T2 relaxation times relate to collagen network integrity and 2) absolute maximum shear strains and T1ρ relaxation times relate to PG loss. These hypotheses were tested by comparing the FE model results both qualitatively (distribution maps) and quantitatively (degeneration volumes) with changes inT2and T1ρrelaxation times.
Furthermore, we examined differences between patients with ACLR and controls in biomechanical parameters and MRI relaxation times.
Collagen network degeneration
Qualitatively, there was a good match between maximum principal stresses andT2
relaxation times. In cartilage locations with maximum principal stresses exceeding 7 MPa degeneration threshold for degeneration, the T2 relaxation time was also increased during the follow-up (seeFig. 7.5-Patients 1-3, 5and 7). ForPatient 4, both maximum principal stresses and the T2 relaxation times showed no degenerative signs.Patient 6may only have mild or no degeneration, evidenced by the maximum stress values near the 7 MPa threshold and by small changes in the T2 relaxation time. For almost all patients, the posterior aspect of the articular cartilage showed increased maximum principal stresses and T2 relaxation times.
Earlier studies have also reported increased OA susceptibility of the posterior aspect of the articular cartilage [126, 301–304], and is attributed to valgus collapse [302, 303]. In all controls both maximum principal stresses and T2
relaxation times were below the assumed degeneration thresholds (Fig. 7.5), indicating that these subjects will not develop OA. Importantly, all subjects in both groups predicted to be at risk or at no risk of OA onset and development by the FE model matched the follow-up MRI information (T2relaxation time). This highlights the importance of patient-specific analysis. In the future, more subjects will be added from other patient groups.
Quantitatively, using 7 MPa and 60 ms thresholds for degeneration, the predicted volumes from the FE models matched the measured volumes from T2 relaxation times, for both patients with ACLR and controls. Some of the differences between the predictions and the MRI finding can be influenced by the assumed degeneration threshold values, which are likely patient-specific, or other limitations listed below. In StudyIIand IV, other threshold values were tested for both the FE model parameters and MRI relaxation times; however, the best match between the FE model predictions and follow-up MRI information was obtained using 7 MPa, 32% and 60 ms, respectively.
There were significant differences between patients with ACLR and controls in terms of maximum principal stresses and changes inT2 and T1ρ relaxation times.
The predicted maximum principal stresses showed that some patients with ACLR are at higher risk of developing OA than others [6, 301]. Relaxation times showed similar patient-specificity. The distributions of peak maximum principal stresses and changes inT2 and T1ρ relaxation times were more clustered in controls with a small standard deviation, and the values were below the assumed degeneration thresholds. This was to be expected, since one of the selection criteria for the controls was that they should not show any degenerative signs at any time point.
Interestingly, using 7 MPa [33, 36, 39, 195, 196] and 10 MPa [35, 305, 306] as two maximum principal stress thresholds, we could differentiate patients into three
different risk groups for progression of OA. This was in agreement with the WORMS evaluation at the 3-year time point in ∼85% of the cartilage compartments. Therefore, this relatively simple and straightforward biomechanical method may be applicable for clinical risk assessment of OA. Even though the definition of these three risk levels was based on the results, it was however made by the authors. Further, it currently only applies to the studied groups. To generalize the grouping criteria, several factors in the model that may need to be adjusted to other patient groups are listed in limitations below.
