New Era Begins in Artificial Intelligence-Powered Heart Disease Screening
Researchers have significantly strengthened the detection and evaluation of cardiovascular diseases by developing the CardioNets model, which combines ECGs with cardiac MRI images.
The deep learning model CardioNets, developed by Ding and his team, has outperformed standard methods in detecting cardiovascular diseases by learning the relationships between ECG data and cardiac MRI images.
CardioNets Model and Working Principle
Ding and his colleagues developed a deep learning model called CardioNets that learns the relationships between electrocardiograms and paired cardiac magnetic resonance images. This innovative model demonstrated a performance far superior to traditional systems that only analyze ECG data.
Cost and Function Comparison of Methods
While cardiac MRI imaging provides a comprehensive assessment of heart structure and function, its high cost prevents its widespread use. In contrast, an ECG is low-cost and widely used, but it falls short in evaluating heart structure. CardioNets overcomes this deficiency by using only ECG input.
Comprehensive Research and Dataset
The researchers evaluated measurement prediction, cardiovascular disease detection, image synthesis, and physician performance in detail. The study utilized 159,819 samples provided by the UK Biobank, MIMIC-IV-ECG, and two Chinese hospitals for the training and validation of the model.
Success Rates and Performance Criteria
CardioNets improved the prediction of 82 cardiac measurements compared to the strongest ECG-based benchmark model. The area under the receiver operating characteristic curve for cardiomyopathy detection was 0.890, while this value was 0.879 for pulmonary hypertension in the MIMIC database.