Résumé
Detailed morpho-functional analysis requires quantification of both hard and soft tissues. Muscular traits (volume, weight, fiber length and angles) are particularly important because they determine mobility. Muscle dissections and chemical treatment have been and are still being used to quantify these traits, in particular to calculate muscle PCSA and force. At the same time, technical advances in μCT scanning now enable to access the same traits less destructively and in 3D. Recently published algorithms, such as the one included in the R package GoodFibes, aim to automatically trace and model muscle fibers or fascicles, allowing to display them graphically, but also to extract lengths and angles. Here we present a semi-automatic pipeline combining Fiji/ImageJ scripts and the GoodFibes package to extract muscle fibers in 3D reconstructions of six species of nesomyid rodents. We compare the results of this pipeline to those obtained by dissection and dissolution of muscles in the same specimens. Although the pipeline allows for the rapid acquisition of larger amounts of data than dissection, the paired results obtained by the two methods are quite different from each other in some specimens. Factors that appear to play a role in explaining large discrepancies are: 1) the orientation of the 3D reconstruction slices relative to that of the muscle fascicles; 2) the delimitation between fascicles, which is affected by staining, contrast and scan definition; 3) the anatomy, especially when the orientation of fascicles varies within a given muscle. These pitfalls mean that users should not take the output of GoodFibes at face value, but fortunately they are fairly easy for users to detect thanks to the visualization capabilities of the package in association with other software.