Ultrafast Magnetic Resonance Imaging.

We propose a radical advance in Magnetic Resonance Imaging. MRI remains slow because it requires successive applications of magnetic field gradients to encode for spatial location. Parallel MRI accelerates imaging by permitting undersampling of k-space using multiple detectors, each with a unique spatial sensitivity to the radio field emitted by the object. This is called spatial sensitivity encoding. A fourfold undersampling means a fourfold acceleration, however a five-fold undersampling cannot be reconstructed. This is because too much reliance is placed on spatial sensitivity encoding, and spatial sensitivity encoding is less efficient than gradient encoding when the detectors are large. On the contrary, we hypothesized that when very large numbers of very small detectors are deployed, it would render spatial sensitivity encoding efficient, and would allow complete elimination of gradient reversals and therefore extremely rapid MR imaging. Since each detector receives all of the signal from all of the object all of the time, signal is expected to be high, and each detector is predicted to receive signal of the same order of magnitude as one large detector in conventional MRI. Moreover, the elimination of gradients removes the major source of electrical noise. Unexpectedly, therefore, we anticipate that signal-to-noise (SNR) will be high. By means of a detailed simulation, in which the exact Hertzian radio field is simulated for each of 32x32x32 voxels interacting with each of 1024 detector loops, we demonstrate that 3-dimensional MR images can indeed be acquired very quickly. Nevertheless, we also simulated imaging in the presence of unforeseen sources of noise, and found that image reconstruction is resistant to noise, even with unrealistically low SNR values. In summary, high quality volumetric MRI is likely to be achievable in milliseconds.