Predictive stress analysis in simplified spinal disc model using physics-informed neural networks

This study develops a physics-informed neural network (PINN) model to predict stress distribution in a simplified spinal disc structure. The model incorporates 3D spatial inputs and enforces equilibrium conditions through a custom loss function. Trained on synthetic elasticity-based data, it achieves an MAE of 0.026 and an R² of 74.6%. Stress patterns under various loading conditions were visualized, with peak stress occurring at z = 1 under top compression. Results demonstrate PINNs' potential for biomechanical modeling, improving predictive accuracy in spinal biomechanics and informing clinical interventions.