Temporal Relationship-Aware Treadmill Exercise Test Analysis Network for Coronary Artery Disease Diagnosis

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Treadmill exercise test (TET) is a common non-invasive method for diagnosing coronary artery disease (CAD).
• TET reports are prone to external influences, leading to potential misdiagnosis or underdiagnosis of CAD.
• Accurate and automated CAD diagnosis from TET ECGs is crucial for timely patient management.

Purpose of the Study:

• To develop a novel automatic approach for diagnosing CAD using TET reports.
• To enhance the accuracy of CAD diagnosis by analyzing temporal relationships within electrocardiograms (ECGs).
• To introduce a new deep learning model, TETDiaNet, for improved CAD detection.

Main Methods:

• A customized preprocessing method was developed to obtain clear ECGs from TET reports.
• A novel neural network, TETDiaNet, was designed to capture temporal relationships within TET ECGs.
• The TETDia block within TETDiaNet mimics clinical diagnostic processes using intra-state and inter-state contextual learning modules.

Main Results:

• The proposed approach demonstrated superior performance in CAD diagnosis on the newly established TET4CAD dataset.
• Experimental results highlighted the significant discriminative value of temporal ECG relationships for CAD diagnosis.
• The TETDiaNet model effectively extracted essential diagnostic information by modeling temporal dynamics in ECGs.

Conclusions:

• The novel automatic CAD diagnosis approach using TETDiaNet shows promise for improving diagnostic accuracy.
• Analyzing temporal relationships within TET ECGs is a valuable strategy for CAD detection.
• The developed TET4CAD dataset provides a resource for further research in automated CAD diagnosis.