Prediction on the medial knee contact force in patients with knee valgus using transfer learning approaches: Application to rehabilitation gaits

Prediction of medial knee contact force using multisource fusion recurrent neural network and transfer learning

Estimation of knee contact force (KCF) during gait provides essential information to evaluate knee joint function. Machine learning has been employed to estimate KCF because of the advantages of low computational cost and real-time. However, the existing machine learning models do not adequately consider gait-related data's temporal-dependent, multidimensional, and highly heterogeneous nature. This study is aimed at developing a multisource fusion recurrent neural network to predict the medial condyle KCF. First, a multisource fusion long short-term memory (MF-LSTM) model was established. Then, we developed a transfer learning strategy based on the MF-LSTM model for subject-specific medial KCF prediction. Four subjects with instrumented tibial prostheses were obtained from the literature. The results showed that the MF-LSTM model could predict medial KCF to a certain high level of accuracy (the mean of ρ = 0.970). The transfer learning model improved the prediction accuracy (the mean of ρ = 0.987). This study shows that the MF-LSTM model is a powerful and accurate computational tool for medial KCF prediction. Introducing transfer learning techniques could further improve the prediction performance for the target subject. This coupling strategy can help clinicians accurately estimate and track joint contact forces in real time.

Osteoarthritis year in review 2023: Biomechanics

Biomechanics plays a significant yet complex role in osteoarthritis (OA) onset and progression. Identifying alterations in biomechanical factors and their complex interactions is critical for gaining new insights into OA pathophysiology and identification of clearly defined and modifiable mechanical treatment targets. This review synthesized biomechanics studies from March 2022 to April 2023, from which three themes relating to human gait emerged: (1) new insights into the pathogenesis of OA using computational modeling and machine learning, (2) technology-enhanced biomechanical interventions for OA, and (3) out-of-lab biomechanical assessments of OA. We further highlighted future-focused areas which may continue to advance the field of biomechanics in OA, with a particular emphasis on exploiting technology to understand and treat biomechanical mechanisms of OA outside the laboratory. The breadth of studies included in this review highlights the complex role of biomechanics in OA and showcase numerous innovative and outstanding contributions to the field. Exciting cross-disciplinary efforts integrating computational modeling, mobile sensors, and machine learning methods show great promise for streamlining in vivo multi-scale biomechanics workflows and are expected to underpin future breakthroughs in the understanding and treatment of biomechanics in OA.

Accurately and effectively predict the ACL force: Utilizing biomechanical landing pattern before and after-fatigue

BACKGROUND AND OBJECTIVE

As a fundamental exercise technique, landing can commonly be associated with anterior cruciate ligament (ACL) injury, especially during after-fatigue single-leg landing (SL). Presently, the inability to accurately detect ACL loading makes it difficult to recognize the risk degree of ACL injury, which reduces the effectiveness of injury prevention and sports monitoring. Increased risk of ACL injury during after-fatigue SL may be related to changes in ankle motion patterns. Therefore, this study aims to develop a highly accurate and easily implemented ACL force prediction model by combining deep learning and the explored relationship between ACL force and ankle motion pattern.

METHODS

First, 56 subjects' during before and after-fatigue SL data were collected to explore the relationship between the ankle initial contact angle (AIC), ankle range of motion (AROM) and peak ACL force (PAF). Then, the musculoskeletal model was developed to simulate and calculate the ACL force. Finally, the ACL force prediction model was constructed by combining the explored relationship and sparrow search algorithm (SSA) to optimize the extreme learning machine (ELM) and long short-term memory (LSTM).

RESULTS

There was almost a stronger linear relationship between the PAF and AIC (R = -0.70), AROM (R2 = -0.61). By substituting AIC and AROM as independent variables in the SSA-ELM prediction model, the model shows excellent prediction performance because of very strong correlation (R2 = 0.9992,  MSE = 0.0023,  RMSE = 0.0474). Based on the equal scaling by combining results of SSA-ELM and SSA-LSTM, the prediction model achieves excellent performance in ACL force prediction of the overall waveform (R2 = 0.9947,  MSE = 0.0076,  RMSE = 0.0873).

CONCLUSION

By increasing the AIC and AROM during SL, the lower limb joint energy dissipation can be increased and the PAF reduced, thus reducing the impact loads on the lower limb joints and reducing ACL injuries. The proposed ACL dynamic load force prediction model has low input variable demands (sagittal joint angles), excellent generalization capabilities and superior performance in terms of high accuracy. In the future, we plan to use it as an accurate ACL injury risk assessment tool to promote and apply it to a wider range of sports training and injury monitoring.