Deep convolution neural network for screening carotid calcification in dental panoramic radiographs

Area of Science:

• Cardiology
• Radiology
• Artificial Intelligence

Background:

• Ischemic stroke is a major cause of death and disability, often linked to carotid artery atherosclerosis.
• Carotid artery calcification (CAC) is a key indicator of atherosclerosis, traditionally detected via ultrasound.
• Recent advancements allow inference of CAC from panoramic dental radiographs.

Purpose of the Study:

• To develop and evaluate an AI-based algorithm for automatic detection of carotid calcifications.
• To utilize deep learning on panoramic dental radiographs for CAC identification.

Main Methods:

• A deep learning convolutional neural network (CNN) with transfer learning (TL) was developed.
• The algorithm was trained on 500 panoramic dental radiographs, with manual labeling of calcifications.
• Performance was evaluated using sensitivity (recall) and specificity metrics.

Main Results:

• The AI algorithm achieved a sensitivity of 0.82 and specificity of 0.97 for individual arteries.
• For individual patients, the algorithm demonstrated a recall of 0.87 and specificity of 0.97.
• The system accurately identified calcifications from routine dental X-rays.

Conclusions:

• AI-based detection of carotid artery calcifications from dental radiographs is feasible and accurate.
• Integration into healthcare and dental settings could aid in reducing stroke incidence and associated mortality.
• This approach offers a novel, non-invasive method for cardiovascular risk assessment.