The Effects of Intravoxel Dephasing and Incomplete Slice Refocusing on Susceptibility Contrast in Gradient-Echo MRI

ture to monitoring intravascular contrast agents or consent) , respectively, that have been acquired in the abvisualizing brain function. In particular, gradient-echo MRI sence and presence of macroscopic magnetic field gradients. sequences sensitized to alterations of cerebral blood oxygenThe images were obtained at 2.0 T (Siemens Magnetom ation (CBO) have resulted in exciting new tools for highSP4000) with use of the standard imaging head coil, together resolution mapping of human brain activation. The underlywith spin-density-weighted RF-spoiled FLASH sequences ing ‘‘functional contrasts’’ are due to regional and transient with T* 2 sensitivity under conditions that are identical to changes in the concentration of deoxyhemoglobin. Although those commonly employed for CBO-sensitive functional many studies have addressed the physiologic origin of MRI brain mapping in this laboratory (3, 4) . susceptibility contrasts and developed models of underlying In the absence of a gradient, Figs. 1a and 2a exhibit a magnetic field (or frequency) distributions, little attention certain degree of inhomogeneity-induced signal loss that has been paid to the way in which such effects become originates from intravoxel dephasing at the relatively long modulated by the imaging process itself. As a consequence, gradient-echo time. The images serve as references for the the influence of the slice-selection gradient on pertinent graeffects of a gradient along the frequency-encoding axis dient-echo images has been largely overlooked. (Figs. 1b and 2b), the slice-selection axis (Figs. 1c and 2c), In a previous study, the signal loss of a gradient-echo and the phase-encoding axis (Figs. 1d–1f and 2d–2f) . In image that stems from an additional macroscopic magnetic line with previous investigations (1, 2) , a further reduction field inhomogeneity was entirely attributed to a miscalibrain MRI signal intensity was only seen for the miscalibrated tion of the slice-selection gradient (1) . More recently, we gradient in the slice direction. A gradient in the frequencyhave identified the strength of the slice-selection gradient as encoding axis results in stretching or shrinking of the object a factor determining the quality of functional difference along that dimension (depending on relative signs) , while maps (2) . Here, we further strengthen the analysis of suscepits application along the perpendicular phase-encoding directibility contrasts by discussing the influence of all three imtion rotates the effective read gradient to an oblique axis. aging gradients. We demonstrate that, in general, gradientBy comparing respective image raw data sets in Fig. 3, it is echo MRI signal changes reflect contributions not only from seen that inhomogeneity components in the frequencyand ‘‘intravoxel dephasing,’’ but also from incomplete refocusphase-encoding direction shift the gradient echo along correing of the slice-selection gradient. Moreover, while intrasponding directions in k space. Fourier transformation results voxel dephasing originates from an increase in broadness of in altered signal phases, but does not lead to signal loss in the underlying field or frequency distribution, incomplete magnitude images. Of course, if very strong inhomogeneities slice refocusing results from an offset in its mean frequency. push a major portion of the echo outside the acquisition In a first step, we experimentally confirm that a homogewindow, the loss of low spatial frequencies results in a comneous shift of the entire frequency distribution results in plete lack of gross structures and a concomitant edge ensignal reduction due to incomplete slice refocusing as it is hancement (e.g., see Figs. 1e, 1f and 2e, 2f) . This phenomeless known than signal loss due to vectorial summation over non has already been noted for strongly asymmetric echoes an ensemble of nuclear spin moments with a broadened along the frequency(5) or phase-encoding direction (6, 7) .