Federated Learning for Enhanced ECG Signal Classification with Privacy Awareness

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Accurate classification of electrocardiogram (ECG) signals is crucial for diagnosing cardiac conditions.
• Existing methods often face challenges with data privacy and centralized data requirements.
• Federated learning offers a decentralized approach to model training, preserving patient data confidentiality.

Purpose of the Study:

• To develop and evaluate a privacy-preserving method for ECG signal classification.
• To investigate the efficacy of stacked convolutional neural networks (CNNs) for extracting features from ECG data.
• To leverage federated learning for collaborative model training without sharing raw patient data.

Main Methods:

• A novel approach combining federated learning with stacked CNNs was proposed for ECG classification.
• Stacked CNN architecture was designed to capture multi-scale temporal features in ECG signals.
• The model was trained collaboratively across distributed local devices using federated learning.

Main Results:

• The proposed federated learning and stacked CNN model achieved a final accuracy of 98.6% after 100 communication rounds.
• Performance significantly surpassed baseline methods in ECG signal classification.
• The approach demonstrated effective feature extraction and high classification accuracy.

Conclusions:

• The developed method offers an accurate and privacy-preserving solution for ECG classification in healthcare.
• Federated learning combined with stacked CNNs is a viable strategy for distributed medical data analysis.
• This approach has the potential to enhance diagnostic capabilities and patient monitoring systems.