Smart ECG Biosensor Design with an Improved ANN Performance Based on the Taguchi Optimizer.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Electrocardiogram (ECG) signal analysis is crucial for cardiac diagnostics.
• Predicting ECG signals at specific auscultation sites from other measurement points presents challenges.
• Existing Artificial Neural Network (ANN) models can suffer from hyperparameter tuning issues and suboptimal accuracy.

Purpose of the Study:

• To design a smart biosensor for predicting ECG signals at a target auscultation site using signals from other sites.
• To enhance ANN model accuracy and address hyperparameter challenges through optimization.
• To identify optimal input combinations for improved ECG signal prediction quality.

Main Methods:

• A hybrid biosensor architecture combining Artificial Neural Networks (ANNs) with a Taguchi optimizer was developed.
• The Taguchi optimizer was employed to fine-tune ANN hyperparameters and optimize input features.
• Various input combinations were systematically evaluated to determine the best predictive configuration.

Main Results:

• The proposed hybrid biosensor significantly improved the prediction accuracy of ECG signals compared to a standard ANN-based biosensor.
• The optimization process identified crucial factors influencing prediction and their importance.
• Simulation results using an actual database demonstrated prediction accuracy exceeding 94%.

Conclusions:

• The developed smart biosensor effectively predicts ECG signals from specific auscultation sites with high accuracy.
• The integration of ANNs with Taguchi optimization offers a robust solution for biosensor design and ECG signal prediction.
• This approach provides a significant advancement in non-invasive cardiac monitoring and diagnostics.