Automated 3-Dimensional Magnetic Resonance Imaging Allows for Accurate Evaluation of Glenoid Bone Loss Compared With 3-Dimensional Computed Tomography.

Abstract

To evaluate clinical measurements of glenoid bone loss based on 3-dimensional (3D) computed tomography (CT) and automatically segmented 3D reconstructions from Dixon fat-water magnetic resonance (MR) imaging. Available CT and MR studies from 16 patients with recurrent anterior shoulder instability were retrospectively reviewed. Three-dimensional reconstructions were formed independently by 2 observers using freely available software and a simple threshold-based segmentation (3D Slicer, version 4.8.0; http://www.slicer.org). Bone loss was estimated with the perfect-circle method. Intra-user and interuser reproducibility was determined with intraclass correlation coefficients. Bland-Altman plots were used to evaluate the similarity between imaging modalities. Differences between MR and CT estimates of bone loss ranged from 0% to 6%. The individual intraclass correlation coefficients showed good to excellent reliability, with intraobserver comparisons between MR- and CT-based bone loss estimates ranging from 0.94 to 0.99. Bland-Altman plots showed 95% confidence intervals from -5% to 6% for differences between MR and CT estimates, with 88% of all measurements (42 of 48) showing a less than 2% difference between MR and CT estimates. The described methodology for obtaining an MR-based 3D reconstruction of the glenoid can evaluate glenoid bone loss similarly to the performance of a 3D CT reconstruction. The results may allow surgeons to simplify the preoperative imaging protocol for patients with recurrent shoulder stabilization and limit the number of shoulder CT scans. Level III, retrospective therapeutic trial.