Efficient Generation of Personalized Numerical Models for SAR10g Estimation in High-Field Parallel MR Excitation

In magnetic fields 3 Tesla and higher, parallel magnetic resonance excitation (PEX) systems employ multiple radiofrequency excitation sources, thus posing increasing challenges regarding patient safety concers. At such high fields in PEX systems, the patient induced B1 field inhomogeneities cause local elevations in temperature measured in terms of the specific absorption rate (SAR), which could cause tissue damage. It has been previously demonstrated that SAR is highly relevant to body tissues, which differ among individuals and body types in terms of quantity and distribution. Due to this, SAR investigations are conducted through electromagnetic (EM) simulations involving numerical models of the PEX setup and the specific body type. For this purpose, personalized numerical models for each individual are required.
This thesis establishes the ground for generating such personalized numerical models, whole-body and head models, through computationally fast and efficient methods. The following aspects are covered:

Specifying tissues in the whole body and the head which are relevant to the 10-gram averaged SAR (SAR10g), through a presented approach called electrical property clustering.
•
Conduction of EM simulations with numerical body and head models involving the previously specified tissues, and the comparison of SAR10g resulting from few operating modes.
•
Establishment of an image-processing pipeline for the generation of personalized numerical voxel models involving the presented specific tissues.
•
Conversion of head voxel models to conformal surface mesh models.