286º taller internacional ENMC: Imagen muscular: inteligencia artificial, segmentación automática y uso compartido de...

Quantitative muscle MRI (qMRI) has emerged as a promising non-invasive biomarker for assessing neuromuscular diseases (NMDs). However, clinical implementation is limited by the significant time required for manual muscle segmentation, which restricts analysis to limited muscle regions rather than comprehensive whole-muscle assessment. The 286th European NeuroMuscular Centre (ENMC) workshop brought together 18 international participants from 10 countries to establish consensus on optimal qMRI acquisition protocols and automated analysis tools, revealing that while most centers utilize qMRI techniques, barriers to manual segmentation include limited expertise and excessive time requirements. Automated segmentation methods using machine learning architectures, particularly 3D U-Net models, have demonstrated promising results for individual muscle segmentation. Multi-center studies are starting to implement standardized protocols, while machine learning approaches can distinguish among many NMDs with higher accuracy than human experts. Data sharing platforms and federated learning approaches address the need for larger NMD cohorts with standardized and vendor-agnostic data formats, while maintaining patient privacy. The integration of automated 3D muscle segmentation tools integrated into clinical workflows represents a transformative advancement to revolutionize diagnosis, disease monitoring, and therapeutic assessment in NMDs. This consensus workshop provides a roadmap for accelerating the translation of qMRI from research tools to clinically implemented biomarkers for NMD management.