Communication Dans Un Congrès Année : 2025

Automatic Segmentation of Lower-Limb Arteries on CTA for Pre-surgical Planning of Peripheral Artery Disease

Résumé

Peripheral artery disease (PAD) often requires revascularization guided by computed tomography angiography (CTA). Manual analysis of lower-limb arteries on CTA is time-consuming, operatordependent, and limited by challenging imaging conditions. While deep learning has advanced vascular segmentation, current models focus on large central vessels and do not capture the full arterial anatomy and pathological conditions in the lower extremities for PAD. In this work, we present a fully automated 3D segmentation pipeline tailored to PAD to delineate the entire lower-limb arterial tree, including the recognition of each main and peripheral branches, along with calcified plaques and stents. Our framework leverages nnU-Net, and addresses the limitations in complex PAD settings by introducing (1) a PAD-specific annotation protocol using a semi-automated tool to reduce inter-observer variability in labeling stents and calcifications, (2) a novel object-level detection metric that accounts for boundary ambiguity in calcified lesions and stents, and (3) the incorporation of the clDice loss to enhance topological preservation, critical for vascular analysis in a pathological setting. We train and evaluate our method on a curated in-house dataset under challenging imaging conditions. Our pipeline demonstrates robust performance, enabling the extraction of clinically relevant features necessary to the pre-surgical planning. It paves the way for fully automated PAD assessment, with potential to improve diagnostic accuracy, reduce time to treatment, and more objective severity scoring for therapy planning.

Fichier principal
Vignette du fichier
AMAI2025_Paper59.pdf (2.8 Mo) Télécharger le fichier
Origine Fichiers éditeurs autorisés sur une archive ouverte
Licence

Dates et versions

hal-05225711 , version 1 (27-08-2025)

Licence

Identifiants

  • HAL Id : hal-05225711 , version 1

Citer

Lisa Guzzi, Maria A Zuluaga, Fabien Lareyre, Gilles Di Lorenzo, Sébastien Goffart, et al.. Automatic Segmentation of Lower-Limb Arteries on CTA for Pre-surgical Planning of Peripheral Artery Disease. MICCAI-AMAI2025 - 4th Workshop on Applications of Medical Artifical Intelligence, Sep 2025, Daejeon, South Korea. ⟨hal-05225711⟩
706 Consultations
342 Téléchargements

Partager

  • More