Early clinical experiences with AI-based EVAR planning using the Endoleak Risk Index support its value for...

The aim of this study was to evaluate our first experience with the use of the artificial intelligence-based Endoleak Risk Index (ERI) in the planning of infrarenal endovascular aortic repair with special regard on its impact on clinical decision-making.

This single-center study evaluated two patient groups treated with Endurant endovascular aortic repair (EVAR). Group 1 comprised a retrospective cohort with at least 3 years of follow-up. The ERI was calculated for this group from preoperative computed tomography angiography scans and compared with the actual outcome. The prospective group 2 included patients scheduled for elective EVAR from March 2024 to March 2025, with preoperative AI-based simulations for endograft sizing including ERI calculation for type 1a endoleak (EL1a) risk prediction. The influence of ERI on clinical decision-making was assessed. Patients with noncontrast computed tomography scans or scans with slice thickness greater than 3 mm were excluded.

Twenty patients were included, with 10 in each group and a median age of 70 years (range, 59-81 years) in group 1 and 72 years (range, 60-84 years) in group 2. In group 1, the ERI was elevated in six of 10 cases, with four patients experiencing perioperative or late EL1as during follow-up. Notably, two patients with elevated ERI did not develop EL1a over follow-up periods of 93 and 44 months, whereas all patients with low ERI remained free of endoleaks. Two patients in group 2 had low ERI and no EL1a, whereas two had elevated ERI for at least one simulated endograft size, leading to a change in treatment (larger endograft) for one patient. The remaining six patients had elevated ERI for all simulated sizes, with one case being unsuitable for infrarenal EVAR. Despite elevated ERI, the initial treatment plan remained unchanged for four patients, one of whom died due to cardiac reasons before implantation. Overall, no patients in group 2 developed EL1a during a median follow-up of 3 months (range, 1-12 months).

This pilot study suggests that ERI calculation can be valuable even in straightforward cases, emphasizing the importance of education in the EVAR planning process. With further validation from larger datasets and advancements in technology, artificial intelligence-based EL1a risk prediction has the potential to significantly enhance EVAR planning in the future, promoting patient safety by providing tailored treatment strategies and supporting the surgeon in his decision-making.