Shape differences in the semitendinosus following tendon harvesting for anterior cruciate ligament reconstruction

Abstract Following hamstring autograft anterior cruciate ligament reconstruction (ACLR), muscle length, cross‐sectional area, and volume are reduced. However, these discrete measures of morphology do not account for complex three‐dimensional muscle shape. The primary aim of this study was to determine between‐limb semitendinosus (ST) shape and regional morphology differences in young adults following tendon harvest for ACLR and to compare these differences with those in healthy controls. In this cross‐sectional study, magnetic resonance imaging was performed on 18 individuals with unilateral ACLR and 18 healthy controls. Bilaterally, ST muscles were segmented, and shape differences assessed between limbs and compared between groups using Jaccard index (0–1) and Hausdorff distance (mm). Length (cm), peak cross‐sectional area (cm2), and volume (cm3) were measured for the entire muscle and proximal, middle, and distal regions, and compared between limbs and groups. Compared to healthy controls, the ACLR group had significantly (p < 0.001, Cohen's d = −2.33) lower bilateral ST shape similarity and shape deviation was significantly (p < 0.001, d = 2.12) greater. Shape deviation was greatest within the distal region of the ACLR (Hausdorff: 23.1 ± 8.68 mm). Compared to both the uninjured contralateral limb and healthy controls, deficits in peak cross‐sectional area and volume in ACLR group were largest in proximal (p < 0.001, d = −2.52 to −1.28) and middle (p < 0.001, d = −1.81 to −1.04) regions of the ST. Overall, shape analysis provides unique insight into regional adaptations in ST morphology post‐ACLR. Findings highlight morphological features in distal ST not identified by traditional discrete morphology measures. Clinical significance: Following ACLR, risk of a secondary knee or primary hamstring injury has been reported to be between 2‐to‐5 times greater compared to those without ACLR. Change in semitendinosus (ST) shape following ACLR may affect force transmission and distribution within the hamstrings and might contribute to persistent deficits in knee flexor and internal rotator strength.


| INTRODUCTION
Following anterior cruciate ligament (ACL) rupture, reconstruction (ACLR) is typically recommended to restore knee stability. 1 In Australia,~90% of ACLR harvest semitendinosus (ST), alone or in combination with gracilis (GR), for tendon autografts. 2,3 Following ACLR, harvested ST and/or GR tendons partly regenerated iñ 70% of patients. 4 Although athletes return to sport after ACLR, 1 long-term deficits in knee flexor and internal rotator strength persist, 5,6 due in part to postoperative morbidity of harvested muscle. Muscle morbidity is traditionally assessed through discrete morphological measures (i.e., length, cross-sectional area [CSA], volume). However, the full extent of morphological maladaptation post-ACLR may not be well described by these discrete measures as they do not assess the complex three-dimensional shape of muscle.
The ST is digastric, long, thin, and fusiform, 7 and theoretically well suited to high shortening velocities and large excursions. 8,9 A knee flexor and hip extensor, ST is also an important antagonist of external knee valgus moments which is recruited to support the knee during tasks such as landing and change of direction commonly performed during field and court sports and associated with ACL rupture. [10][11][12] Following tendon harvest for ACLR, ST experiences atrophy, 5,6 retraction, 13 fatty infiltration, 14 and impaired voluntary activation. 15 Post-ACLR strength deficits have been correlated with ST CSA and volume, indicating ST morphology affects its functional capacity in this clinical population. 5,16 Chronic deficits in ST CSA and volume have been observed as long as 7-years after ACLR, despite partial donor tendon regeneration in most patients. 17 Compared to intact ST, regenerated ST tendon tends to insert more proximal and medial on the tibia, within the popliteal fascia, and, in some cases, may not regenerate. 4,18 Altered ST morphology after ACLR may reduce torque generating capacity of the hamstrings via reductions in muscle forcegenerating capacity, moment-arm, or both. Quantifying the three-dimensional shape of the ST may better describe structural and functional consequences following ACLR.
Shape modeling has been used to study human anatomy following cardiac, 19 orthopedic, 20 and complex maxillofacial surgeries, 21 and to investigate musculoskeletal structures in the lower limb 22 and foot. 23 To date, no study has evaluated threedimensional muscle shape differences between limbs for those following hamstring ACLR. The aims of this study were to (1) determine bilateral ST shape similarity and regional morphology in those with ST-GR ACLR and compare to healthy controls; and (2) in those with an ACLR, determine bilateral ST shape similarity and regional morphology in those with and without postoperative tendon regeneration. We hypothesized (1) bilateral ST shape similarity would be lower in those with an ACLR when compared to healthy controls, and regional morphology differences would be more pronounced in the distal region of ST.
Additionally, (2) within the ACLR cohort those with ST tendon regeneration when compared to those with no ST regeneration will have greater bilateral ST shape similarity and regional muscle morphology.

