An automatic fascicle tracking algorithm quantifying gastrocnemius architecture during maximal effort contractions

Background. Ultrasound has become the gold-standard for making dynamic measurements of muscle structure during functional movements in biomechanical studies. Manual measurements of fascicle length and pennation angle are time intensive which limits the clinical utility of this approach while also limiting sample sizes. The purpose of this study was to develop a novel tracking paradigm to quantify individual fascicle length and pennation measurements during maximal voluntary contractions and demonstrate is repeatability between days and reproducibility between different examiners. Methods. Five healthy young adults performed maximal isokinetic contractions at 0, 30, 120, 210, and 500 degrees about their ankle on an isokinetic dynamometer while their gastrocnemius muscle was observed using ultrasound. Individual muscle fascicles were identified in the first frame, and tracked using the automatic fascicle tracking algorithm and a manual approach by three observers on three separate days. Repeatability within examiners across days and reproducibility across examiners and days was evaluated using intraclass correlation coefficients. Agreement between manual and automatic tracking was evaluated using the coefficient of multiple correlations. Supervised automatic tracking was performed on all videos by one examiner to evaluate the fidelity of automatic tracking in practice. Results. We found both manual and automatic measurements of fascicle length and pennation angle to be strongly repeatable within examiners and strongly reproducible across examiners and days (ICCs>0.76). There was greater agreement between manual and automatic measurements of fascicle length than pennation angle, however the mean CMC value for both was still found to be strong in both cases (CMC>0.8). Supervision of automatic tracking greatly showed very strong agreement between manual and automatic measurements of fascicle length and pennation angle (CMC>0.94). Conclusions. We have developed a novel automatic fascicle tracking algorithm that quantifies fascicle length and pennation angle of individual muscle fascicles during dynamic contractions across a range of velocities. We demonstrated that this fascicle tracking algorithm is repeatable and reproducible across different examiners and different days and showed strong agreement with manual measurements, especially when tracking is supervised by the user so that tracking can be reinitialized if poor tracking fidelity is observed.