AAA detection gets boost from commercially available AI algorithm
A commercially available artificial intelligence algorithm could enhance the incidental detection of abdominal aortic aneurysms (AAA) in busy settings where they are more susceptible to being overlooked.
CT scans of the abdomen and pelvis are routinely conducted in emergency settings and often produce some sort of incidental finding. Among these findings are AAAs, which are estimated to be incidentally detected in around 4% of abdominal CT scans conducted in emergency settings. This figure, however, may be much higher, as some experts contend that smaller AAAs are present on as many as 35% of abdominal CT scans from EDs, but do not get reported.
The authors of a new study in Clinical Imaging believe this is an area where opportunistic AI can help.
“Abdominal aortic aneurysm (AAA) is a common incidental finding on CT imaging performed in the acute care setting. Artificial intelligence (AI) algorithms have been developed to automatically measure aortic lumen size and thus facilitate AAA detection,” Joseph Cavallo, MD, with the Department of Radiology and Biomedical Imaging at Yale, and colleagues explained. “However, few studies have evaluated the performance of such tools in a large clinical setting.”
To test their theory, researchers retrospectively applied a commercially available opportunistic AI algorithm (AIDOC, Tel Aviv, Israel) to all the abdominal and pelvic CT scans completed at a tertiary academic center's ED between July 2020 and May 2021. Natural language processing software was used to review the original reports for the presence of AAA; the software flagged discrepancies between AI’s interpretations and the original reports for review by a designated emergency radiologist.
Of the 4,023 cases analyzed, 98.3% of the original reports did not mention the presence of AAA. The algorithm flagged 16 of these cases as discrepancies needing review, of which 31% were determined to be truly positive by the ED radiologist. The group estimated that the algorithm enhanced AAA detection by 7.4% without disrupting workflows.
“Although the additional AAAs identified by the algorithm in this study were classified as low risk based on their size, they meet the threshold for continued surveillance. Extrapolated over a larger population, some of these low-risk aneurysms would presumably progress to sizes requiring intervention,” the group wrote, adding that similar algorithms could improve screening rates without significantly impacting radiologists’ reading burdens as well.
The authors acknowledged that the algorithm has limitations, such as imperfect measurements (typically overestimations), but they signaled optimism for how it could improve opportunistic detection of AAAs requiring monitoring and/or treatment, nonetheless.
Learn more about the findings here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.