Iterative reconstruction methods for rosette trajectories in functional MRI.

Abstract

Blood Oxygenation Level Dependent (BOLD) functional MRI (Magnetic Resonance Imaging) is a noninvasive imaging method to explore the function of the human brain using the change of tissue <italic>T</italic>₂* as the contrast mechanism. Fast imaging of the whole brain volume is important in fMRI, since the brain function can be spread over the entire brain for a very simple task. Simultaneous Multislice Acquisition using Rosette Trajectory (SMART) was proposed as one of the fast multislice imaging methods, but the non-iterative conjugate phase (CP) reconstruction resulted significant amount of artifacts from the off-resonance slices. In this thesis, we develop a physics model based iterative reconstruction method for SMART (iSMART) to reduce the artifacts and demonstrated the method at 3T. Simulation studies and functional experiments were performed to show that the method reduces most of the artifacts at the price of slight decrease in the in-plane spatial resolution and increased computation. The accuracy of iSMART depends on the accuracy of the system model it uses in the reconstruction. To increase the accuracy of the system model, we also developed a robust, dynamic <italic>R</italic>₂*-and-field-map-corrected image reconstruction method. This method is based on the spatio-temporally regularized iterative reconstruction algorithm. The spatio-temporally regularized iterative reconstruction estimates not only provides accurate field maps, but also field-map corrected images and the <italic>R</italic>₂* map. Therefore, the spatio-temporally regularized iterative reconstruction can be used for dynamic <italic>R</italic>₂* mapping for functional MRI experiments, where dynamic <italic>R</italic>₂* mapping has a few advantages over the conventional <italic>T</italic>₂* weighted BOLD imaging. It also can be used as dynamic field-map corrected BOLD imaging method. In simulations, we show that the method can improve the accuracy of the field map from the initial guess of the field map. In functional experiments, we show evidence that the proposed method can dynamically estimate and correct for the field map changes during a functional study. Further improvement on the accuracy of the system model can be achieved by the proposed 2D k-space trajectory measurement method. It is well known that time-varying magnetic fields induce eddy currents, which can distort the designed gradient waveforms. The proposed k-space trajectory measurement method utilizes pencil excitation at several spatial locations, and it measures main field fluctuations and crosstalk between two gradient channels. It requires reasonable scan time, and does not require extra equipment such as surface coil or a point phantom. A phantom experiment in a 3T scanner shows that even with the vendor-provided eddy current correction, there is remaining uncorrected distortion in the gradient waveforms.