Real-time muscle fascicle length measurement via machine learning

To develop assistive devices that can account for muscle dynamics, we developed a machine learning algorithm that measures muscle fascicle lengths in real-time from live ultrasound imaging. Our algorithm’s predictions yielded up to 95% correlation and R2 = 0.89 with respect to hand-tracked images, opening the door towards real-time feedback for device control with muscle dynamics in-the-loop.