Reliability of MR Quantification of Rotator Cuff Muscle Fatty Degeneration Using a 2-point Dixon Technique in Comparison with the Goutallier Classification: Validation Study by Multiple Readers

Rationale and Objectives: Presurgical assessment of fatty degeneration is important in the management of patients with rotator cuff tears. The Goutallier classification is widely accepted as a qualitative scoring system, although it is highly observer-dependent and has poor reproducibility. The objective of this study was to quantify fatty degeneration of the supraspinatus muscle using a 2-point Dixon technique in patients with rotator cuff tears by multiple readers, and to evaluate the reproducibility compared to Goutallier classification. Materials and Methods: Two hundred patients with rotator cuff tears who underwent magnetic resonance imaging (MRI), including 2-point Dixon sequence at 3.0-T, were selected retrospectively, Qualitative and quantitative analyses of fatty degeneration were performed by two radiologists and three orthopedic surgeons independently. The fat quantification was performed by measuring signal intensity values of in phase (S(In)) and fat image (S(Fat)), and calculating fat fraction as S(Fat)/S(In). The reproducibility of MR quantification was analyzed by the intra- and interclass correlation coefficients and Bland-Altman plots. Results: The interobserver agreement of the Goutallier classification among five readers was moderate (k = 0.51), whereas the interclass correlation coefficient regarding fat fraction value quantified in 2-point Dixon sequence was excellent (0.893). The mean differences in fat fraction values from the individual segmentation results were from -0.072 to 0.081. Proposed fat fraction grading and Goutallier grading showed similar frequency and distribution in severity of rotator cuff tears. Conclusions: Fat quantification in the rotator cuff muscles using a 2-point Dixon technique at 3.0-T MRI is highly reproducible and clinically feasible in comparison to the qualitative evaluation using Goutallier classification.