MRI Radiomics-Based Diagnosis of Knee Meniscal Injury.

Abstract

Objective: This study aims to explore a grading diagnostic method for the binary classification of meniscal tears based on magnetic resonance imaging radiomics. We hypothesize that a radiomics model can accurately grade meniscal injuries in the knee joint. By extracting T2-weighted imaging features, a radiomics model was developed to distinguish meniscal tears from nontear abnormalities.

Materials and methods: This retrospective study included imaging data from 100 patients at our institution between May 2022 and May 2024. The study subjects were patients with knee pain or functional impairment, excluding those with severe osteoarthritis, infections, meniscal cysts, or other relevant conditions. The patients were randomly allocated to the training group and test group in a 4:1 ratio. Sagittal fat-suppressed T2-weighted imaging sequences were utilized to extract radiomic features. Feature selection was performed using the minimum Redundancy Maximum Relevance (mRMR) method, and the final model was constructed using the Least Absolute Shrinkage and Selection Operator (LASSO) regression. Model performance was evaluated on both the training and test sets using receiver operating characteristic curves, sensitivity, specificity, and accuracy.

Results: The results showed that the model achieved area under the curve values of 0.95 and 0.94 on the training and test sets, respectively, indicating high accuracy in distinguishing meniscal injury from noninjury. In confusion matrix analysis, the sensitivity, specificity, and accuracy of the training set were 88%, 92%, and 87%, respectively, while the test set showed sensitivity, specificity, and accuracy of 89%, 82%, and 85%, respectively.

Conclusions: Our radiomics model demonstrates high accuracy in distinguishing meniscal tears from nontear abnormalities, providing a reliable tool for clinical decision-making. Although the model demonstrated slightly lower specificity in the test set, its overall performance was good with high diagnostic capabilities. Future research could incorporate more clinical data to optimize the model and further improve diagnostic accuracy.