C2-Net database

The lack of an objective and ready-to-use tool for preoperative planning in C2 pedicle screw placement surgery is notable. We developed C2-Net, a deep learning model for rapidly and accurately assessing C2 pedicle screw placement feasibility from CT images. C2-Net incorporates image segmentation and screw placement probability assessment modules. Validated using 3D-printed models with manually placed screws as ground truth, C2-Net achieved 89.4% accuracy, 90.0% sensitivity, and 89.0% specificity, outperforming junior surgeons and comparable to senior surgeons. The model provides visual interpretations through attention maps, enhancing explainability. C2-Net demonstrates potential in differentiating C2 pedicle structural variations and offers a valuable assistive tool for clinical decision-making in spinal surgery planning.