Real-time arrhythmia detection using convolutional neural networks

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Cardiovascular diseases are a leading global cause of death.
• Current diagnostic methods are not ideal for continuous, out-of-hospital monitoring.
• Real-time detection of arrhythmias is crucial for long-term cardiac health management.

Purpose of the Study:

• To develop a real-time system for detecting arrhythmias using convolutional neural networks (CNNs).
• To evaluate the runtime performance and computational cost of the arrhythmia detection workflow.
• To demonstrate the feasibility and generalizability of CNN-based arrhythmia detection for in-home monitoring.

Main Methods:

• Utilized convolutional neural networks (CNNs) to classify arrhythmia conditions from electrocardiogram (ECG) images.
• Conducted extensive experiments to evaluate the computational cost of each workflow step for real-time processing.
• Validated the trained model using data from a customized wearable sensor in a lab setting.

Main Results:

• Achieved feasible real-time arrhythmic detection using CNNs.
• Demonstrated high accuracy and efficiency of the approach.
• Confirmed the generalizability of the model with wearable sensor data.

Conclusions:

• CNNs can effectively support real-time arrhythmic detection from ECG images.
• The developed approach is accurate, efficient, and suitable for in-home heart monitoring.
• This research integrates machine learning with traditional diagnostics for improved cardiovascular care.