TUHH: Machine Learning Speeds up EM Field Exposure Prediction in Human Head

TUHH > Forschung und Transfer > Neues aus der Forschung

Machine Learning Speeds up EM Field Exposure Prediction in Human Head

Advancements in wireless power transfer (WPT) system and brain machine interface (BMI) have triggered many applications to promote the human daily life. Meantime, the development of these modern technologies entails critical considerations in terms of bio-electromagnetic (Bio-EM) compatibility and EM safety. To quantify the electromagnetic field (EMF) exposure level, specific absorption rate (SAR) is widely employed to measure absorbed power per unit mass (W/kg). Past studies have shown that the SAR value of human tissues is sensitive to material uncertainty, which has attracted numerous worldwide studies that employ numerical simulations and experimental tests. Following this, there is a growing interest in implementing machine learning (ML) method in SAR prediction of human tissues.

Since 2021, researchers from the Institut für Theoretische Elektrotechnik at TUHH have conducted studies on the exposure of human tissues to EM fields using high-resolution full-wave models [1-3]. Recently, the problem's complexity has been further extended to include material uncertainty. Specifically, for plane wave illumination at 13.56 MHz, an artificial neural network (ANN) has been employed to predict SAR in multiple head models, achieving a high accuracy rate of 98%, as shown in Fig. 1. The work, entitled "SAR Prediction in Human Head Tissues with Varying Material Parameters Using an Artificial Neural Network"[4], will be presented in June at Bio-EM 2023 in Oxford, UK.

Figure 1: Machine-learning based high accurate SAR prediction in human head[1-4].

With the trained ANN model, the field exposure level in human tissues can be examined quickly. Moreover, due to the small electric size of the human head at the MHz frequency range, highly simplified models can be extracted with great promise. This work is in progress and can enable efficient SAR predictions for even highly realistic models. Future work will focus on expanding the scope of ANN modeling to a much larger category of human models and testing the approach with realistic near-field exposures up to GHz frequencies. These ongoing studies will undoubtedly advance our understanding of the interactions between EM fields and human tissues, with important implications for Bio-EM research.

Ansprechpartner

M.Sc. Hamideh Esmaeili

Dr. Cheng Yang

Institut für Theoretische Elektrotechnik

Technische Universität Hamburg (TUHH)

Blohmstraße 15, 21079 Hamburg, Germany

E-Mail: hamideh.esmaeili(at)tuhh.de

cheng.yang(at)tuhh.de

Reference:

[1] C.Yang, "Bridging the Modeling Gap: Huygens' Principle for Brain Implants", News from research [online] https://www.tuhh.de/tuhh/forschung/neues-aus-der-forschung/bridging-the-modeling-gap-huygens-principle-for-brain-implants-2.

[2] C. Yang, M. Schierholz, E. Trunczik, L.M. Helmich, H.D. Brüns, and C. Schuster, “Efficient and Flexible Huygens’ Source Replacement of mm-scale Human Brain Implant”, Joint IEEE International Symposium on Electromagnetic Compatibility, Signal & Power Integrity, EMC Europe, 2021.

[3] H. Esmaeili, C. Yang and C. Schuster, "Flexible Numerical Evaluation of Human Head Exposure to a Transmitter Coil for Wireless Power Transfer at 13.56MHz," International Symposium on Electromagnetic Compatibility, EMC Europe, 2022.

[4] H. Esmaeili, C. Yang and C. Schuster, "SAR Prediction in Human Head Tissues with Varying Material Parameters Using an Artificial Neural Network", Accepted for presentation on Bioelectromagnetics 2023, UK.

TU Hamburg

Technische Universität

Hamburg (TU Hamburg)

Am Schwarzenberg-Campus 1

21073 Hamburg

Machine Learning Speeds up EM Field Exposure Prediction in Human Head

Ansprechpartner

TU Hamburg

Soziale Netzwerke

Weiterführende Links