Related Experiment Videos
Multi-scale Feature Learning with CNN-RNN-Attention Framework for ECG-based Cancer Therapy-Related Cardiac
Summary
A new deep learning model using ECG signals can detect cancer therapy-related cardiac dysfunction (CTRCD). This cost-effective approach offers a reliable alternative to echocardiography for monitoring heart health during cancer treatment.
Area of Science:
- Cardiology
- Oncology
- Artificial Intelligence
Background:
- Cancer therapy-related cardiac dysfunction (CTRCD) is a serious side effect of anticancer drugs.
- Echocardiography, the standard diagnostic tool, is operator-dependent, time-consuming, and expensive.
- Electrocardiogram (ECG) offers a more accessible and cost-effective alternative for cardiac monitoring.
Purpose of the Study:
- To develop a deep learning model for detecting CTRCD using ECG signals.
- To create a reliable and cost-effective method for monitoring cardiac function during cancer therapy.
- To improve the interpretability of deep learning models in cardiac diagnostics.
Main Methods:
- A hybrid deep learning model integrating Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) was developed.
- An attention mechanism was incorporated to weigh the importance of different ECG features.
- Attention weights were visualized to enhance model interpretability and identify key diagnostic features.
Main Results:
- The proposed deep learning model effectively detects CTRCD from 12-lead ECG data.
- Ablation studies confirmed the effectiveness of the integrated CNN-RNN architecture and attention mechanism.
- Visualization of attention weights identified significant ECG features contributing to CTRCD classification.
Conclusions:
- The developed deep learning model shows promise as a cost-effective and reliable tool for CTRCD detection.
- This approach can aid in the early identification of cardiac side effects in cancer patients.
- The findings support the use of ECG-based AI for routine cardiac monitoring in oncology.