Glenoid Version and Inclination Can be Accurately Predicted for Shoulder Arthroplasty from Preoperative Computed Tomography Scans Utilizing Virtual Implant PositioningTM Despite Missing Angulus Inferior Data Using Statistical Shape Modeling.

Abstract

Background: Computed tomography (CT)-based preoperative planning has become increasingly popular for shoulder arthroplasty. Scans which omit the angulus inferior of the scapula are common and can result in a rejection, but the effect of small amounts of missing scapula on the calculation of native version and inclination are unclear. The goals of this study were: (1) to determine the effect of increasing amounts of missing angulus inferior data on computed version and inclination and (2) to compare two methods of registering the scapula and calculating native version and inclination with increasing amounts of missing angulus inferior data.

Methods: 100 shoulder CT scans uploaded for clinical use were selected from a large database to encompass a broad spread of glenoid pathologies for the study. The mean retroversion was -9.7 degrees (range -37.8 to 5.5 degrees) and mean inclination 9.0 degrees (range -5.7 to 30.0 degrees). Each CT scan was first processed as it would be clinically within the Virtual Implant Positioning preoperative planning software. Each scapula then successively had 10mm, 20mm and 30mm of angulus inferior removed from the CT scan. The new version and inclination were then assessed using two methods: (1) the best fit scapular plane with an assumed point for the angulus inferior and (2) using a statistical shape model (SSM). Mean differences in calculated version and inclination were then compared to the true native version and inclination, and between the best fit plane method and SSM method for each level of inferior scapula cropping.

Results: For the best fit plane method, no statistically significant mean differences were noted between 0 and 10 or 20mm of angulus inferior cropping, however 30mm of cropping resulted in statistically significant different mean differences in version (p < 0.001) and inclination (p = 0.040). For the SSM, there were no significant differences in the mean differences in version or inclination for any of the cropping levels. For all comparisons, the SSM statistically out-performed the best fit plane method.

Conclusions: Less than 20 mm of missing angulus inferior data does not have a statistically significant effect on calculations of native version and inclination for preoperative CT-based planning utilizing Virtual Implant Positioning for shoulder arthroplasty. For patients with missing angulus inferior data, use of an SSM for scapula registration results in statistically significantly less mean and maximum differences from the true version and inclination compared to a 3D least squares best fit plane method.