Direct Measurement Of Knee Ligament Forces During Cycling Using Robotic Systems With AI-Predicted Joint Kinematics

PURPOSE: This study aimed to directly measure knee ligament forces during cycling exercise using a robot-based joint testing system with AI-predicted kinematics. METHODS: Three-dimensional (3D) motion of the femur and tibia during cycling was obtained in 24 adults using a validated method that optimally aligns CT-based bone models with two biplane fluoroscopy images. A statistical shape model (SSM) was developed using CT data from 57 cadavers to describe knee morphology with 10 shape coefficients. A multi-layer perceptron (MLP) was trained using these kinematic and morphological datasets to predict 3D joint kinematics during cycling based on joint morphology. The model's predictions were evaluated using root-mean-square errors (RMSEs). A cadaveric knee was CT scanned and input into the validated MLP to obtain joint kinematics, which were then used to drive the specimen in performing cycling exercise using a robotic system. By progressively removing the ligaments, the forces on the posterior cruciate ligament (PCL), anterior cruciate ligament (ACL), medial collateral ligament (MCL), and lateral collateral ligament (LCL) at different pedal crank angles during cycling were measured. RESULTS: The SSM showed accurate shape reconstruction with a mean RMSE of 0.7 mm. The MLP showed high prediction accuracy for joint kinematics, with mean RMSEs for translational and rotational movements below 0.4° and 0.1 mm, respectively. The forces on the PCL showed a higher mean value compared to those of the other ligaments, with a peak value of 550 N at a crank angle of 250° (Fig. 1). CONCLUSION: This study is the first to predict personalized joint kinematics during cycling from joint morphology to drive cadaveric knee joints via a robotic system for the direct measurement of ligament forces. The proposed method demonstrated high accuracy in predicting personalized joint kinematics and showed strong potential for measuring ligament forces during various functional exercises.