PG loss and cartilage defects
Qualitatively, the relationship between absolute maximum shear strains and changes inT1ρrelaxation times was not as strong, particularly for patients with ACLR. Only Patients 1and2showed a good correspondence between absolute maximum shear strain and T1ρ relaxation time which may be attributed to these patients having cartilage lesions. ForPatient 1 the FE models predicted a decrease in FCD content near the lesion and on the surface of the lateral tibial cartilage using the strain-based mechanism. Highly localized FCD loss around the lesion was also predicted by the fluid-velocity driven mechanism. T2and PG-sensitiveT1ρrelaxation times were increased near the side of the lesion during follow-up. ForPatient 2, slight increases in both T1ρ and T2 relaxation times in the tibial cartilages during the follow-up support the FE model predictions of minor FCD loss near the cartilage defect on the lateral femoral cartilage compartment. Therefore, our results indicate that at a compartment level, absolute maximum shear strain may not indicate PG loss or changes in tissue integrity. On the other hand highly localized strain levels may be more important, and could indicate highly localized cell death and PG loss, even though global strain levels would not change. Moreover, based on the results of Study IV,Patient 1has a high-risk defect that directly contributes to knee OA onset and development, whilePatient 2has a moderate-risk defect that may not directly increase the risk of knee joint OA. Earlier studies also document the importance of location, shape and size of lesion on the mechanical response of articular cartilage [151, 233, 258, 307].
With regard to the T1ρ relaxation times, there were significant differences between patients with ACLR and controls, similar to T2. Generally, the values of T1ρwere 10 ms higher than those ofT2at both 1- and 3-year follow-up time points.
The range and distribution ofT1ρ were similar to those of T2. This would suggest that both T2 and T1ρ may be more indicative of the overall state of articular cartilage integrity than individual components [162, 171, 172, 176].
8.3 LIMITATIONS
Number of subjects. InStudy I, only one subject was included. However,Study Iis a methodological study showing that it is possible to obtain similar results from kinetic and kinetic-kinematic finite element models. Adding more subjects would not change the main conclusions ofStudy I. InStudies IIandIV, only two subjects were included. Both are proof-of-concept studies where the FE model predictions matchedT2 andT1ρ follow-up information. On the other hand, inStudy III, finite element models were generated for 13 subjects and the predictions matched MRI follow-up information, which is more than in most biomechanical modeling studies.
Material formulation for articular cartilage. A common limitation in Studies I-IIIis the material formulation for each soft tissue, with TIPE material formulation for articular cartilage, TIE material formulation for menisci and spring formulation for ligaments. A more accurate material formulation would be to model all tissues as fibril reinforced poroelastic materials [39, 62, 204, 204, 216, 230, 276, 308]. Recently, there have been some attempts to implement patient-specific collagen orientation and PG distribution of articular cartilage in FE models [199, 212]. However, this increases the complexity and the computation times of FE models, making clinical implementation challenging. Importantly, the TIPE material used in these studies is based on a previous study, in which the tissue response was found similar to the more complex FRPVE material [190]. The TIE formulation for menisci adequately describes the behavior of the tissue during short-term loading and can produce similar results as more complex formulations [164, 309, 310]. Both the spring and FRPVE formulation for ligaments produced similar results in terms of articular cartilage response in the knee [311].
Gait and implementation. Analyzing only walking could be considered a limitation. For these patients, motion capture was also done for cutting, single-leg-squats and drop-jump activities. However, it is unclear which motion implementation method would be the most suitable to incorporate these activities in FE modeling in a fast and reliable manner. On one hand, the kinetic-kinematic approach (forces and angles, same as used here), may not be feasible, since both internal-external and varus-valgus rotations may be unreliable due to skin marker movement and/or measurement cross-talk [235, 237]. On the other hand, the kinetic approach (forces and moments) could be used. However, this requires musculoskeletal modeling to estimate muscle forces and the inclusion of the patella-femoral joint. Again, this would increase the model complexity, making clinical implementation difficult. Several studies [15, 245, 247, 312, 313] have shown that patients with ACLR have kinematic differences between the injured and contralateral side, which may increase the risks of OA in both the tibiofemoral and patellofemoral joint compartments and could account for some discrepancies between the FE models and follow-up information. These activities, and other ACLR rehabilitation exercises will be included in future FE modeling studies.
Despite only modeling the stance phase of gait, the FE models were able to identify well patients at risk of OA onset and development and those that remain healthy in a 2-year longitudinal follow-up.
