KEY HIGHLIGHTS

• In a prospective study of 79,163 post-contrast CT examinations published in Clinical Imaging, 40.7% of superior mesenteric artery occlusion (SMAO) cases were missed on the initial interpretation.
• Only 2.4% of QA-detected cases were CT angiography exams, compared with 46.3% of cases identified on the initial read, suggesting the SMA is less consistently included in the search pattern on conventional post-contrast CT.
• An AI model optimized for SMAO detection identified 83 missed cases in six weeks, providing an additional safety net when this rare but critical finding falls outside the radiologist's initial search pattern.

New research published in Clinical Imaging suggests that superior mesenteric artery occlusion (SMAO), a life-threatening vascular emergency requiring rapid diagnosis and treatment, may be overlooked on CT interpretation more often than previously recognized. The findings highlight a significant opportunity for radiologists and radiology practices to strengthen evaluation of the superior mesenteric artery and potentially improve patient outcomes.

Conducted by researchers from vRad and led by Robert J. Harris, Director of Imaging AI, the study prospectively evaluated 79,163 post-contrast CT examinations with abdominal coverage from 766 imaging facilities over a six-week period. A quality assurance AI model analyzed each exam and flagged cases in which SMAO was suspected but not mentioned in the original radiology report. Those flagged examinations were then reviewed by a second radiologist as part of the quality assurance process. Through this approach, researchers confirmed 83 cases of SMAO that had been missed on the initial interpretation, representing 40.7% of all cases identified during the study period.

The findings suggest that SMAO may be overlooked more often than previously recognized and highlight opportunities to improve detection through both radiologist awareness and AI-supported quality assurance tools.

Acute SMAO can progress rapidly to bowel ischemia and infarction, with significant morbidity and mortality when it is not identified and treated promptly. It is also rare—the study measured an incidence of 0.26%—and often inconspicuous. When the vessel is fully occluded, its attenuation can approximate that of the surrounding mesentery, leaving little visual cue to draw the eye. In early acute cases, the rest of the abdomen and pelvis may look unremarkable apart from a small focal thrombus.

The study’s most actionable finding concerns why these findings might have been missed. None of the cases where the pathology was missed on the original radiologist report had SMAO or mesenteric ischemia mentioned in the clinical indication, compared with 17.4% of cases diagnosed correctly on the first read. QA-detected cases were also far less likely to be CT angiography exams—2.4%, versus 46.3% of initially diagnosed cases. Taken together, the authors conclude that search pattern, rather than an inability to recognize SMAO once the vessel is carefully assessed, is the principal contributor to missed cases.

That distinction is important for practices, because it points to a fixable problem. Radiology practices may benefit from specifically encouraging radiologists to include the SMA more consistently in their search pattern, said vRad Chief Medical Office, Dr. Benjamin Strong, particularly on conventional post-contrast CT where the intensity changes are more subtle than on CTA.

Additional analysis of our medical malpractice confirms SMA occlusion misses to be a meaningful contributor to poor patient outcomes and malpractice liability. Our focus on this issue—through AI algorithm development and encouraging our radiologists to include the SMA in their search pattern on post-contrast abdominal CT—has significantly reduced it as a problem in our practice.

— Benjamin W. Strong, MD, Chief Medical Officer, vRad

The published research also evaluated the AI model itself. Deployed prospectively, it demonstrated 67.6% sensitivity and 99.6% specificity. Exams with a positive model result reached a radiologist substantially faster, with a median delay of 5.1 minutes compared with 27.9 minutes for model-negative exams — a measure of the value of AI-based worklist prioritization for time-sensitive findings.

The SMAO model is one of 19 pathologies that run automatically on all relevant examinations entering the vRad system, providing worklist prioritization and an AI QA safety net for vRad reading radiologists. AI does not replace the radiologist’s search pattern or diagnostic judgment, but for a rare and easily overlooked finding it can provide an additional opportunity to identify a consequential result when it is missed on the initial search. External radiology practices using The vRad Platform also benefit from these AI models in the form of worklist prioritization.

The SMAO model has been nominated for AuntMinnie’s Best New Radiology Software award.

The full study is available in Clinical Imaging: https://www.sciencedirect.com/science/article/abs/pii/S0899707126001026

ABOUT VRAD

vRad (Virtual Radiologic) is the nation’s leading teleradiology practice with 500+ U.S. board-certified or eligible physicians, the majority of whom are subspecialty trained. Our practice is accredited as a telemedicine provider by The Joint Commission and delivers high-quality diagnostic imaging services to 1,600 facilities and radiology groups across the United States. vRad has 23 issued patents for telemedicine and radiology technologies; and is a leading innovator in the areas of artificial intelligence, machine learning, imaging data analytics, and software to improve the quality of patient care, value for our clients, and the experience of our physicians. www.vrad.com.

View source version on businesswire.com: https://www.businesswire.com/news/home/20261027673556/en/