Generating a Synthetic Lumbar CT from a Standard MRI Protocol

Background: Eliminating the need for a preoperative spine CT scan prior to spine surgery offers significant financial advantages, reduced radiation exposure, and a more streamlined workflow. This can be achieved by converting a standard-protocol MRI scan into a synthetic CT (sCT), which provides a precise representation of bone structures. The sCT is intended for use in the preoperative planning and navigation of spine surgeries, eliminating the need for an additional CT scan. Methods: The transformation is based on a neural network architecture that converts MRI data into high-resolution sCT images. Although the resolution of standard magnetic resonance imaging ranges from 3 to 5 mm, sCT achieves sub-millimeter accuracy (below 1 mm). In this study, we present the results of the generation of sCT and compare them with conventional CT scans of the same patients. Results: A comprehensive comparison was conducted using both 3D and 2D measurements on 500 different vertebrae. The 3D evaluation used 3D surface distance measurement to assess the full vertebrae and key anatomical elements. The 2D analysis focused on critical distances that define the vertebral body, pedicles, and spinous processes. All measurements were performed automatically and validated by orthopedic surgeons, ensuring clinical relevance. Conclusions: This study shows that converting a standard-protocol MRI scan into an sCT provides a precise representation of bone structures.