Training Towards Precise Control Over Muscle Activity with Real Time Biofeedback Based on Ultrasound Imaging

Abstract

Automated algorithms for analysis of B-mode muscle ultrasound videos are computationally expensive. Thus, we propose a technique based on the comparison of pixels intensities between frames, that processes ultrasound recordings almost instantaneously, without special processing hardware. We verify the suitability of this method for feedback training of muscle activation. 15 able-bodied volunteers took part in the study practising contractions of different torque outputs of the gastrocnemius muscle. Within 3 s of each trial, they received visual feedback reflecting the ultrasound video analysis, visualizing how close their attempt was to the target, allowing them to modify their strategy for the next trial. A tendency to get closer to the target contraction intensity was recorded for 60.1%, 74.4%, 67.8% of trials for minimal, weak and medium contractions, respectively. Considering the number of successful trials (within 10% of the target), 11 participants improved in the second session in the weak and 10 in the medium contraction task. We demonstrated that the method provides easily interpreted semi-real time feedback of muscle activation that could help with muscle training.