Abstract

Reconstructing spinal defects continues to be a significant challenge in modern neurosurgery due to the complexity of their anatomy and individual differences in vertebral structures. This work embraces an innovative end-to-end pipeline in which artificial intelligence (AI), computer tomography (CT) imaging, and additive manufacturing are combined to create patient specific spinal implants made of PEEK. AI-driven segmentation and reconstruction models with cascaded 3D convolutional neural networks (CNNs) are used to automate the segmentation and reconstruction of anatomical models based on clinical CT datasets. High-fidelity anatomical models can be reconstructed for the computer-aided design (CAD) and 3D printing of custom implants tailored to each individual patient's anatomy; ultimately benefiting surgical accuracy, accuracy of implant fit, and postoperative outcomes using PEEK. PEEK has several advantages due to its radiolucency, biocompatibility, and mechanical compatibility with cortical bone. The Auto-Implant method allows for enhanced precision through curvature-sensitive modeling for ideal anatomical and cosmetic match. Evaluation on a synthetic and clinical dataset achieved sub-mm reconstruction accuracy and numerous endorsements from clinicians and appropriate anatomical accuracy of final implant designs, achieved through post-processing approaches and Gaussian curvature. We have demonstrated a reproducible and scalable AI assisted pipeline to personalize the manufacture of patient specific implants and perhaps the outcomes of surgical procedures, whilst facilitating more extensive personalized spinal reconstructions. Future directions will include continuing to expand clinical validation, enhance user interactivity with clinician input and build structured clinical trials to facilitate greater uptake and regulatory approval. This research has taken significant steps in facilitating the adoption of AI within surgical planning and implant design and has greatly advanced personalized, precision-based neurosurgical care.