Author: Sara Allievi, Stefano Bonvini, Gloria Miori, Laura Orsingher, Andrea Passerini, Igor Raunig, Daniele Ravanelli, Erich Robbi, Annalisa Trianni 👨🔬
Affiliation: Department of Information Engineering and Computer Science, University of Trento, Vascular Surgery Department, S.Chiara Hospital, APSS, Medical Physics Department, S.Chiara Hospital, APSS 🌍
Purpose:
This study evaluates the performance of an AI-driven tool in segmenting and analyzing tissue composition in abdominal aortic aneurysm (AAA) patients, specifically focusing on the sealing zones (proximal and distal regions) to enhance Endovascular Aneurysm Repair (EVAR) planning. Accurate tissue characterization in these regions is crucial for determining the suitability of the vessel wall for graft fixation, reducing the risk of complications.
Methods:
Preoperative contrast-enhanced CT angiography (CTA) scans from 20 AAA patients were analyzed using an AI tool designed to segment and classify tissues in the sealing zones. The tissue was categorized into healthy, thrombotic, and calcified regions, following the ESVS 2024 guidelines. AI results were compared with manual annotations by two experienced vascular surgeons. Cohen’s Kappa coefficient was used to assess the agreement between the AI and manual classifications. Time efficiency between automatic and manual segmentations was compared using the Wilcoxon test.
Results:
The AI tool showed substantial agreement with manual assessments, achieving Cohen’s Kappa values of 0.77 for thrombotic tissue and 0.63 for calcified tissue. This indicates that the AI tool reliably characterizes tissue composition in sealing zones, offering a standardized method to support clinicians in identifying areas with different tissue types essential for EVAR planning. Additionally, the AI tool significantly reduced segmentation time, with a median of 25 minutes (95% CI: 20-30) compared to 51 minutes (95% CI:47-62) for manual segmentation (p<0.001).
Conclusion:
The AI-based tool is a promising aid for EVAR planning, providing accurate and repeatable tissue characterization of sealing zones. By automating tissue classification, it enhances surgical decision-making, improves planning efficiency, and may contribute to better patient outcomes through more precise graft placement and reduced complications, while saving time for clinicians.