Artificial intelligence could accurately identify gastrointestinal stromal tumors (GIST) from non-GIST neoplasms via endoscopic ultrasound imaging, according to a study recently published in Scientific Reports.

“These findings suggest that once the endoscopic ultrasound-based artificial intelligence (EUS-AI) system is built to differentiate GISTs from non-GISTs in one part of the gastrointestinal tract, it can be applied to other parts of the gastrointestinal tractable to other AI systems with different algorithms for subepithelial lesions (SELs) of the upper gastrointestinal tract,” the authors wrote.

This prospective study included 52 patients of both sexes, previously diagnosed with nongastric SELs, out of which duodenum was the most common location as it was present in 26 patients, followed by esophagus in 15, and colon in 11 cases.


Continue Reading

Read more about GIST diagnosis

This approach showcased an accuracy of 94.4%, a sensitivity of 100%, and a specificity of 86.1%, with an area under the curve of 0.98. Moreover, lesions located in the esophagus obtained the best results with 100% accuracy, sensitivity, and specificity. On the other hand, the duodenum and colon had an accuracy of 96.2% and 90.9%, respectively, while both structures had a sensitivity of 100% and specificity of 0%.

Furthermore, the size of the lesions positively correlated with the diagnostic accuracy of this method. However, the location did not seem to affect it, suggesting that disease frequency may not modify this data, Minoda and colleagues explained.

EUS-AI identifies cystic degeneration and increased lesion size, yielding room for even more improvement. Previous research proved its efficacy for conducting the differential diagnosis of gastric GISTs. Hence, the authors hypothesized it might also be useful for other organs.

“This study showed that EUS-AI is useful for the differential diagnosis of GISTs in the gastrointestinal tract other than the stomach, although it was trained for gastric SELs,” the study team concluded.

Reference

Minoda Y, Ihara E, Fujimori N, et al. Efficacy of ultrasound endoscopy with artificial intelligence for the differential diagnosis of non-gastric gastrointestinal stromal tumors. Sci Rep. Published online October 5, 2022. doi:10.1038/s41598-022-20863-8