strain on radiology departments," they wrote.

More articles about AI:
Waystar acquires AI tech firm to streamline prior authorization
Cigna, Celgene join $23M funding round for AI-powered precision healthcare startup
4 ways AI is improving the customer experience

Copyright © 2024 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.

 

Featured Whitepapers

Featured Webinars