Key result
MultiCardioFusionNet outperforms other deep learning models for CHD diagnosis with ~0.93 external AUC.
Its licence doesn't allow us to show a copy here.
Why the study?
Accurate, accessible, and cost-effective diagnostic approaches for coronary heart disease remain needed to improve morbidity and mortality outcomes.
Does MultiCardioFusionNet improve the diagnostic accuracy of coronary heart disease compared to other deep learning models in clinically suspected patients?
Does MultiCardioFusionNet improve the diagnostic accuracy of coronary heart disease compared to other deep learning models in clinically suspected patients?
A multimodal deep learning framework integrating ECG and basic clinical data provides highly accurate, non-invasive identification and risk assessment of coronary heart disease.
Coronary heart disease (CHD) remains a leading cause of morbidity and mortality worldwide, highlighting the need for accurate, accessible, and cost-effective diagnostic approaches. This study proposes MultiCardioFusionNet, a multimodal deep learning framework that integrates electrocardiographic (ECG) signals and clinical information (gender, age, and body mass index) for intelligent CHD diagnosis. To capture complementary disease-related characteristics, a three-branch architecture is designed to jointly extract Frequency-Domain features, temporal features, and clinical representations from multimodal data. The resulting multimodal representations are integrated through an attention-guided fusion strategy to exploit complementary information across spectral, temporal, and clinical domains. The framework was developed and evaluated using real-world multicenter clinical data. In the internal validation cohort, MultiCardioFusionNet achieved an AUC of 0.9327, while independent external validation yielded an AUC of 0.9261, demonstrating promising generalizability capability. The proposed model consistently outperformed several representative deep learning methods, including Gated Recurrent Unit, Convolutional Neural Network, Convolutional Neural Network - Long Short-Term Memory, Convolutional Neural Network - Bidirectional Long Short-Term Memory and Convolutional Neural Network - Transformer (GRU, CNN, CNN-LSTM, CNN-BILSTM, and CNN-transformer). These results indicate that integrating electrophysiological and clinical information can substantially enhance CHD diagnosis. The consistent performance across the internal test cohort and the independent multi-center external validation cohort suggests that MultiCardioFusionNet may serve as a non-invasive decision-support approach for assisting CHD identification and risk assessment among clinically suspected patients undergoing further diagnostic evaluation.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2026) studied Coronary heart disease (CHD). MultiCardioFusionNet vs. Other deep learning methods (GRU, CNN, CNN-LSTM, CNN-BILSTM, and CNN-transformer) was evaluated on Area under the curve (AUC) for CHD diagnosis. MultiCardioFusionNet achieved an AUC of 0.9327 in internal validation and 0.9261 in multicenter external validation for CHD diagnosis, outperforming other representative deep learning methods.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: