A Surrogate Model Based on Artificial Neural Network for RF Radiation Modelling with High-Dimensional Data.

Area of Science:

• Biomedical Engineering
• Computational Electromagnetics
• Medical Imaging

Background:

• Accurate specific absorption rate (SAR) estimation is crucial for assessing electromagnetic field exposure in the brain.
• Uncertainty in electroencephalography (EEG) electrode positioning introduces significant variability in brain SAR calculations.
• Conventional uncertainty quantification methods can be computationally intensive, limiting their practical application.

Purpose of the Study:

• To develop and validate a novel artificial neural network (ANN) architecture for quantifying brain SAR uncertainty.
• To address the challenges posed by high-dimensional data arising from uncertain electrode positions.
• To provide a computationally efficient alternative to traditional uncertainty quantification techniques.

Main Methods:

• An artificial neural network (ANN) model was designed to handle high-dimensional input data related to electrode positions.
• The ANN was trained to predict the uncertainty in brain SAR values.
• The proposed method was compared against conventional uncertainty quantification approaches.

Main Results:

• The proposed ANN method effectively quantifies uncertainty in brain SAR values induced by electrode placement variations.
• The ANN approach demonstrates a significant reduction in computational expense and execution time compared to conventional methods.
• The method shows promise for efficient and accurate uncertainty quantification in bioelectromagnetics.

Conclusions:

• Artificial neural networks provide a powerful and efficient tool for uncertainty quantification in complex bioelectromagnetic simulations.
• The developed ANN method offers a practical solution for addressing the computational burden associated with high-dimensional uncertainty in brain SAR calculations.
• This approach facilitates more reliable assessments of electromagnetic exposure from EEG devices.