Fully automated algorithm estimates muscle fascicle length from ultrasound image

Abstract While ultrasound is a useful tool for visualizing muscle in vivo, traditional analysis involves substantial manual labor. Semi-automated algorithms have been introduced in recent years, reducing the amount of time required for extracting pennation angles and fascicle lengths from ultrasound images. Unfortunately, semi-automated algorithms still require some user actions and thereby subjective decision making. We here present a freely available, fully automated feature detection algorithm that involves Hessian filtering to highlight line-like objects within the ultrasound image. Hough transform is used to determine muscle fascicle angles and feature detection is used to determine the location and angle of aponeuroses. As a demonstration, we test the algorithm on ultrasound images obtained from vastus lateralis muscle in healthy individuals (N = 9) during isometric knee extension moment production (0 – 45 N-m) at three knee angles (15-25 deg). Pennation angle, muscle thickness and fascicle length vary with knee moment and knee angle in line with previous observations. Specifically, fascicle length decreases with larger knee moments and increases towards knee flexion. We expect the proposed algorithm to be useful for estimating muscle fascicle lengths during cyclic movements like human locomotion.