The 3-Dimensional Fatty Infiltration in the Overall Supraspinatus Can Be Predicted by Localized Sectional Accumulation Units

Abstract

Background: Fatty infiltration (FI) of the rotator cuff (RC) muscles is one of the most common risk factors for a retear following RC repair. Recent methodological developments focus on using 3-dimensional measurements of the overall FI of RC muscles instead of using single-plane-based measurements. However, the required labor-intensive segmentation and time-consuming post-processing steps need to be optimized for routine clinical use. Methods: We collected all 6-point Dixon magnetic resonance imaging (MRI) slices of the overall supraspinatus in 46 patients with atraumatic small-to-massive RC tears. Using emerging techniques, the overall 3-dimensional FI of the supraspinatus (overall FI, using all N slices) was assessed as the reference. Each sagittal segment of the supraspinatus was defined as a sectional accumulation unit (SAU). The localized FI in SAUs with different scales of N/3, N/6, and N/12 slices (SAU-FIs) was then calculated using piecewise accumulation by stacking neighboring slices after the overall supraspinatus had been sequentially segmented using MRI slices of the same thickness. The capacity of the SAU-FIs to predict the overall FI, and the ideal locations for prediction, were evaluated using linear regression models after the associations were examined. Goodness-of-fit of the regression models was appraised by the coefficient of determination (R2) and root-mean-square error (RMSE). The agreement between the predicted and measured overall FI was assessed using Bland-Altman analysis and the standard deviation of the percent differences (sd%). Results: The localized SAU-FIs of the N/3, N/6, and N/12 SAUs generally displayed comparable distributions throughout the normalized distal-proximal long axis of the supraspinatus. The localized SAU-FIs showed substantial correspondence with the overall FI, and the highest correlations were found in the 2/3 SAU (Pearson r and Spearman ρ: 0.95, 0.98), 3/6 and 4/6 SAUs (Pearson r and Spearman ρ: 0.97), and 5/12 to 7/12 SAUs (Pearson r and Spearman ρ: 0.95 to 0.96). The strongest predictors to estimate the overall supraspinatus FI in the regression analysis were these SAU-FIs located in the middle third, which demonstrated good fits to the overall FI (all R2 ≥ 0.90; RMSE ≤ 1.69). The best agreements between the overall FI predicted by the regression models and the measured overall FI were found in these SAUs (2/3 SAU: sd% = 4.84%; 3/6 and 4/6 SAUs: sd% = 5.14%; 5/12 to 7/12 SAUs: sd% = 6.44%). Conclusions: Specific SAUs near the center of the supraspinatus (2/3, 3/6 and 4/6, and 5/12 to 7/12 SAUs), which displayed the best agreement between the predictions and actual measurements of overall FI values, can serve as appropriate surrogates to estimate the overall FI of the supraspinatus in small-to-massive RC tears. The potential to assess the overall FI of the supraspinatus using specific localized SAUs may improve the speed of analytical strategies for accurately assessing the overall FI of RC muscles and thus enable their routine clinical use in the future. Level of Evidence: Prognostic Level IV. See Instructions for Authors for a complete description of levels of evidence.