AI-Assisted Detection of Supraspinatus Tendon Pathologies Using a Hierarchical Deep Learning Model to Improve Clinical Applicability: Development and Evaluation Study

Category	Detail
Condition	Supraspinatus tendon disorders
Key Mechanisms	Hierarchical 3D deep learning model based on ResNet-18 architecture for MRI classification
Target Population	Adults, particularly those over 60 years
Care Setting	Musculoskeletal imaging

Supraspinatus tendon abnormalities are prevalent, affecting approximately 20% of the general population.
MRI is the preferred imaging modality for evaluating rotator cuff pathology.
AI can enhance the identification of supraspinatus tendon disorders through automated pattern recognition.
The developed AI model classifies tendon status into intact, tendinopathy/partial-thickness tears, and full-thickness tears.
Score-CAM is used to improve interpretability and clinician trust in AI predictions.

Conservative measures for tendinopathy and partial-thickness tears; surgical intervention for full-thickness tears.

Regular follow-up imaging may be necessary to assess progression of tendon disorders.

Inadequate interpretability of AI models may limit clinician trust and adoption.

Adults, especially those aged over 60 years with shoulder pain.

Management strategies vary based on the classification of tendon status.

Utilize MRI with coronal T2-weighted sequences for optimal visualization of supraspinatus tendon disorders.
Incorporate AI tools to assist in the diagnosis and treatment planning for rotator cuff injuries.

Clinical Scorecard: Utilizing a Hierarchical Deep Learning Approach for AI-Enhanced Identification of Supraspinatus Tendon Disorders: A Study on Development and Evaluation for Clinical Use