Energy-efficient Real-time Myocardial Infarction Detection on Wearable Devices

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence

Background:

• Myocardial Infarction (MI) is a leading cause of death, often silent and recurrent.
• Continuous, real-time monitoring for MI is crucial for early detection and intervention.
• Existing wearable solutions for MI detection lack the necessary speed and energy efficiency for long-term use.

Purpose of the Study:

• To develop a fast and energy-efficient method for Myocardial Infarction detection using wearable devices.
• To introduce a Binary Convolutional Neural Network (BCNN) tailored for real-time MI detection.
• To evaluate the performance of the proposed BCNN against state-of-the-art methods on real hardware.

Main Methods:

• Implementation of a Binary Convolutional Neural Network (BCNN) for MI detection.
• Validation using the PTB diagnostic Electrocardiogram (ECG) database.
• Performance evaluation on actual wearable hardware, measuring speed and energy consumption.

Main Results:

• The proposed BCNN methodology demonstrates superior performance compared to existing state-of-the-art approaches.
• Real-hardware evaluation confirms the BCNN's faster processing speed.
• The BCNN achieves up to a 12x improvement in energy efficiency over current methods.

Conclusions:

• The developed BCNN provides a highly effective, fast, and energy-efficient solution for real-time Myocardial Infarction detection on wearable devices.
• This advancement enables more practical and sustainable long-term cardiac monitoring.
• The BCNN represents a significant improvement for wearable-based cardiac health technology.