AI is helping pathologists find harmful lesions the eye can't see
As a pathologist, Dr. Christopher Owens has spent his career trying to improve his methods of catching precancerous colorectal polyps in patients. If he could catch cases earlier, fewer people would develop colorectal cancer, he said.
Dr. Christopher Owens, an associate medical director at Quest Diagnostics, explains that certain precancerous colorectal polyps called sessile serrated adenomas (SSAs) can be difficult to detect. These lesions, which resemble benign hyperplastic polyps, may not exhibit symptoms in their early stages and account for an estimated 1.9 million annual cases of colorectal cancer worldwide.
In a Northeastern University study published in Lecture Notes in Computer Science, researchers discovered that vision transformer models, a type of AI technology, may be more effective than older AI models in identifying these precancerous lesions. Vision transformer models, which can understand the relationships between different parts of an image, achieved a 70% sensitivity rate in detecting SSAs, compared to the 40% sensitivity rate of the best-performing convolutional neural network (CNN) models.
CNNs, which analyze images segment by segment, may struggle to identify these subtle differences in cell architecture. However, the study also highlights concerns from the American Gastroenterological Association about AI-assisted detection systems potentially increasing the identification of low-risk polyps, leading to more frequent and costly colonoscopies with minimal benefit.
The researchers are currently working with clinicians to validate these findings in real-world clinical settings and assess their impact on patient care.
Written by urgent.news from Medical Xpress's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
