Artificial Intelligence Chest X-Ray Opportunistic Screening Model for Coronary Artery Calcium Deposition: A...

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Coronary artery calcification (CAC) is a key indicator of cardiovascular risk.
• Current screening methods often require specialized imaging like computed tomography (CT).
• There is a need for accessible, opportunistic screening methods for cardiovascular risk assessment.

Purpose of the Study:

• To develop an opportunistic screening model for predicting coronary calcium burden and cardiovascular risk.
• To utilize readily available frontal chest X-rays (CXR) and patient demographics for this prediction.

Main Methods:

• A novel multitask learning framework was developed and trained on 2121 patients with paired CT scans and CXR images.
• The model used coronary artery calcification (CAC) scores as ground truth.
• Internal training data from Mayo Clinic was validated on external datasets from Emory University Healthcare and Taipei Veterans General Hospital.

Main Results:

• The model achieved moderate classification performance for CAC scores (0, 1-99, 100+) with average f1-scores ranging from 0.65 to 0.71 across datasets.
• Area under the receiver operating curves (AUC) for risk identification (0 vs. 100+ CAC) ranged from 0.71 to 0.83.
• The open-source fusion AI-CXR model demonstrated robust performance on external datasets, outperforming existing state-of-the-art models on the internal cohort.

Conclusions:

• The proposed AI-CXR model shows promise as a robust, first-pass opportunistic screening method for cardiovascular risk.
• This approach leverages routine chest X-rays for accessible cardiovascular risk assessment.
• The model's performance on diverse external datasets suggests its potential for widespread clinical application.