| Participants
This level 3 cross-sectional study involved a secondary analysis of previously published data from individuals with a history of ACLR 5 and healthy controls. 22,24 As this was the first study to explore shape modeling of the ST following ACLR, it was not possible to base sample size estimates on previous reports. Instead, the secondary outcome of muscle volume was used. Previous studies have reported effect size of 1.52 6 and 1.71 25 when comparing surgically reconstructed to uninjured contralateral limbs and to healthy controls, respectively. Therefore, conservative sample size estimates were based on an anticipated effect size of 0.8 and a sample size of 12 was deemed sufficient to provide a statistical power of ≥0.8 when p < 0.05. All available data were used thereby increasing the total study population to 18  all participants providing their written informed consent before any testing.

| Volumetric shape modeling
To prepare the muscle meshes for volumetric shape similarity measurements between limbs, meshes were exported to 3-Matics Research (Materialise, v13). In healthy controls, a right-sided muscle was mirrored to the left. In the ACLR cohort, the injured muscle was mirrored to the uninjured side. To focus on shape difference rather than gross size differences between limbs in the ACLR cohort, the injured muscle was linearly scaled to approximate the length of the healthy contralateral muscle. The two muscle meshes were then rigidly aligned using N-point registration within 3-Matics Research using 3 points on the proximal musculotendinous junction (MTJ) and 3 points on the distal MTJ systematically selected by the user. A part comparison analysis was conducted within 3-Matics Research to produce a heatmap illustrating areas of greatest difference between the two meshes. The aligned meshes were then exported to Python (Python Software Foundation, v2.7) to perform shape similarity assessment.

| Shape similarity analysis
Shape similarity between injured and uninjured (ACLR) and left and right

| Discrete regional morphology analysis
Using the Measure and Analyze toolboxes within the Mimics software, muscle length (cm) and volume (cm 3 ) were calculated from the wrapped and smoothed muscle meshes. Fit Centreline tool was used to calculate the centroid of the muscle or tendon boundary for each axial slice. These centroids were then linked and linearly interpolated to create the intercentroid pathway, which represented the total length of the respective muscle or tendon.
The CSA at any given point along the intercentroid pathway was determined by fitting a plane to the outer boundary of the muscle but orthogonal to the pathway. This process was performed across the length of the pathway for each muscle with the largest value being recorded as peak CSA (cm 2 ). Muscle regions were defined by percentage of muscle length: proximal (100%-66%), middle (<66%-33%), and distal (<33%-0%). Morphological measures were normalized to limb length (m) for muscle length (cm . m − ¹) and mass (BW) (kg) x limb length (m) 27 for peak CSA (cm² . kg − ¹ . m − ¹), and volume (cm 3. kg − ¹ . m − ¹), respectively.

| Statistical analysis
All statistical analyzes were performed using SPSS v27 (IBM Corp.).
Results are presented as mean ± standard deviation (SD) for parametric variables, and frequencies and proportions for categorical or binary variables. Bilateral shape similarity measures (Jaccard index, RMSE, and Hausdorff distance) were compared between ACLR and healthy control cohorts using independent t-tests. A mixed-design MANOVA was used to assess interaction between group (Control/ACLR) and limb (left/right or uninjured/injured) for total and regional (proximal, middle, distal) discrete morphology variables (length, peak CSA, and volume). When significant interactions were detected, post hoc pairwise comparisons were used. If no significant difference between healthy control limbs was found, all subsequent analysis used an average of the measure across two limbs. To evaluate postoperative tendon regeneration, regenerated and non-regenerated subgroups were compared for bilateral ST shape similarity and all discrete morphology measures (total and regional) assessed using independent t-tests.

| Shape similarity
Compared to healthy controls, the ACLR cohort had significantly lower bilateral ST shape similarity ( Table 2   No between-group differences were found for ST length, peak CSA, and volume between the uninjured contralateral limb and healthy control limbs.

| Regional differences in muscle morphology
A significant group-by-limb (p < 0.001) interaction was observed for all regional muscle morphology measures.    (Figure 4).

