Preview

PDF, English - main document Download (34MB) | Lizenz: Minimizing the Adverse Effects of Electric Fields in Magnetic Resonance Imaging using Optimized Gradient Encoding and Peripheral Nerve Models by Davids, Mathias underlies the terms of Creative Commons Attribution-NonCommercial-NoDerivatives 4.0

Abstract

Magnetic Resonance Imaging (MRI) is an important imaging modality in both the clinic and in research. MRI technology has been trending toward increasing field strengths to improve the signal-to-noise ratio of the MR signal and fast excitation/encoding strategies to more flexible target anatomical regions during excitation to reduce the total imaging time. While largely successful, both strategies rely on the application of increasingly strong and rapidly switched magnetic fields: the radio frequency (RF) field for excitation and the gradient field for encoding. The technology for generating these fields (and rapidly switching them) has advanced to the point that we are limited by biological responses to the switching fields. For the gradient field, the electric field generated in the tissue causes peripheral nerve stimulation (PNS) causing mild but bothersome sensations at low levels, up to pain or cardiac malfunction at higher levels. The electric fields created by the much faster time-varying RF cause heat deposition, ultimately denaturing proteins and causing tissue damage. In this thesis, methods are presented to characterize and minimize these two problems associated with the switched magnetic fields in MRI. The deposited RF energy (Specific Absorption Rate, SAR) incurred during shaped excitations can be significantly reduced by optimizing gradient and RF waveforms for inner-volume excitations that allow imaging of a sub-volume of the body without wrapping artifacts. The adverse effects of the switching gradient fields are addressed by designing time-optimal gradient encoding waveforms and by developing a method to predict and characterize PNS using field simulations and a full-body nerve model allowing these critical effects to be addressed at the gradient coil design stage. In the first part, time-optimal gradient trajectories are demonstrated that use the gradient hardware at the maximum available performance. The skeleton of the trajectory is defined by a set of k-space control points. The method optimizes gradient waveforms that traverse the k-space control points in the minimum possible amount of time. By using an analytic representation of the gradients (piece-wise linear), the design process is fast and numerically robust. The resulting trajectories sample k-space efficiently while using the gradient system at maximum performance. Compared to the leading Optimal Control method, the proposed method generates gradient waveforms that are 9.2% shorter. The computation process is ∼100x faster and does not suffer from numerical instabilities such as oscillations. In the second part, a method is developed that jointly optimizes parallel transmission RF and gradient waveforms for fast and robust 3-D inner-volume excitation of the MRI signal in minimal time and with minimal energy deposition. The optimization of the k-space trajectories is based on a small number of shape parameters that are well-suited for joint optimization with the RF waveforms. Within each iteration of the trajectory optimization, a small tip-angle least-squares RF pulse design problem is solved. Using optimized 3-D cross (shells) trajectories, a cube shape (brain shape) region was excited with 3.4% (6.2%) NRMSE in less than 5 ms using a 7 T scanner with 8 Tx channels and a clinical gradient system (Gmax = 40 mT/m, Smax = 150 T/m/s). Incorporation of off-resonance robustness in the pulse design significantly altered the k-space trajectory solutions and improved the practical performance of the pulses. In the final part, a framework is presented that simulates PNS thresholds for realistic gradient coil geometries and thus allows, for the first time, to directly address PNS in the coil design process. The PNS framework consists of an accurate body model for simulation of the induced electric fields, an atlas of peripheral nerves, and a neurodynamic model to predict the nerve responses to imposed electric fields. With this model, measured PNS thresholds of two leg/arm solenoid coils and three commercial actively-shielded MR gradient coils could be reproduced with good accuracy. The proposed method can be used to assess the PNS capability of gradient coils during the design phase, without building expensive prototype coils.