Scientists report a new deep learning tool based on Bayesian U-Net architecture that can segment individual muscles from CT images. The high accuracy of the results offers a new level of personalized biomechanical modeling for patients for better therapies and athletes for better performance.