Validation of an automated spinal fusion finite element modeling framework for the prediction of pedicle screw loosening

This study aimed to validate a patient-specific biomechanical simulation pipeline for predicting pedicle screw loosening risk in posterior spinal fusion. In particular, the research questioned whether this simulation, which integrates CT-derived bone properties, musculoskeletal force data, and finite element analysis, can outperform conventional CT attenuation measurements taken in both the vertebral body and along the planned screw trajectory.

We conducted a retrospective database analysis, including patients who underwent lumbar spinal fusion with preoperative and postoperative CT scans. Screw loosening was identified through manual testing during revision surgeries. CT attenuation was measured manually in the vertebral body and automatically along the screw trajectory. The biomechanical model integrated patient-specific musculoskeletal force data, CT-derived bone properties, and finite element analysis to estimate local bone loading relative to yield stress. The pipeline’s predictive performance was evaluated using receiver operator characteristic (ROC) curves.

The study included 161 pedicle screws, with 48 classified as loosened. Patient-specific biomechanical modeling demonstrated superior predictive capabilities (ROC AUC = 0.919) compared to screw trajectory HU measurements (ROC AUC = 0.783) and vertebral body HU measurements (ROC AUC = 0.760). Patient-specific biomechanical modeling offers a more comprehensive assessment of screw loosening risk by integrating multiple influential factors compared to simple CT attenuation measurements.