Benchmarking open-source automated thigh muscle MRI segmentation algorithms

A study compared eight open-source tools for automatic segmentation of thigh muscle from magnetic resonance imaging (MRI). Six of them operate independently and two are designed as auxiliary algorithms after initial muscle labeling. Researchers tested these tools on three-dimensional MRI volumes from three base datasets and subsequently on derived datasets, evaluating both quantitative metrics and practical usability. Results revealed substantial performance differences between individual methods and datasets, with many methods showing reduced accuracy on pathological cases. Domain-specific models trained on large datasets consistently outperformed general-purpose models and newer architectures. The study revealed practical trade-offs between accuracy, generalizability, and ease of use that are critical for clinical adoption, and highlighted concerns about deployment on underrepresented populations and pathologies.