Labial-gland artificial intelligence model screening for autoimmune thyroiditis among patients with connective tissue...

Area of Science:

• Pathology
• Artificial Intelligence
• Immunology

Background:

• Connective tissue disease (CTD) patients have an increased risk of autoimmune thyroiditis (AIT).
• Early identification of AIT risk in CTD patients is crucial for timely intervention.
• Current diagnostic methods may not fully capture subtle pathological indicators.

Purpose of the Study:

• To develop a deep learning-based prediction model for assessing AIT risk in CTD patients.
• To utilize whole section images (WSI) of labial gland pathological tissue for risk prediction.
• To enhance early clinical identification and management strategies for AIT in CTD.

Main Methods:

• Retrospective analysis of 121 CTD patients' labial gland pathological sections.
• Classification of patients into positive (Ab+) and negative (Ab-) groups based on thyroid autoantibodies (TgAb, TPOAb).
• Application of EfficientNet-B5 for feature extraction, combined with multi-instance and ensemble learning.

Main Results:

• The developed deep learning model achieved an AUC of 0.829 on both internal and external validation sets.
• The model effectively identified key pathological features in labial gland tissue associated with high AIT risk.
• Demonstrated excellent prediction performance for AIT risk in CTD patients.

Conclusions:

• Deep learning analysis of labial gland WSI is effective for predicting AIT risk in CTD patients.
• The model offers a novel technical approach for early clinical risk stratification.
• Provides a theoretical basis for optimizing diagnosis and treatment decisions in AIT-affected CTD patients.