SCOPE-MRI: Bankart Lesion Detection as a Case Study in Data Curation and Deep Learning for Challenging Diagnoses

2025-04-29

AI Summary

Bankart lesions exhibit subtle and often inconspicuous features on standard MRI, leading to diagnostic challenges. Method: We introduce ScopeMRI—the first publicly available shoulder MRI dataset validated by intraoperative arthroscopic examination and expert annotation—and propose a dual-path deep learning framework that separately models standard MRI and MR arthrography (MRA) modalities. The framework integrates multi-planar 3D convolutional neural networks with vision transformers to enable modality-specific feature learning and cross-plane feature fusion. Contribution/Results: Our method achieves an AUC of 0.91, sensitivity of 83%, and specificity of 91% on standard MRI alone—matching or surpassing radiologists’ diagnostic performance on MRA; the MRA submodel attains an AUC of 0.93. External validation confirms robust generalizability. We release both the dataset and modular, reproducible code to advance clinical translation and open research in musculoskeletal radiology AI.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

While deep learning has shown strong performance in musculoskeletal imaging, existing work has largely focused on pathologies where diagnosis is not a clinical challenge, leaving more difficult problems underexplored, such as detecting Bankart lesions (anterior-inferior glenoid labral tears) on standard MRIs. Diagnosing these lesions is challenging due to their subtle imaging features, often leading to reliance on invasive MRI arthrograms (MRAs). This study introduces ScopeMRI, the first publicly available, expert-annotated dataset for shoulder pathologies, and presents a deep learning (DL) framework for detecting Bankart lesions on both standard MRIs and MRAs. ScopeMRI includes 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 diagnosis. Separate DL models for MRAs and standard MRIs were trained using a combination of CNNs and transformers. 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). The models achieved an AUC of 0.91 and 0.93, sensitivity of 83% and 94%, and specificity of 91% and 86% for standard MRIs and MRAs, respectively. Notably, model performance on non-invasive standard MRIs matched or surpassed radiologists interpreting MRAs. External validation demonstrated initial generalizability across imaging protocols. This study demonstrates that DL models can achieve radiologist-level diagnostic performance on standard MRIs, reducing the need for invasive MRAs. By releasing ScopeMRI and a modular codebase for training and evaluating deep learning models on 3D medical imaging data, we aim to accelerate research in musculoskeletal imaging and support the development of new datasets for clinically challenging diagnostic tasks.

Problem

• Detecting Bankart lesions on standard MRIs using deep learning
• Reducing reliance on invasive MRAs for challenging diagnoses
• Creating a public dataset for musculoskeletal imaging research

Innovation

• First public expert-annotated dataset for shoulder pathologies
• Deep learning combines CNNs and transformers for lesion detection
• Multi-view ensemble improves MRI and MRA diagnostic accuracy