2025 ISAKOS Congress in Munich, Germany

2025 ISAKOS Biennial Congress Paper


Artificial Intelligence Based MRI-Based Volumetric Assessment of Rotator Cuff Musculature Demonstrates Predictive Value for Preoperative and Postoperative Functional Outcomes Following Arthroscopic Rotator Cuff Repair

Grant J Dornan, MS, Vail, CO UNITED STATES
Marco-Christopher Rupp, MD, Munich, Bavaria GERMANY
Lauren Watkins, PhD, Stanford UNITED STATES
Maximilian Hinz, MD, Munich GERMANY
Kacie Sorfleet, B.S., Vail, CO UNITED STATES
Marilee P. Horan, MPH, Vail, CO UNITED STATES
Scott Tashman, PhD, Vail, CO UNITED STATES
Peter J. Millett, MD, MSc, Vail, CO UNITED STATES

Steadman Philippon Research Institute, Vail, CO, UNITED STATES

FDA Status Not Applicable

Summary

This study investigated an artificial intelligence-based MRI volumetric assessment of RC musculature, demonstrating improved association with patient-reported outcomes following ARCR compared to traditional manual classifications of muscle health.

Abstract

Background

The success of arthroscopic rotator cuff repair (ARCR) is significantly influenced by the health status of the musculature. Traditional classifications of muscle health, such as Goutallier and Thomazeau grading, are subjective and categorical, often leading to unreliable assessments. The use of MRI-based volumetric analysis offers a potentially more accurate and predictive approach to evaluating the rotator cuff (RC) musculature. This study aimed to evaluate the predictive value of artificial intelligence-based MRI volumetric assessments of RC musculature for functional outcomes following ARCR.

Methods

This retrospective study included 103 patients (mean age 59 ± 7 years, 33% female) with full-thickness rotator cuff tears (RCT) of the supraspinatus tendon, with or without extension into other tendons, who underwent ARCR. Preoperative and postoperative patient-reported outcomes (PROs) were collected, including the American Shoulder and Elbow Surgeons (ASES) score, Disabilities of the Arm, Shoulder, and Hand (Quick-DASH) score, Single Assessment Numeric Evaluation (SANE), and Short Form-12 Physical Component Summary (SF-12 PCS). MRI assessments included manual Goutallier grading, Thomazeau grading, critical shoulder angle (CSA), mediolateral tear size, and automated volumetric scores (muscle size, fat infiltration, and relative volume contribution) using AI-based 3D segmentation.

Results

Of the 103 patients included, 97 completed the minimum 2-year follow-up. Significant improvements were observed in the ASES score (from 57 ± 19 to 93 ± 13, p<0.001), Quick-DASH (from 37 ± 19 to 8 ± 11, p<0.001), SANE (from 57 ± 24 to 88 ± 20, p=0.001), and SF-12 PCS (from 43 ± 9 to 53 ± 8, p<0.001). Manual Goutallier grades correlated significantly with automated volumetric scores in the supraspinatus (p<0.001) and infraspinatus muscles (p<0.001), but not in the subscapularis (p=0.63). Muscle size (MS) of the infraspinatus (ISP) showed significant positive correlations with SF-12 PCS (rho=0.243, p=0.014) and was significantly associated with attainment of patient acceptable symptom state (PASS) for ASES (PASS group -0.61±1.04; non-PASS group -1.59±1.13; p=0.013). Fat infiltration (FI) of the ISP showed significant negative correlations with SF-12 PCS (rho=-0.2, p=0.044). MS of the supraspinatus (SSP) showed significant positive correlations with SF-12 PCS (rho=0.246, p=0.013) and was significantly associated with PASS attainment for ASES (PASS group -1.06±1.08; non-PASS group -1.86±1.04; p=0.025). None of the manual measurements of fatty infiltration and atrophy demonstrated a significant association with postoperative outcomes (p>0.05).

Conclusion

AI-based MRI volumetric assessment of RC musculature correlated well with manual measurements and provided improved predictive value for both preoperative and postoperative functional outcomes following ARCR. This approach may inform preoperative risk stratification and decision-making in ARCR.