A machine learning system for artificial ligaments with desired mechanical properties in ACL reconstruction applications.

The anterior cruciate ligament is one of the important tissues to maintain the stability of the human knee joint, but it is difficult for this ligament to self-heal after injury. Consequently, transplantation of artificial ligaments (ALs) has gained widespread attention as an important alternative treatment method in recent years. However, accurately predicting the intricate mechanical properties of ALs remains a formidable challenge, particularly when employing theoretical frameworks such as braiding theory. This obstacle presents a significant impediment to achieving optimal AL design. Therefore, in this study, a high-precision machine learning model based on an artificial neural network was developed to rapidly and accurately predict the mechanical properties of ALs. The results showed that the proposed model achieved a reduction of 45.22% and 50.17% in the normalized root mean square error on the testing set when compared to traditional machine learning models (Random Forest and Support Vector Machine), demonstrating its higher accuracy. In addition, the design of ALs with desired mechanical properties was achieved by optimizing the braiding parameters, and its effectiveness was verified through experiments. The mechanical properties of the prepared ALs were able to fully meet the desired targets and were at least 2% higher. Finally, the influence weights of different braiding parameters on the mechanical properties of ALs were analyzed by feature importance.