PEACE: Private and Energy-Efficient Algorithm for Cardiac Evaluation on the EDGE using Modified Split Learning and Model Quantization

International Conference on Information and Communication Systems (ICICS), pp. 01–06

Abstract

The classification of Electrocardiograms (ECG) signals is crucial for diagnosing and monitoring cardiovascular diseases. Recently, there has been increasing interest in leveraging efficient machine learning models for ECG classification. Optimizing computational resources in this domain is crucial for real-time analysis and reducing power consumption. In this paper, we introduce an efficient and personal ECG classification system that combines privacy-preserving distributed machine learning with model quantization to decrease computational complexity. The system includes an efficient modified split learning mechanism that saves 71.7% of the communications overhead and 46.5% of the computational complexity at the client compared to the original split learning system. Furthermore, we employed post-training quantization to enhance the efficiency of the client model, resulting in a remarkable 50x reduction in dynamic power consumption. Overall, our proposed system can enable the use of ECG classifiers on portable medical devices much more efficiently while preserving the privacy of patient data.

Authors 5

1. Ahmad Ayad
   Affiliation as printed

   Chair of Information Theory and Data Analytics

2. Mahdi Barhoush
   Affiliation as printed

   Chair of Information Theory and Data Analytics

3. Johnson Loh Aachen Chair of Integrated Digital Systems and Circuit Design (IDS)

   RWTH Aachen University

   Affiliation as printed

   Chair of Integrated Digital Systems and Circuit Design RWTH Aachen University,Aachen,Germany,D-52074

4. Tobias Gemmeke Aachen Chair of Integrated Digital Systems and Circuit Design (IDS)

   RWTH Aachen University

   Affiliation as printed

   Chair of Integrated Digital Systems and Circuit Design RWTH Aachen University,Aachen,Germany,D-52074

5. Anke Schmeink
   Affiliation as printed

   Chair of Information Theory and Data Analytics