Limb salvage prediction in peripheral artery disease patients using angiographic computer vision.

This article focuses on the development and evaluation of a computer vision (CV) model using artificial intelligence (AI) to predict limb outcomes in patients with peripheral artery disease (PAD) based solely on pedal angiograms. Utilizing a MobileNetV2-based model trained on 518 angiograms with known 3-month outcomes, the study demonstrated that the AI model achieved high accuracy (93% test accuracy) in distinguishing salvaged limbs from amputations and significantly outperformed the inframalleolar modifier in the Global Limb Anatomic Staging System (IM GLASS) in correlating predicted and actual outcomes. The model’s ability to analyze complex angiographic patterns without additional clinical data suggests potential applications in personalized treatment decision-making, institutional expertise adaptation, and quality assurance during vascular interventions. Limitations include reliance on still images, a 3-month outcome window, and exclusion of clinical variables, highlighting areas for future research.