F I G U R E 3
Mean difference of regional morphological measures in anterior cruciate ligament reconstruction (ACLR) participant (injured-uninjured). (A) Regional segments from a representative ACLR participant depicting semitendinosus (ST) muscle length in proximal (blue 100%-66%), middle (green < 66%-33%), and distal (red < 33%-0%) regions, (B) normalized peak cross-sectional area (CSA), and (C) normalized muscle volume across each region. CI: Confidence interval of the mean difference. The shortened ST is likely due to muscle-belly retraction immediately following harvesting and proximal migration of the distal musculotendinous junction. 37 Postoperative ST atrophy may shorten its fascicle length thereby reducing active operating range. 38 Reduction of physiological CSA will lower muscle force capacity and therefore impair ST generated posterior tibial drawer which supports the ACL. 39 The greatest deficits in ST peak CSA and volume were found in proximal and middle regions, as these regions comprise the majority of ST muscle mass (based on uninjured and healthy control data) and can be expected to experience the greatest postoperative loss in absolute terms. Interventions aimed at promoting hypertrophy in the ST or its agonists might be an important component of postoperative rehabilitation. 40 As hypothesized, we observed significantly longer muscle with greater volume in those with ST tendon regeneration post-ACLR compared to those without. Regionally, ACLR limbs with tendon regeneration had significantly greater proximal peak CSA and greater muscle volume in proximal and middle regions compared to those without regeneration. This highlights the importance of the tendon preserving surgical techniques and re-affirms the need to study the mechanism of tendon regeneration following graft harvesting.
This study had limitations that should be considered. First, the retrospective design means we cannot determine whether observed differences in ST morphology among ACLR, uninjured contralateral, and healthy controls were due to graft harvest or were present at ACL injury. However, healthy controls had no significant betweenlimb differences in ST shape or discrete morphology, suggesting observed interlimb differences in ACLR are unlikely to have predated injury. Second, we did not exclude participants with a history of hamstring strain, or other significant lower limb injury that did not require surgery, which may have confounded our results as hamstring atrophy and remodeling has been observed 5-23 months following hamstring strain. 40,41 Third, we did not account for previous training history of participants, which may contribute to altered ST shape and morphology. 27,42 However, training effects on ST are likely overshadowed by the substantial morphological changes following ACLR.
Fourth, due the retrospective nature of the analysis, a standardized follow-up time was unable to be implemented, this may have influenced the rate of tendon regeneration observed within this study. Fifth, although great effort and instruction were taken to ensure that positioning was maintained throughout the MRI procedure, interlimb differences in limb positioning was not quantified and could therefore introduce error into the shape analysis of the ST muscle. However, given both groups and legs would have been exposed to the same conditions, it is unlikely to affect bilateral or between-group comparisons, but might affect absolute values of shape variation. Finally, assessing functional implications of altered ST morphology was beyond the scope of this study. Future studies should aim to understand the consequences of ACLR tendon harvesting on the ST function during dynamic "high risk" movements and if muscle shape change and regional morphology are correlated to lower secondary ACL and hamstring injuries. This is in part due to the ST being an important knee flexor and antagonist of external valgus torque, [10][11][12] possibly protecting the ACL during complex movement tasks such as landing and cutting maneuvers. Further analyzes are warranted to examine the influence of shape change of ST on the internal muscle architecture parameters such as fascicle lengths and arrangement using imaging modalities such as diffusion tensor imaging or ultrasound.

| CONCLUSION
Shape similarity analysis provided unique perspective on ST morphology post-ACLR, highlighting shape differences compared to healthy ST in the distal region not captured by traditional discrete morphology measures. Different shape following ACLR might alter force transmission and distribution in the ST, which may have implications for load sharing amongst the hamstrings and knee flexor strength-endurance. Tendon regeneration in participants with ACLR was shown to result in both improved bilateral shape similarity regional morphology of the ST. Future work might focus on methods to reduce tendon harvest induced muscle morbidity and promote tendon regeneration.