Temporal Relationship-Aware Treadmill Exercise Test Analysis Network for Coronary Artery Disease Diagnosis.
Jianze WeiBocheng PanYu GanXuedi LiDeping LiuBotao SangXingyu GaoPublished in: Sensors (Basel, Switzerland) (2024)
The treadmill exercise test (TET) serves as a non-invasive method for the diagnosis of coronary artery disease (CAD). Despite its widespread use, TET reports are susceptible to external influences, heightening the risk of misdiagnosis and underdiagnosis. In this paper, we propose a novel automatic CAD diagnosis approach. The proposed approach introduces a customized preprocessing method to obtain clear electrocardiograms (ECGs) from individual TET reports. Additionally, it presents TETDiaNet, a novel neural network designed to explore the temporal relationships within TET ECGs. Central to TETDiaNet is the TETDia block, which mimics clinicians' diagnostic processes to extract essential diagnostic information. This block encompasses an intra-state contextual learning module and an inter-state contextual learning module, modeling the temporal relationships within a single state and between states, respectively. These two modules help the TETDia block to capture effective diagnosis information by exploring the temporal relationships within TET ECGs. Furthermore, we establish a new TET dataset named TET4CAD for CAD diagnosis. It contains simplified TET reports for 192 CAD patients and 224 non-CAD patients, and each patient undergoes coronary angiography for labeling. Experimental results on TET4CAD underscore the superior performance of the proposed approach, highlighting the discriminative value of the temporal relationships within TET ECGs for CAD diagnosis.
Keyphrases
- coronary artery disease
- percutaneous coronary intervention
- end stage renal disease
- cardiovascular events
- coronary artery bypass grafting
- ejection fraction
- chronic kidney disease
- prognostic factors
- neural network
- high intensity
- peritoneal dialysis
- aortic stenosis
- physical activity
- emergency department
- machine learning
- cardiovascular disease
- healthcare
- oxidative stress
- type diabetes
- heart failure
- deep learning
- body composition
- left ventricular
- patient reported