A Machine Learning Model Demonstrates Excellent Performance in Predicting Subscapularis Tears Based on Pre-Operative
Jacob F Oeding1, Ayoosh Pareek2, Micah J Nieboer3
1School of Medicine, Mayo Clinic Alix School of Medicine, Rochester, Minnesota, U.S.A.; Oslo Sports Trauma Research Center, Norwegian School of Sport Sciences, Oslo, Norway.
Summary
Machine learning accurately predicts subscapularis tears using only preoperative magnetic resonance imaging (MRI) findings. This model achieved 85% accuracy, highlighting MRI
Area of Science:
- Orthopedic surgery
- Medical imaging
- Machine learning applications in healthcare
Background:
- Subscapularis tears are common rotator cuff injuries.
- Accurate preoperative diagnosis is crucial for effective treatment planning.
- Current diagnostic methods can be limited.
Purpose of the Study:
- To develop a machine learning model for predicting subscapularis tears.
- Utilize preoperative imaging and physical examination data.
- Assess the model's predictive performance.
Main Methods:
- A cohort of 202 shoulders undergoing rotator cuff repair was analyzed.
- Demographic, physical examination, and imaging data were collected.
- An XGBoost machine learning model was developed and validated.
Main Results:
- The XGBoost model achieved high accuracy (0.85) in predicting subscapularis tears using MRI data alone.
- Key predictive features included MRI signs of tearing, MRI quality, and biceps pathology.
- Adding physical examination or patient characteristics did not significantly improve predictive ability.
Conclusions:
- Machine learning models can accurately predict subscapularis tears based on preoperative MRI.
- Focusing on key imaging features enhances predictive performance.
- Preoperative MRI is a powerful tool for identifying subscapularis tears.
