Toward Non-Invasive Diagnosis of Bankart Lesions with Deep Learning

TL;DR

The study tackles the diagnostic challenge of Bankart lesions, traditionally best identified via invasive MRI arthrograms, by developing deep learning models that detect lesions on both standard MRIs and MRAs. Using a Swin Transformer trained on knee MRI data and fine-tuned separately for each modality, the authors ensemble predictions from sagittal, axial, and coronal views, achieving AUCs of $0.87$ for standard MRIs and $0.90$ for MRAs on a hold-out test set. On standard MRIs, the model reached about $85.9 ext{%}$ accuracy and $83.3 ext{%}$ sensitivity, rivaling radiologist performance on MRAs, while maintaining high specificity; the MRAs results were similarly strong. The findings suggest DL can reduce reliance on invasive imaging, improve diagnostic access, and potentially lower costs, though external validation and model interpretability are needed before clinical adoption.

Abstract

Bankart lesions, or anterior-inferior glenoid labral tears, are diagnostically challenging on standard MRIs due to their subtle imaging features-often necessitating invasive MRI arthrograms (MRAs). This study develops deep learning (DL) models to detect Bankart lesions on both standard MRIs and MRAs, aiming to improve diagnostic accuracy and reduce reliance on MRAs. We curated a dataset of 586 shoulder MRIs (335 standard, 251 MRAs) from 558 patients who underwent arthroscopy. Ground truth labels were derived from intraoperative findings, the gold standard for Bankart lesion diagnosis. Separate DL models for MRAs and standard MRIs were trained using the Swin Transformer architecture, pre-trained on a public knee MRI dataset. Predictions from sagittal, axial, and coronal views were ensembled to optimize performance. The models were evaluated on a 20% hold-out test set (117 MRIs: 46 MRAs, 71 standard MRIs). Bankart lesions were identified in 31.9% of MRAs and 8.6% of standard MRIs. The models achieved AUCs of 0.87 (86% accuracy, 83% sensitivity, 86% specificity) and 0.90 (85% accuracy, 82% sensitivity, 86% specificity) on standard MRIs and MRAs, respectively. These results match or surpass radiologist performance on our dataset and reported literature metrics. Notably, our model's performance on non-invasive standard MRIs matched or surpassed the radiologists interpreting MRAs. This study demonstrates the feasibility of using DL to address the diagnostic challenges posed by subtle pathologies like Bankart lesions. Our models demonstrate potential to improve diagnostic confidence, reduce reliance on invasive imaging, and enhance accessibility to care.

Figures (4)

• Figure 1: Bankart lesion on standard MRI (left) and MRI arthrogram (right) in the axial view. Images are from the same patient and depict the same tear. White circles reflect annotations identifying the tear, provided by a shoulder/elbow fellowship-trained orthopedic surgeon.
• Figure 2: Data Collection and Labeling Protocol.
• Figure 3: Model Training & Inference (a) Schematic of 2D model training setup using 3D MRIs. (b) Schematic of model inference.
• Figure 4: Receiver operating characteristic (ROC) curves for single-view models and the multi-view ensemble, compared to radiologist performance on (a) standard MRIs and (b) MRAs. The single-view models correspond to those included in the multi-view ensemble. Shaded regions around each curve represent 95% confidence intervals, calculated through bootstrapping with 1000 iterations. Radiologist performance is marked with red x symbols, illustrating sensitivity and false positive rates derived from original radiology reports. The dashed diagonal line indicates the performance of a random classifier (AUC = 0.50).