Learning-based estimation of dielectric properties and tissue density in head models for personalized radio-frequency dosimetry.

Researchers

Journal

Modalities

Models

Abstract

Electromagnetic dosimetry at radio frequencies is an important task in human safety and compliance of related products. Recently, computational human models generated from medical images are often used for such assessment, especially for considering the inter-variability. However, the common procedure to develop personalized models is time consuming as it includes excessive segmentation of several components representing different biological tissues. This limits the feasibility of real-time assessment for personalized dosimetry. Deep learning methods have shown to be a powerful approach for pattern recognition and signal analysis. Convolutional neural networks with deep architectures are proved to be robust for feature extraction and image mapping in several biomedical applications. In this study, we develop a learning-based approach for fast and accurate estimation of tissue dielectric properties from magnetic resonance images. Smooth distribution of dielectric properties in head models, which was realized by the procedure without tissue segmentation, improves the smoothness of specific absorption rate (SAR) distribution as compared to commonly-used procedure. The estimated SAR distributions as well as that averaged over 10-g of tissue in cubic shape are found to be of high consistency with those computed using a conventional methods employing a segmentation.
© 2020 Institute of Physics and Engineering in Medicine.