AI-Supported Image Analysis Helps Distinguish Acute Diverticulitis from Colon Cancer

An artificial intelligence (AI) support system significantly improved radiologists’ ability to differentiate between benign and malignant etiology of bowel wall thickening on computed tomography images, according to a study published online in JAMA Network Open.

Researchers from the Technical University of Munich in Germany developed the deep learning algorithm to discern colon carcinoma from acute diverticulitis using contrast-enhanced computed tomography images of 585 patients.

The deep learning model predicted colon carcinoma and acute diverticulitis with a sensitivity of 83.3% and specificity of 86.6% for the test set, according to the study. In comparison, the average sensitivity and specificity for a group of 10 radiologist readers with different levels of experience were 77.6% and 81.6%, respectively.

With support from the deep learning model, however, readers’ sensitivity increased to 85.6% and specificity increased to 91.3%, the study showed. Additionally, AI support reduced the number of false-negatives and false-positive findings.

“Due to the divergent management of patients with colon carcinoma, the reduction of false-negative findings is of utmost importance,” researchers wrote. “Without an AI support system, the false-negative rate was 22% for all readers, 26% for the residents, and 14% for the board-certified radiologists. Artificial intelligence support led to substantial reduction in the false-negative rate to 14.3% for the combined reader group, 16.1% for the residents, and 10.0% for the board-certified radiologists.”

With deep learning model support, negative predictive value improved from 78.5% to 86.4%, and positive predictive value improved from 80.9% to 90.8%, the study found.

“Our model significantly increased the diagnostic performance of all readers,” researchers wrote, “proving the feasibility of AI-supported image analysis.”

Reference
Ziegelmayer S, Reischl S, Havrda H, et al. Development and validation of a deep learning algorithm to differentiate colon carcinoma from acute diverticulitis in computed tomography images. JAMA Netw Open. Published online January 3, 2023. doi:10.1001/jamanetworkopen.2022.53370