A novel convolutional neural network for reconstructing surface electrocardiograms from intracardiac electrograms and...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Point-of-care monitoring for cardiac pathologies often relies on implanted devices.
• Reconstructing 12-lead surface electrocardiograms (ECG) from intracardiac electrograms (EGM) can enhance patient monitoring.
• Current methods for ECG reconstruction have limitations in accuracy and efficiency.

Purpose of the Study:

• To develop a novel convolutional neural network (CNN) framework for multivariate input-output mapping.
• To implement this CNN for accurate ECG reconstruction from EGM and vice versa.
• To create a new diagnostic tool for ECG classification and arrhythmia detection.

Main Methods:

• A novel CNN framework was designed for multivariate data transformation.
• The model was trained and evaluated on ECG and EGM data from 14 patients.
• The framework was also tested for non-patient-specific reconstruction and for EGM synthesis from ECG.
• Feature analysis identified an overcomplete basis for ECG space, utilized for arrhythmia classification.

Main Results:

• The CNN model achieved high accuracy in reconstructing 12-lead ECG from EGM, outperforming previous methods.
• Accurate reconstruction was possible even with a single EGM lead.
• The model demonstrated comparable accuracy in synthesizing EGM from 12-lead ECG.
• The derived feature basis achieved 0.98 average accuracy in classifying arrhythmias on the MIT-BIH database.

Conclusions:

• The proposed CNN framework offers an efficient and accurate method for ECG reconstruction and EGM synthesis.
• This technology has the potential to significantly improve point-of-care cardiac monitoring.
• The developed diagnostic tool shows promise for accurate arrhythmia classification.