AI-Assisted Detection of Supraspinatus Tendon Pathologies Using a Hierarchical Deep Learning Model to Improve Clinical Applicability: Development and Evaluation Study - Summary - MDSpire
Advertisement
AI-Assisted Detection of Supraspinatus Tendon Pathologies Using a Hierarchical Deep Learning Model to Improve Clinical Applicability: Development and Evaluation Study
To develop and evaluate a hierarchical deep learning model for identifying supraspinatus tendon disorders using MRI.
Approach:
Model Development: A hierarchical 3D deep learning model based on the ResNet-18 architecture was constructed, trained on coronal T2-weighted MRI scans.
Classification Categories: The model classifies tendon status into three categories: intact tendons, tendinopathy or partial-thickness tears, and full-thickness tears.
Interpretability Enhancement: Score-weighted Class Activation Mapping (Score-CAM) was integrated to visualize imaging features driving model predictions.
Key Findings:
The model was trained on a dataset of coronal T2-weighted MRI scans from 1192 cases.
Interpretation:
Limitations:
The study utilized a retrospective design with a dataset primarily from a single institution.
Generalizability may be limited due to the homogeneity of the imaging protocols.
