Summary
A computational model incorporating subject-specific geometries of the articular surfaces, menisci, and ligament insertions can identify knees prone to increased ACL force with valgus loading.
Abstract
Introduction
Young athletes participating in cutting and pivoting sports are vulnerable to ACL rupture. Articular surface geometries are risk factors for non-contact ACL injury and also drive knee mechanics. Thus, measuring knee mechanics, like ACL force, under loads associated with noncontact ACL injury is critical to identify those at the greatest risk of injury. Computational models incorporating subject-specific knee geometries can estimate ACL force non-invasively and under loads that are potentially injurious. However, the credibility of these models needs to be established. To this end, we used an experimental-computational knee modeling workflow to estimate knee mechanics and compare these estimates to a corresponding cadaveric experiment. We hypothesized that estimates of ACL force and tibiofemoral kinematics from a computational model are correlated with corresponding measurements from a cadaveric experiment.
Methods
Regarding the cadaveric experiments, ten unpaired cadaveric knees (5 males and 5 females; age: 33 ± 7 years) were procured. Anatomical coordinate systems were defined from bony landmarks. Knees were mounted to a robotic manipulator and tested using established protocols. Compression (100 N) and a valgus moment (8 Nm) were applied serially at 15° of flexion. All degrees of freedom except flexion were left unconstrained. Outcomes were: 1) ACL force at peak applied load, 2) internal tibial rotation (ITR), and 3) anterior tibial translation (ATT). Regarding computational modelling, a previously published modelling pipeline was utilized. Briefly, tibiofemoral bone, cartilage and meniscal geometries were segmented from magnetic resonance imaging (MRI). Subject-specific ligament insertions and origins were identified. Soft tissues were assigned population-mean structural properties, and slack lengths were standardized. The same coordinate systems were registered between computational and cadaveric experiments. The same loads were applied, and the same outcomes were collected as for the cadaveric experiment. The correlation between model and experiment outcomes was assessed via Pearson correlation coefficient (a=0.05) and regression coefficients were reported.
Results
The correlation coefficient (r) between computational model estimates and cadaveric experiment measurements for ACL force was 0.81 (p = 0.005). Of the five kinematic variables, ITR (r = 0.73, p = 0.02) and ATT (r = 0.68, p = 0.03) were correlated between computational model and cadaveric experiment. The correlations for the other kinematic variables were not statistically significant. ACLF, ITR, and ATT had regression coefficients of 0.64 ±0.16 N/N, 0.64 ±0.21 °/°, and 0.33 ±0.13 mm/mm, respectively.
Discussion
The strong correlation for ACL force (r = 0.81) indicates that our geometry-specific computational model can distinguish those knees prone to increased ACL force in response to valgus loading. Regression coefficients <1 indicate that models with population-mean ligament properties may systematically underestimate ITR and ACL force. Tuning parameters like ligament stiffness and slack length may improve agreement, if necessary. In conclusion, a computational model that includes subject-specific geometric data may be informative in identifying individuals prone to elevated ACL loads. This finding is promising for use of computational modeling as a tool for surgical planning in the clinical setting since MRI data are typically obtained following ACL injury.