An Artificial Intelligence Enabled Detection System to Identify Thoracic Aortic Aneurysms and Gaps in Timely Referral for Cardiac Surgery Evaluation

Objective
Thoracic aortic aneurysms (TAA) are often detected incidentally and can progress asymptomatically posing a significant risk for life-threatening aortic dissection or rupture. Despite established guidelines for surgical referral, a considerable number of patients experience delays in evaluation, representing a critical gap in care. Automated surveillance systems using artificial intelligence (AI) offer a promising solution to address this gap. This study evaluated the effectiveness of an AI software platform integrated with electronic health record (EHR) data to automatically identify patients with moderate or severe TAA and flag those without evidence of appropriate surgical follow-up.

Methods
We conducted a retrospective study of a quality improvement initiative within a three-hospital health system from July 2019 to July 2025. A rules-based, natural language processing algorithm analyzed all echocardiogram and chest CT reports to identify and classify TAA based on discrete aortic measurements. Severe TAA was defined as an ascending/root diameter ≥ 5.0 cm or an indexed diameter > 4.0 cm/m², and moderate TAA was defined as a diameter of 4.5 to < 5.0 cm or an indexed diameter > 2.72 to 4.0 cm/m². A care gap was identified if a patient meeting these criteria had no documented interaction with the cardiac surgery service in the EHR.

Results
The algorithm identified 1,047 patients with a TAA lacking surgical evaluation by a cardiac surgeon (Figure 1, top). Subsequent analysis of the EHR revealed that 937 (89%) of the these patients were never referred for consultation with a member of the cardiac surgery team. Among those that were ultimately referred for cardiac surgical evaluation, the median time to follow-up was 267 days for patients with moderate TAA, and 60 days for severe TAA. Patient age (p=0.049) and TAA severity (p<0.001) at the time of screening were associated with time to evaluation (Figure 1, bottom).]

Conclusions
The implementation of an AI-enabled system is a feasible and effective strategy for systematically identifying patients with significant TAA. By streamlining identification and prompting timely evaluation, this system supports adherence to established clinical guidelines and may help ensure appropriate surgical referral, ultimately reducing the risk of adverse aortic events.