Automated multi-objective pedicle screw planning

Background: Posterior spinal fusion is a common procedure, but manual pedicle screw planning is time-consuming, cognitively demanding, and may not fully leverage patient-specific anatomy. Pedicle screws are vulnerable to loosening, a complication linked to poor anchorage. This study aimed to develop and evaluate an automated pedicle screw planning framework for multi-level fusion, designed to simultaneously optimize clinical safety, biomechanical stability, and screw head alignment. Methods: The proposed automated screw planning workflow was compared with clinical standard manual planning on 20 clinical computed tomography (CT) scans comprising 238 pedicle screws. The automated pipeline utilized neural network-based segmentation, template mesh fitting, and multi-objective optimization balancing clinical safety, biomechanical stability, and rod conformity. Clinical safety was assessed using the Gertzbein-Robbins classification system. Comparative analysis evaluated bone density coverage via Hounsfield unit (HU) measurements and screw head alignment through angular misalignment calculations. Results: Clinical safety assessment demonstrated high anatomical accuracy, with 87.8% grade A and 12.2% grade B screws per the Gertzbein-Robbins scale, with a median grade B breach of 0.8 mm. The automated method also achieved a median relative improvement of 74.9% (P