In terms of gait implementation, for the kinetic-kinematic driven FE models, used in Studies I-III, the rotations were acquired from motion capture.
Measurement inaccuracies such as cross-talk and skin marker movement may influence the internal-external and varus-valgus rotations from motion capture [235–237]. In all studies, the total joint force was produced primarily by the translational forces and ligament forces, due to their pre-strains and stiffnesses. In Studies Iand IV, the kinetic model used also included quadriceps muscle forces pulling the patella. However, the hamstring muscle group also contributes to the total joint contact force [298]. This may explain why in StudyI, the patellofemoral joint contact force was ∼0.5 BW lower than some of the literature reported values for both ModelsAandB. We acknowledge that musculoskeletal modeling could be used to estimate all muscle forces and the FE model could include all these muscles [209, 210]. In the kinetic-kinematic driven models used inStudies IIand III, adding other muscle groups should not influence the distribution of stresses and strains, since the muscle contributions are indirectly considered by the implemented rotations. Furthermore, adding the patellofemoral joint and using musculoskeletal modeling would drastically increase total times for generating and running the FE models (∼ 4 times longer/patient). Despite this, all FE models produced values within the range reported in the literature for rotations, joint reaction forces and contact pressures (see Study I and Study III in appendix).
Importantly, inStudies II-IV, the FE model predictions were in good agreement with the follow-up information from MRI. However, in the future the patellofemoral joint should be included in the analysis using faster FE model generation methods [195].
Under- and Overloading. On one hand, several studies related both impact and chronic overloading of articular cartilage with collagen network damage and OA onset and development [34, 224, 306, 314]. It has been shown that subject BMI is directly linked with overloading and knee OA [108, 110]. On the other hand, several studies have reported underloading in the ACLR limb in early follow-up during walking, running and cutting maneuvers [16, 140] and has been related to joint degeneration. Therefore, our assumption of cartilage overloading may not be always correct.
Inflammation. ACL injury and reconstruction is in most cases accompanied by an increased pro-inflammatory cytokine activity [132,141,315–320]. The effects of these biochemical markers on OA onset and progression could be considered in the FE models [202,203,321–323]. Recently, a new mechanobiological model combined both biochemical and biomechanical degeneration pathways to simulate OA progression inin-silicoinjured cartilage samples [324]. The methodology should be incorporated in future FE studies of ACLR. However, faster methods should be developed for incorporating these mechanisms, e.g., through machine learning.
T2andT1ρrelaxation time. The resolution of theT2andT1ρrelaxation time maps could lead to partial volume effects that may cause inaccurateT2andT1ρvalues near areas with high relaxation time differences (i.e. cartilage interface with synovial fluid or bone) [154, 162, 166, 325–327]. Higher resolution may not be feasible as this would drastically increase image acquisition time. In terms of MRI mapping, several studies [328,329] have reported high repeatability and reliability of MRI used to evaluate the cartilage morphology (i.e. WORMS). To the authors knowledge, the repeatability of T2 and T1ρ mapping for evaluating compositional changes in
knee joint cartilage has not been studied earlier. This could affect the estimation of cartilage degeneration from MRI in studies II-IV. However, several recent studies noted the repeatability ofT1-mapping of both fibroglandular and adipose tissues of breasts [330–332].
Degeneration thresholds. Constant degeneration thresholds were used in MRI analyses. Both T2 and T1ρ relaxation time values depend on the specific implementation of the respective measurement [326]. Further,T2has demonstrated susceptibility to the orientation of the tissue in the magnetic field [162]. However, we evaluated the change in T2 and T1ρ relaxation time, which is arguably more important than the exact values and the same measurement protocol and analysis was used at both follow-up time points, alleviating issues related to potential differences in the results.
The biomechanically simulated tissue degeneration used constant thresholds of stress and strain. These thresholds, and material properties of cartilage, could be adjusted in the future at least according to age, gender or physical activity [36, 333–
337].