Chemical Shift in MRI

In echo planar imaging, the very low bandwidth along the phase encode axis results in substantial chemical shift artifacts. At 1.5 Tesla, for example, using a 30 Hz/pixel bandwidth, fat and water are displaced by about 8 pixels. Further, the voxel sizes in EPI are usually rather large, for the reasons discussed above. Using a more or less typical 3 mm voxel, fat and water may be displaced from one another by 2.5 cm. This problems scales with field strength, so that in a 3 Tesla scanner the fat water chemical shift approaches 5 cm. Since most body tissues contain at least some water and fat, it is absolutely necessary to correct for the chemical shift problems.

Fortunately, there are a number of good technologies to manage chemical shift [13]. In the vast majority of cases, only the water component of the MR signal is of clinical interest. This is always the case in functional neuroimaging, where the lipid content of the brain is very low and the dominant source of fat signal is the component found in skin. It is therefore reasonable to simply suppress the fat signal outright. Usually, this is done by applying a fat saturation pulse prior to imaging. Because the chemical shift between fat and water is quite large, one can transmit a 90° pulse at the fat frequency without significantly affecting the water signal. After this pulse the fat signal will be in the transverse plane and it can be dephased easily by applying a gradient pulse. Until the fat signal has had time enough to recover its longitudinal magnetization it will not appear in the images. This so-called chemical shift saturation method does require excellent magnetic field homogeneity so that the frequencies of fat and water are well-resolved. Fortunately, today’s imaging instruments easily meet this requirement.

An alternative method of suppression is to use STIR (short TI inversion recovery)[14]. This approach takes advantage of the T1 difference between fat and other body tissues. An inversion (180°) pulse is applied immediately prior to the EPI imaging sequence timed such that the magnetization of fat is recovering through zero at the time of the 90° excitation pulse. Because the fat has no magnetization at that time, the 90° pulse does not result in the formation of any signal from fat. While STIR is a very effective method of fat suppression, it has side effects that make it less desirable. First of all, it alters the contrast of the images overall, as it adds T1 contrast. Secondly, the method works best if the inversion pulse is applied only when the tissue is fully magnetized. The latter requires that inversion recovery be used only with long TR images.