Multi-projection magnetic resonance inverse imaging

Using highly parallel radiofrequency (RF) detection, magnetic resonance inverse imaging (InI) can achieve 100-millisecond temporal resolution with the whole brain coverage. This is achieved by trading off partition encoding steps and thus the spatial resolution for a higher acquisition rate. The reduced spatial information is typically estimated by solving under-determined inverse problems using RF coil sensitivity information. The reconstructed InI images under the minimum l-2-norm constraint typically demonstrate a lower spatial resolution. Here we propose the multi-projection inverse imaging (mInI) method to combine different projection images to reduce the loss of spatial resolution of InI. Specifically, coronal, sagittal, and transverse projection images are acquired from different runs of the functional MRI (fMRI) acquisitions using a 32-channel head coil array. Simulations show that, compared to the InI reconstruction using the minimum l-2-norm, mInI improves the spatial resolution of the reconstructed image significantly. Going from one projection to three projections, the spatial resolution quantified by the full-width–half-maximum of the point-spread function (PSF) is improved from 2.6 pixels to 1.4 pixels (4 millimeter per pixel nominal resolution). Considering the shape of the PSF, the effective spatial resolution improves from 16.9 pixels to 4.7 pixels. In vivo fMRI experiments using a two-choice reaction time task shows visual and sensorimotor cortical activity spatially consistent with typical EPI data, yet mInI offers the 100 millisecond temporal. The mInI data with three projections reveal that the hemodynamic response at the lateral geniculate nuclei (LGN) and at the visual cortex precedes that at the sensorimotor cortex by 1300 ms and 700 ms respectively. mInI can be applied to BOLD-contrast fMRI experiments to characterize the dynamics of the activated brain areas with a high spatiotemporal resolution.