Related Experiment Videos
MAF-Net: Multimodal cross-attention-based fusion network for cardiovascular disease classification
Chang Qu1, Xin Zhang1, Yansong Lu1
1School of Artificial Intelligence, Changchun University of Science and Technology, Changchun, Jilin, China.
Insights
This study introduces MAF-Net, a novel multimodal deep learning model for cardiovascular disease classification. By integrating clinical data with electrocardiogram (ECG) features, MAF-Net significantly improves diagnostic accuracy for various arrhythmias.
Area of Science:
- Cardiology and Medical Informatics
- Artificial Intelligence in Healthcare
- Biomedical Signal Processing
Background:
- Cardiovascular diseases (CVDs) are a leading global cause of mortality, necessitating accurate and rapid diagnostic tools.
- Electrocardiograms (ECGs) are crucial for CVD detection but traditional single-modality analysis limits accuracy by ignoring clinical data interactions.
- Integrating multimodal data, including clinical information and ECG signals, is essential for enhancing CVD diagnostic performance.
Purpose of the Study:
- To develop and evaluate MAF-Net, a Multimodal Cross-Attention-based Fusion Network, for improved cardiovascular disease classification.
- To fuse clinical data features with ECG signal features using a novel bidirectional cross-attention mechanism.
- To enhance the accuracy of classifying five major arrhythmia super-categories: Normal, Myocardial Infarction, ST-T Segment Changes, Conduction Disturbance, and Hypertrophy.
Main Methods:
- Proposed MAF-Net, a deep learning model with three components: X Branch for clinical data processing (second-order polynomial features, channel attention), Y Branch for multi-scale ECG feature extraction (convolutional modules, Bi-LSTM, multi-head attention), and a Bidirectional Modality Fusion Module.
- The fusion module employs a bidirectional cross-attention mechanism, utilizing clinical features as Query and ECG features as Key/Value for deep data integration.
- The model was evaluated on a dataset classifying five super-categories of arrhythmias.
Main Results:
- MAF-Net achieved high performance across key metrics: 90.75% ± 0.32% accuracy, 84.58% ± 0.41% precision, and 87.12% ± 0.38% recall.
- The model demonstrated a strong F1 score of 0.8069 ± 0.005 and a ROC-AUC value of 0.9407 ± 0.002.
- Experimental results indicate that MAF-Net outperforms existing methods in cardiovascular disease classification.
Conclusions:
- The proposed MAF-Net effectively integrates clinical and ECG data, significantly improving cardiovascular disease classification accuracy.
- The bidirectional cross-attention fusion mechanism is key to capturing complex inter-modal interactions.
- MAF-Net shows significant potential for clinical decision support systems in cardiology.
Abstract:
Cardiovascular disease ranks among the leading causes of death globally, posing a severe threat to human health. Consequently, rapid and accurate identification of cardiovascular disease has become a critical research endeavor. Electrocardiograms (ECGs), as a non-invasive detection tool, are widely used in cardiovascular disease detection due to their convenience and effectiveness. However, existing methods are often limited to single-modality analysis, neglecting the interaction between clinical data (such as age, gender, weight, etc.) and ECG features in classification tasks, resulting in limited recognition accuracy. Integrating multimodal data is key to improving CVD diagnostic accuracy. To address this, we propose MAF-Net (Multimodal Cross-Attention-based Fusion Network), a multi-class classification model that fuses clinical data features with ECG signal features for cardiovascular disease classification. The model comprises three components: (1) X Branch (Clinical Data Processing): Generates high-order interaction features via a second-order polynomial feature cross-layer and employs channel attention-weighted selection to identify key clinical factors;(2) Y Branch (Multi-scale ECG Feature Extraction): Parallel multi-scale convolutional modules (64@7, 128@3, 256@3) capture local morphological features, while Bi-LSTM models long-range temporal dependencies, supplemented by multi-head attention to focus on pathological segments;(3) Bidirectional Modality Fusion Module: Employing a bidirectional cross-attention mechanism, it uses clinical features as Query and ECG features as Key/Value to deeply fuse clinical and ECG data features. On the dataset, experiments targeting five super-categories of arrhythmias- NORM (Normal), MI (Myocardial Infarction), STTC (ST-T Segment Changes), CD (Conduction Disturbance), HYP (Hypertrophy). showed an accuracy rate of 90.75% ± 0.32%, precision of 84.58% ± 0.41%, and recall of 87.12% ± 0.38%, with an F1 score of 0.8069 ± 0.005 and a ROC-AUC value of 0.9407 ± 0.002. Results indicate that this model outperforms existing methods across key metrics, demonstrating its potential for application in clinical decision support.
Related Concept Videos
01:21Imaging Studies for Cardiovascular System IV: CMRI
01:28Imaging Studies for Cardiovascular System V: CT
01:25Cardiomyopathy I: Introduction and Classification