Real-Time Prediction of Lumbar Spine Stress and Displacement Using a Proof-of-Concept Digital Twin Biomechanical...

To develop a proof-of-concept digital twin framework for real-time prediction and visualization of lumbar spine biomechanics in both native and surgically treated models and to numerically verify its fidelity against coupled MBD-FEA simulations.

A hybrid framework combining multibody dynamics (MBD) and finite element (FE) simulations with two sequential Kriging surrogate models was developed. Surrogate Model 1 predicted spinal loads from three-dimensional lumbar rotations; Surrogate Model 2 estimated von Mises stress and displacement from these loads for native and postoperative (TLIF + pedicle screw) spine models. Evaluation consisted of numerical verification against full-order MBD-FEA simulations and a plausibility check by benchmarking against literature-reported L4-L5 intradiscal pressure (IDP) ranges.

For predefined compound 3-DOF (flexion/extension, lateral bending, and axial rotation) lumbar rotational motions, the mean absolute errors (MAEs) of nodal stress and displacement predicted by the digital twin relative to the coupled high-fidelity MBD-FEA outputs were ≤ 0.5% for the native spine and ≤ 0.2% for the postoperative spine. The L4-L5 IDP ranged from 108 to 513 kPa across all motions, indicating physiological levels overlapping with previously reported ranges (30-1530 kPa). However, the predicted values were lower than some reported maxima in flexion (approximately 1500 kPa).

This framework demonstrated strong predictive performance in terms of numerical agreement, indicating how faithfully the surrogate model reproduces the high-fidelity coupled MBD-FEA reference solutions, and presented a proof-of-concept pipeline capable of real-time execution and visualization. However, experimental and clinical validation of the accuracy of IMU-derived segment kinematics and in vivo biomechanical outputs should be performed in future work.