Mapping myocardial activation distributions using neural networks: 2-D simulation results

Area of Science:

• Cardiovascular Physiology
• Computational Biology
• Artificial Intelligence

Background:

• Accurate mapping of myocardial activation is crucial for understanding cardiac electrophysiology.
• Traditional methods for mapping cardiac activation can be invasive or limited in resolution.
• Simulating cardiac activation patterns aids in developing and testing new diagnostic tools.

Purpose of the Study:

• To evaluate the capability of neural networks in accurately mapping myocardial activation.
• To assess the performance of neural networks in determining the sequence and location of cardiac activation.
• To explore the use of QRS complexes in simulating normal and altered myocardial activation patterns.

Main Methods:

• A two-dimensional (2-D) fractal-based computer model was employed to generate myocardial activation data.
• Training datasets were created using two scenarios: random foci and hierarchical conduction blocking.
• Neural network learning was assessed using trained weights for both training and testing datasets.

Main Results:

• Neural networks achieved a mean error of less than 5% in identifying the site and timing of myocardial activation.
• High pointwise mean correlation coefficients were observed, exceeding 0.98 for conduction network cases and 0.84 for point foci cases.
• The network's ability to generate accurate activation maps was validated against target maps.

Conclusions:

• Neural networks demonstrate high accuracy in calculating cardiac activation maps.
• Electrocardiogram (ECG) lead data can be effectively utilized by neural networks for mapping diverse activation patterns.
• These findings suggest a promising non-invasive approach for detailed cardiac electrophysiological analysis.