Automated 3D MRI Allows for Accurate Evaluation of Glenoid Bone Loss as Compared to 3D CT

Abstract

Objectives:Glenoid bone loss is frequently present in the setting of recurrent shoulder instability. The magnitude of bone loss is an important determinant of the optimal surgical treatment. The current gold-standard for measurement of glenoid bone loss is three-dimensional (3D) reconstruction of a computed tomography (CT) scan. CT scans, however, carry an inherent risk of radiation and increased cost for a second modality. Magnetic resonance imaging (MRI) offers excellent soft tissue contrast and may allow resolution of bony structures to generate 3D reconstructions without a risk of ionizing radiation. We hypothesized that automated 3D MRI reconstruction would offer similar measurements of glenoid bone loss as recorded from a 3D CT scan in a clinical setting.Methods:A retrospective review was performed for fourteen patients who had both pre-operative MRI scan and CT scan of the shoulder. All MR scans were performed on a 1.5 T scanner (Siemens) utilizing a Dixon chemical shift separation sequence and the out-of-phase images with 0.90 mm slice thickness. Reconstructions of the glenoid were performed from axial images (Figure 1A) using an open-platform image processing system (3D Slicer; slicer.org). A single point on the glenoid was selected and a standard threshold was used to build a 3D model (Figure 1B). High-resolution CT scans underwent 3D reconstruction in Slicer based on Houndsfield Unit thresholding. Glenoid bone loss on both scans was measured with the Pico method by defining a circle of best fit using the inferior 2/3 of the glenoid and determining the percent area missing from this circle. Pearson’s correlation coefficient was utilized to determine the similarity between MR and CT based measurements. Statistical significance was defined as p<0.05.Results:The correlation between 3D MR and CT-based measurements of glenoid bone loss was excellent (r = 0.95, p<0.0001). The mean bone loss as measured by the 3D MR was 13.2 +- 7.2% and was 12.5 +- 8.6% for the 3D CT reconstruction (p=0.32). Bone loss in this cohort ranged from 3.7-25.4% on 3D MR and 1.4-26.0% on 3D CT. The root-mean-square difference between measurements was 2.7%.Conclusion:There was excellent agreement between automated 3D MR and 3D CT measurements of glenoid bone loss and minimal differences between these measurements. This reconstruction method requires minimal post-processing, no manual segmentation, and is obtained with widely-available MR sequences. This method has the potential to decrease the utilization for CT scans in determining glenoid bone loss.Figure 1.(A) An axial T1-weighted Dixon subtraction MR image of the glenoid is used to develop a three-dimensional reconstruction of the glenoid. The three-dimensional reconstruction is shown for one patient for both MR scan (B) and CT scan (C), both produced in 3D Slicer with a single click on the glenoid and a threshold-based reconstruction method.