Application of Convolutional Neural Network for Decoding of 12-Lead Electrocardiogram from a Frequency-Modulated...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Remote patient monitoring is crucial for healthcare, especially for elderly and visually impaired individuals.
• Existing methods for transmitting biosignals like electrocardiograms (ECG) can be limited by infrastructure requirements.
• Sonification of ECG offers a novel approach for wireless data transmission using readily available audio channels.

Purpose of the Study:

• To develop an AI-driven algorithm for 12-lead ECG sonification to enhance diagnostic reliability in remote monitoring.
• To create a robust system for transmitting ECG data wirelessly over audio channels, such as standard telephone calls.
• To validate the accuracy and efficacy of the sonification and reconstruction process for ECG signals.

Main Methods:

• Developed an ECG-to-Audio transformer algorithm using frequency modulation (FM) for eight independent ECG leads.
• Designed an Audio-to-ECG transformer algorithm employing a 1D convolutional neural network (CNN) to decode audio ECG streams.
• Trained the CNN model in unsupervised regression mode using the PTB-XL 12-lead ECG database (21,837 recordings).

Main Results:

• The AI algorithms achieved high accuracy in reconstructing ECG signals from audio, with low amplitude errors (RMSE = 3-7 μV, PRD = 2-5.2%).
• QRS detection showed high sensitivity (Se) and positive predictive value (PPV) exceeding 99.7%.
• P-QRS-T fiducial points' time deviation was less than 2 ms, demonstrating precise signal reconstruction across diverse patient data and arrhythmias.

Conclusions:

• The developed 12-lead ECG sonification technique provides a reliable wireless interface for remote patient monitoring.
• The AI-driven approach ensures diagnostic data integrity, supporting automated tools and medical experts in accurate diagnoses.
• This method facilitates accessible and secure transmission of vital ECG data, improving patient care, particularly for vulnerable populations.