Related Experiment Videos
Using an artificial neural network to detect activations during ventricular fibrillation.
M T Young1, S M Blanchard, M W White
1Department of Biological and Agricultural Engineering, North Carolina State University, Raleigh, North Carolina 27695-7625, USA.
Computers and Biomedical Research, an International Journal
|April 25, 2000
Summary
Artificial neural networks effectively identify cardiac electrical activation patterns during ventricular fibrillation. Staged training significantly improved the accuracy of these networks in detecting arrhythmias, aiding sudden death research.
Area of Science:
- Cardiology
- Computational Biology
- Biomedical Engineering
Background:
- Ventricular fibrillation (VF) is a life-threatening cardiac arrhythmia causing sudden cardiac death.
- Accurate identification of electrical activation patterns during VF is crucial for improved understanding and treatment.
- Current methods for analyzing cardiac electrical activity during VF have limitations.
Purpose of the Study:
- To evaluate the efficacy of artificial neural networks (ANNs) in identifying cardiac tissue activation patterns during ventricular fibrillation.
- To assess the impact of different data preprocessing methods (Rule-Based Method, Current Source Density Method, Transmembrane Current Method) on ANN performance.
- To investigate the benefits of a novel 'staged training' approach for enhancing ANN accuracy in VF analysis.
Main Methods:
- Development and training of feedforward ANNs using backpropagation.
- Utilizing Rule-Based Method and Current Source Density Method for initial data preprocessing.
- Employing Transmembrane Current Method in conjunction with other methods for advanced data preprocessing.
- Implementing a staged training strategy with varying training datasets across different stages.
Main Results:
- Both trained ANNs achieved high classification accuracy, correctly identifying over 92% of new test examples.
- The use of staged training led to a notable improvement in the performance of both ANNs.
- The ANNs demonstrated a robust ability to detect and classify activation patterns during simulated ventricular fibrillation.
Conclusions:
- Artificial neural networks show significant promise as a tool for identifying cardiac electrical activation during ventricular fibrillation.
- Staged training is an effective method for improving the accuracy and reliability of ANNs in analyzing complex cardiac arrhythmias.
- This approach could lead to better diagnostic tools and therapeutic strategies for managing sudden cardiac death associated with VF.