A Magnetic Resonance Imaging-Based Clinical Prediction Model Accurately Identifies Patellar Instability Risk Using...

Area of Science:

• Orthopedic imaging analysis
• Machine learning in medicine
• Biomechanical assessment of knee joints

Background:

• Patellar instability is a common orthopedic condition.
• Accurate diagnosis relies on identifying specific radiographic risk factors.
• Magnetic resonance imaging (MRI) offers detailed anatomical visualization.

Purpose of the Study:

• To develop a machine learning model for predicting patellar instability using MRI measurements.
• To identify and quantify the importance of radiographic risk factors associated with patellar instability.

Main Methods:

• Retrospective analysis of knee MRI scans from patients with and without patellar instability.
• Measurement of parameters related to patella alta, malalignment, and trochlear dysplasia.
• Application of logistic regression and machine learning models (e.g., random forest) with SHAP for variable importance analysis.

Main Results:

• Multivariable logistic regression identified lower patellotrochlear index, greater Insall-Salvati ratio, greater tibial tubercle-trochlear groove (TT-TG) distance, and lower trochlear depth as significant predictors.
• The random forest model achieved an area under the receiver operating characteristic curve of 0.85.
• Insall-Salvati ratio, TT-TG distance, and trochlear depth were the most important variables identified by the machine learning model.

Conclusions:

• A machine learning model can reliably predict MRI-based parameters associated with patellar instability.
• Key risk factors like Insall-Salvati ratio, TT-TG distance, and trochlear depth are crucial for both statistical and machine learning models.
• The developed model shows potential for improving diagnostic accuracy and personalizing patient care.