Deep learning for bone segmentation: towards automating failure load simulations

Emile Saillard (1) , Marc Gardegaront (1, 2) , François Bermond (2) , David Mitton (2) , Jean-Baptiste Pialat (3, 4) , Cyrille B. Confavreux (4) , Hélène Follet (5, 1) , Thomas Grenier (6, 3)

1 LYOS - Physiopathologie, diagnostic et traitements des maladies osseuses / Pathophysiology, Diagnosis & Treatments of Bone Diseases
2 LBMC UMR T9406 - Laboratoire de Biomécanique et Mécanique des Chocs
3 CREATIS - Centre de Recherche en Acquisition et Traitement de l'Image pour la Santé
4 HCL - Hospices Civils de Lyon
5 INSERM - Institut National de la Santé et de la Recherche Médicale
6 MYRIAD - Modeling & analysis for medical imaging and Diagnosis

In this paper, we propose a dedicated pipeline of pre-processing, deep learning-based segmentation and post-processing for human femurs and vertebras segmentation of CT-scans volumes. For both tasks, we experimented with three U-Net architectures and showed that out-of-the-box models enable automatic and high-quality volume segmentation, if carefully trained.

Mots clés: bone segmentation; CT-scan; deep learning; femur; spine

Domaines: Biomécanique [physics.med-ph]

Emile Saillard, Marc Gardegaront, François Bermond, David Mitton, Jean-Baptiste Pialat, et al.. Deep learning for bone segmentation: towards automating failure load simulations. 48e Congrès de la Société de Biomécanique, Oct 2023, Grenoble, France. pp.S27-S29. ⟨hal-04288748⟩