A deep-learning artificial intelligence model led to a “significant increase in diagnostic performance” for radiologists distinguishing between acute diverticulitis and colon cancer via CT images, according to results in JAMA Network Open.

“Acute diverticulitis (AD) is a frequent gastrointestinal cause for hospital admission with a substantial disease burden,” Sebastian Ziegelmayer, MD, a radiology resident at the Institute of Diagnostic and Interventional Radiology at the Technical University of Munich, and colleagues wrote. “Contrast-enhanced CT is the imaging modality of choice, and imaging signs include bowel wall thickening, fat stranding, enlarged local lymph nodes and the presence of diverticula, none of which is specific to AD.

“However, radiologic differentiation from its most important differential diagnosis, colon cancer, remains difficult due to an overlap of imaging features.”

In a single-center, retrospective study, Ziegelmayer and colleagues evaluated the diagnostic performance, sensitivity, specificity, positive predictive value and negative predictive value of a deep-learning algorithm in differentiating between colon cancer (CC) and AD on CT images.

They analyzed medical records of 585 patients (mean age, 63.2 years; 58.3% men) who underwent surgery between July 2005 and October 2020, of whom 267 had AD and 318 had CC. Researchers divided data sets into training (74.4%), validation (15.4%) and testing (10.2%) groups.

Images of diseased bowel segments and surrounding mesentery from CT scans were used to create a 3-D convolutional neural network (CNN) as an AI support system, and 10 board-certified radiologists and radiology residents were tasked with classifying the testing cohort first without then with this support.

Compared with the mean radiologist sensitivity and specificity, the 3-D CNN reached a higher sensitivity for the test set (83.3%; 95% CI, 70-96.6 vs. 77.6%; 95% CI, 72.9-82.3) and specificity (86.6%; 95% CI, 74.5-98.8 vs. 81.6%; 95% CI, 77.2-86.1).

For all readers, AI support improved sensitivity from 77.6% to 85.6% and specificity from 81.6% to 91.3%. It also reduced false-negative findings from 78.5% to 86.4% and false-positives from 80.9% to 90.8%.

“We developed a 3-D CNN that can be implemented as an AI support system for the differentiation of CC and AD based on CT images,” Ziegelmayer and colleagues wrote. “Artificial intelligence support led to a significant increase in diagnostic performance of board-certified radiologists and radiology residents.”