SHEESH: Spinal Hardware Evaluation for Exact Screw Hits

Abstract

Cervical spine surgery is a discipline that requires a high degree of precision, especially when it comes to the placement of pedicle and lateral mass screws. Accurate screw placement helps ensure appropriate correction and stabilization of the spine while simultaneously avoiding critical ”no-go zones” (e.g. spinal cord, vertebral artery, etc.) which, if hit by a poorly placed screw, could result in severe trauma to the patient. Currently, surgeons must rely on manual intra-operative measurements to determine relevant instrumentation data such as screw dimensions, entry points, and trajectories. However, gathering measurements this way can be time consuming, especially during complex cases, and is subject to both general human error as well as surgeon- to-surgeon variability. This thesis presents SHEESH (Spinal Hardware Evaluation for Exact Screw Hits), a novel computational pipeline that automates cervical spine pre-surgical planning. The system operates in four phases: (1) individual vertebrae are segmented from volumetric CT data using an open source computer vision model and converted to 3D surface meshes; (2) each vertebra is processed to determine the appropriate anatomical reference frame through spinal canal centroid localization; (3) a pipeline of geometric algorithms isolate clinically relevant substructures such as pedicle isthmus cross-sections and lateral mass posterior cortical faces to extract the important dimensional data (i.e. width, height, depth, and safe trajectory length); (4) screw trajectories are calculated. Ultimately, SHEESH processes the subaxial cervical spine (C3–C7) and produces pedicle and lateral mass screw recommendations whose dimensions fall within established anatomical norms reported in morphometric literature.