Semi-automatic muscle segmentation in MR images using deep registration-based label propagation. (August 2023)

• Record Type: Journal Article
• Authors: Decaux, Nathan; Conze, Pierre-Henri; Ropars, Juliette; He, Xinyan; Sheehan, Frances T.; Pons, Christelle; Salem, Douraied Ben; Brochard, Sylvain; Rousseau, François
• Is Part Of: Pattern recognition. Volume 140(2023)
• Publication Date: 2023-08
• Subjects: Semi-automatic segmentation -- Musculoskeletal system -- Label propagation -- Deep registration; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2023.109529
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: Registration-based label propagation is used for intra-subject muscle MR segmentation. 3D few-shot segmentation is reached by propagating 2D labels using deep registration. Propagation is guided by image intensity, muscle shape and registration consistency. Bidirectional propagation uses registration quality estimation as weighting guidance. An unsupervised pre-training stage initializes the deep registration framework. Abstract: Fully automated approaches based on convolutional neural networks have shown promising performances on muscle segmentation from magnetic resonance (MR) images, but still rely on an extensive amount of training data to achieve valuable results. Muscle segmentation for pediatric and rare diseases cohorts is therefore still often done manually. Producing dense delineations over 3D volumes remains a time-consuming and tedious task, with significant redundancy between successive slices. In this work, we propose a segmentation method relying on registration-based label propagation, which provides 3D muscle delineations from a limited number of annotated 2D slices. Based on an unsupervised deep registration scheme, our approach ensures the preservation of anatomical structures by penalizing deformation compositions that do not produce consistent segmentation from one annotated slice to another. Evaluation is performed on MR data from lower leg and shoulder joints. Results demonstrate that the proposed semi-automatic multi-label segmentation model outperforms state-of-the-art techniques.