Relative performance of automated fiber-tracking approaches to 3D muscle architecture and their utility for comparative biomechanics.

Abstract

Muscle architecture is a key determinant of muscle performance, informing the calculation of functional traits such as force production and excursion distance. Due to its importance in understanding vertebrate biomechanics and ecology, several methods have been developed for reconstructing muscle fascicle architecture from CT and other contrast imaging data. Two automated methods have gained prominence and have been widely applied across vertebrate taxa: the Xfiber extension for the commercial Avizo/Amira software and the Goodfibes R package. Here, we present a comparison of muscle fascicle reconstructions using Xfiber and Goodfibes across a variety of CT datasets varying in muscle morphology and data quality to test the comparative performance of each method. We compared fiber length, fiber orientation, and fiber tortuosity/curvature results for each method across all datasets, with the first published comparison of orientation measurements between the two software packages. We found that Xfiber and Goodfibes produced broadly similar fiber length and orientation measurements across high quality datasets, but Goodfibes results diverged at lower resolutions and could not follow highly curved muscle fascicles. Tortuosity/curvature measurements were more divergent between the two tools, partially as a result of differing filtering methods. Ultimately, we find both methods to be highly promising for future comparative studies, though Goodfibes users will need to be slightly more considerate of data quality and muscle morphology.