High-performing gastroenterologists given video-based machine-learning models can identify 20 percent more polyps than when using traditional methods, according to a study presented at Digestive Disease Week, June 2 to June 5 in Washington, D.C, Medscape reports.
Here's what you should know:
1. UC Irvine Health's William Karnes, MD, and colleagues applied the artificial intelligence model to 4,100 images with polyps and 4,500 without. The resulting polyp detection rate when GI physicians used the AI model was 96 percent.
2. Researchers then had four GI physicians with polyp detection rates of at least 50 percent examine the same videos without the AI video overlay. The researchers were tasked with identifying polyps in the nine videos.
3. The physicians using the machine-learning model identified every unique polyp. Compared to the GI physicians not using the model, unique polyp detection increased from 36 to 45. The machine helped identify the three unidentified polyps with high confidence and six with low confidence.
4. The AI model had a false-positive rate of 7 percent, and a negative predictive value of 99.5 percent. However, Dr. Karnes believes AI could be an additional diagnostic tool. He said, "When combined with accurate optical pathology, AI may eventually bring us to real-time measures of adenoma detection rate."