NEW YORK (Reuters Health) – A machine-learning algorithm that analyzes thyroid-nodule ultrasound images correctly identifies likely cancerous nodules 97% of the time – which is on par with thyroid biopsy with fine-needle aspiration, according to research presented at the Endocrine Society annual meeting.
“We demonstrated that using artificial intelligence (AI) analysis of ultrasound images to rule out thyroid cancer and avoid biopsy is definitely possible,” Dr. Nikita Pozdeyev of the University of Colorado Anschutz Medical Campus, in Aurora, said in a news release.
“This technology could assist radiologists and endocrinologists in choosing which thyroid nodules should undergo biopsy, especially those in the community who may not review a large number of thyroid ultrasound images,” he added.
The researchers trained their AI model on more than 30,000 ultrasound images of 621 thyroid nodules to classify the nodules as “cancer” or “no cancer” and externally validated the results on a different set of ultrasound images of 145 nodules.
The deep learning classifier achieved a negative predictor value (true negative) of 94% to 99% – comparable to that of the benign fine-needle aspiration biopsy with “acceptable” specificity of 52% to 61%, Dr. Pozdeyev reported.
Microcalcifications and spongiform appearance were “reliably” recognized by the AI algorithm as malignant and benign features, respectively, he noted.
Dr. Pozdeyev noted that thyroid nodules are very common and fine-needle aspiration biopsy is used to diagnose thyroid cancer. However, most biopsies show benign (noncancerous) results. If confirmed in other studies, use of this AI tool could help patients avoid unnecessary biopsy for thyroid nodules.
SOURCE: https://endo2022.endocrine.org/ Endocrine Society Annual Meeting (ENDO 2022), June 13, 2022.
Source: Read Full Article