Reliability of a new semi-automated algorithm to analyze vastus lateralis muscle architecture using extended field-of-view ultrasound.

Abstract

Fascicle curvature is often disregarded in muscle architecture analyses, leading to errors in fascicle length (FL) and pennation angle (PA). This study aimed to examine the reliability of a semi-automated method for analyzing vastus lateralis (VL) muscle architecture from extended field-of-view (EFOV) ultrasound images, using a Python-based algorithm that accounts for fascicle curvature. This method was compared to manual linear analysis performed in ImageJ. Additionally, the influence of fascicle curvature versus linearity across different VL regions was investigated. EFOV ultrasound images of the VL muscle from 102 athletes were analyzed by three raters. Muscle thickness (MT) was assessed along the entire VL belly, while FL and PA were measured in the proximal, middle, and distal regions. Intraclass correlation coefficients (ICC) and Bland-Altman were used to evaluate reliability and agreement. ANOVA was used to assess fascicle curvature across VL regions. Each rater demonstrated good-to-excellent reliability between the Python algorithm and ImageJ. The raters mean inter-software reliability was excellent for FL (ICC = 0.97), PA (ICC = 0.95), and MT (ICC = 0.99). Bland-Altman analysis revealed minor discrepancies between software. In the curved versus linear fascicle analysis, the proximal region exhibited pronounced fascicle curvature and was the only region where FL and PA were significantly underestimated when assuming linearity. The semi-automated Python algorithm provides a reliable tool for analyzing VL muscle architecture from EFOV ultrasound images while accounting for fascicle curvature. Assuming fascicle linearity may lead to underestimation of FL and PA, particularly in the proximal VL region.