Quantitative evaluation of fatty change in rotator cuff muscles using the Otsu thresholding method

BACKGROUND: Accurate assessment of fatty change in rotator cuff muscles is critical for clinical decision-making and predicting postoperative outcomes. While the Goutallier classification is widely used for grading fatty change, it is subjective and exhibits inter-rater variability. This study aims to evaluate the reliability and validity of a quantitative method using Otsu's thresholding for measuring fatty area in magnetic resonance imaging scans, comparing its performance to the Goutallier classification. METHODS: Thirty-four patients with large to massive rotator cuff tears underwent preoperative magnetic resonance imaging. Two shoulder surgeons independently evaluated the supraspinatus, infraspinatus, teres minor, and subscapularis muscles using 2 methods: (1) Goutallier classification and (2) a quantitative analysis using ImageJ, where fatty area was segmented using Otsu's thresholding method. Inter-rater reliability was assessed using weighted kappa for Goutallier classification and intraclass correlation coefficient for the fatty area percentages. Segmentation accuracy was measured using the Intersection over Union (IoU). Additionally, the relationship between Goutallier grades and fatty area percentage was analyzed using one-way analysis of variance with post hoc Tukey test. RESULTS: The Otsu method demonstrated significantly higher inter-rater reliability across all muscles (intraclass correlation coefficient >0.90), compared to the moderate agreement observed with Goutallier classification (weighted kappa: 0.64-0.79). The IoU values ranged from 0.70 to 0.85, indicating robust segmentation accuracy, particularly for the supraspinatus and subscapularis muscles. There was a strong correlation between Goutallier grades and fatty area percentage, with all grades showing significant differences (P < .05). The correlation between IoU and relative error in fatty area measurement was weak, suggesting that segmentation accuracy did not heavily influence the precision of fatty area quantification. CONCLUSION: Otsu's thresholding method offers a reliable, objective, and efficient approach for quantifying fatty change in rotator cuff muscles, outperforming the Goutallier classification in terms of inter-rater reliability. This method provides a more standardized measurement and could supplement or replace subjective grading systems in clinical practice. Future research should focus on integrating this method into clinical workflows and exploring its predictive value for surgical outcomes.