Deep learning for giant cell arteritis diagnosis on temporal artery biopsy

Area of Science:

• Pathology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Giant Cell Arteritis (GCA) is a large-vessel vasculitis requiring prompt diagnosis to prevent severe complications.
• Temporal artery biopsy (TAB) is the standard diagnostic method, but its interpretation is complex and expertise-dependent.

Purpose of the Study:

• To evaluate the diagnostic accuracy of a deep learning (DL) model for GCA using TAB images.
• To assess the DL model's ability to identify key pathological features indicative of GCA.

Main Methods:

• A DL model was trained using whole-slide images (WSI) from 366 TABs (137 GCA, 229 controls).
• The model employed CTransPath for feature extraction and an attention-based multiple-instance learning mechanism.
• Performance was validated on an external cohort of 58 TABs (21 GCA, 37 controls).

Main Results:

• The DL model achieved high diagnostic accuracy, with a mean AUROC of 0.987 on the training set and 0.994 on the external test set.
• The model demonstrated interpretability, with attention scores highlighting inflammatory infiltrates in the arterial media layer.
• Only one case was misclassified in the external testing cohort.

Conclusions:

• The developed DL framework achieves state-of-the-art performance for GCA diagnosis from TAB images.
• The model's interpretability supports its potential as an assistive tool for pathologists.
• This AI approach could streamline GCA diagnosis, reduce pathologist workload, and enhance diagnostic consistency.