Postprocessing technique to correct for background gradients in image-based R*(2) measurements.

Background static magnetic field gradients are a source of signal loss in gradient-echo imaging, as they typically result from discontinuity in the magnetic susceptibility at air-tissue boundaries. Moreover, these induced gradients severely compromise the measurement of R*(2), the effective transverse relaxation rate, which is of interest in many biomedical applications of MRI. Since the slice thickness is usually larger than the in-plane pixel dimensions, gradients parallel to the slice-select direction are of particular concern. In this work, a post-processing technique is introduced which attempts to correct the signal on the assumption that the background gradients are approximately linear across the voxel and the signal decay in the absence of these gradients is exponential. In this case, the time-domain signal is weighted by a sinc function characterized by the amplitude G(b) of the background gradient, which is typically not known a priori. The algorithm searches for the estimate of G(b) which yields the optimum fit of the corrected experimental data to an exponential. It is shown to be effective as long as this gradient is below a critical threshold. Evaluation in a phantom and in the human brain at 1.5 and 4 T demonstrates that this method can restore R(2)* in spite of the apparent rate constant exceeding the true value by up to 100%. Contrary to prospective correction techniques, the approach presented in this study does not prolong scan time.