Biomechanical modeling of musculoskeletal (dys-)function requires detailed information on muscle architecture¹, which is defined as the internal arrangement of muscle fibers at the organ level. Diffusion Tensor Imaging (DTI) in combination with tractography facilitates 3D visualizations of the muscle architecture^2,3. However, reconstructed tracts tend to continue along the tendon or aponeurosis when using Fractional Anisotropy (FA) and fiber curvature as a cutoff. Therefore, in order to obtain accurate determination values of fiber length, not only the external boundaries of individual muscles, but also the internal tendons and aponeurosis, have to be carefully segmented⁴. To avoid the overestimation of fiber length and tedious segmentation, we recently introduced a fast semi-automatic method to identify tendinous structures⁵. The method is based on the notion that skeletal muscle fiber tract density should be constant, as muscle fibers span the whole distance from tendon to tendon. On the other hand, in tendons the tract density increases because calculated tracts converge and continue into the tendinous structure due to partial volume effects. Therefore, muscle and tendons can be distinguished using a tract-density (TD) threshold (figures 1a and 1b), facilitating fiber length determination from tendon to tendon. The aim of this study was to assess the repeatability of muscle fiber length estimation in the lower legs using TD as a threshold. Furthermore, changes in fiber length as a result of changes in foot position were assessed.

Five healthy male volunteers were scanned with a 3T Achieva MRI scanner (Philips). Measurements of the lower leg were performed twice (30 min between measurements) with the foot in 15° dorsiflexion, neutral, and 30° plantarflexion positions, using a custom-built device.

Scan parameters were: DTI: SE-EPI; FOV: 192x156mm²; TE/TR: 51.63/11191ms; matrix size: 64x52; slices: 50; voxel size: 3x3x5mm³; SENSE: 1.5; gradient directions: 12; bvalue: 400s/mm²; Fat suppression: SPAIR and olefenic fat-suppression⁶. mDixon: FFE; FOV 192x156 mm²; TR: 7.7 ms; TE1/ΔTE: 2.1/1.7 ms; matrix size 192x192; slices: 100; voxel size: 1x1x2.5mm³. Total scan time: 33 minutes.

Data preprocessing and tensor fitting was performed using DTITools for Mathematica 10.3⁷ and tractography was performed using VIST/e⁸. TD-maps were made by whole volume tractography (seeding 1 tract/mm³) with the following constraints: 0.1<FA<0.7, fiber angle <20° per 0.6mm step, and fiber length >2cm. TD values were defined as the number of tracts in a voxel and values were normalized to the average tract density of the entire volume (mostly muscle, therefore TDmuscle ≈ 1). The Soleus (SOL), Fibularis Longus (FL), Extensor Digitorum (EDL) and the Tibialis Anterior (TA) were roughly segmented using ITKsnap. Next, tractography was performed per muscle and adding TD <1.5 as a constraint. Repeatability was explored using Bland-Altman analysis. Changes in fiber length between the 3 foot positions was tested using a repeated multivariate analysis of variance (MANOVA).

For all scans TD-maps were successfully created (Figure 1a-c) and tractography was performed (Figure 2a-f), avoiding artificial long and short fibers using TD as a stopping criteria. The Bland-Altman plots for the muscles (Figure 3), show small variations and clustered muscles. CV for the 4 muscles together was 12.6%, for the individual muscles they were 9.4%, 11.3%, 11.6% and 17%, for SOL, FL, EDL and TA, respectively. Average fiber length of SOL, FL, EDL and TA are shown in Figure 3 as a function of different ankle positions. Overall, the plantarflexors (SOL and FL) show significant increase in fiberlength (P<0.05) from plantarflexion to neutral position, while the dorsiflexors (EDL and TA) show a significant (P<0.05) increase in fiber length from dorsiflexion to neutral.The use of TD-maps for automatic segmentation of tendinous structures is reproducible. Furthermore, it allowed detection of changes in fiber lengths due to passive stretch of the muscle fibers with respect to the neutral ankle position. The observed behavior of fiber length change is in agreement with the known antagonistic function of the muscles. Furthermore, the mean fiber lengths found in this study are in agreement with previous cadaveric and ultrasound studies (Figure 4)^3,9. The proposed technique can be used for input in biomechanical models as well as in longitudinal studies e.g. follow-up after training, rehabilitation, and surgery. Another important strength of our method is that it also enables visualization of the tendinous structures, potentially enabling automatic measurements of the pennation angle⁴. In conclusion, tractography constrained by TD allows for repeatable estimates of fiber length.