Detection of carotid plaques on panoramic radiographs using deep learning

Area of Science:

• Artificial Intelligence in Medical Imaging
• Oral and Maxillofacial Radiology
• Cardiovascular Disease Screening

Background:

• Panoramic radiographs (PRs) incidentally detect carotid artery calcification (CAC) in 3-15% of patients.
• Dental professionals often miss CAC diagnoses due to limited training.
• Early detection of CAC is crucial for managing carotid artery atherosclerosis.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) model for detecting CAC on PRs.
• To utilize a vision transformer-based AI approach for improved diagnostic accuracy.
• To compare the AI model's performance against existing convolutional neural network (CNN) models.

Main Methods:

• Trained an AI model using Faster R-CNN and Swin Transformer on 6,404 PRs.
• Manually annotated CAC on 185 PRs with CAC and 185 PRs without CAC.
• Evaluated model performance using precision, F1-score, recall, AUC, and AP, comparing against CNN-based AI.

Main Results:

• The Faster R-CNN and Swin Transformer model achieved high performance: precision 0.895, recall 0.881, F1-score 0.888, AUC 0.950, and AP 0.942.
• The developed AI model significantly outperformed previously reported CNN-based AI approaches.
• The model demonstrated improved detection performance for CAC on PRs.

Conclusions:

• The novel AI model shows enhanced detection capabilities for CAC on PRs.
• Integrating this AI tool into dental imaging can assist professionals in identifying CAC.
• This technology has the potential to improve early detection and clinical management of carotid artery atherosclerosis.