Spectroscopy

A vast amount of literature about MR spectroscopy (MRS) exists. However, so far only a very few studies deal with volume selective in vivo proton spectroscopy of various diseases, in particular MS. The stimulated echo, as originally described by Hahn (50), is an echo which follows after 3 RF excitations. Real in vivo proton MRS was made possible by the pioneering work of Frahm et al. ( 5 l), who introduced three orthogonal slice selective 90" R F pulses, which give a stimulated echo from the intersection of the three slices. Shimming, and in particular localized shimming, is of extreme importance for obtaining high-quality spectra in vivo. Magnetic field inhomogeneities reduce both spectral resolution and the S/N ratio by reducing peak intensities and broadening lines. Typically, global shimming of the entire head is performed first, resulting in a water proton Full linewidth at Half Maximum (FWHM) of 12-18 Hz. Afterwards, localized shimming is performed giving a FWHM of 3-6 Hz. RF and gradients of high quality are needed to obtain good volume selection. Nevertheless, theoretically, truncation of the R F envelope will always give some signals outside the nominal voxel. In VII we have evaluated this contamination in a phantom study and we found no significant contamination. The best method for evaluating volume selection is probably by imaging the slice through the VOI in all three directions (52). This can be achieved by adding a phase-encoding and a readout gradient to the sequence. The phase-encoding table and the compensatory gradient for the readout gradient should be placed in the second TE period. Integration of the slice profiles and comparison with the nominal VOI will give the amount of contamination (52). Using this method, Van Zijl(52) found that 80% of the signal originated from a nominal 5X5X5 mm cube and more than 93%) within a 7X7X7 mm cube. However, Van Zijl seems not to have taken into account that the readout and phase-encoding gradients themselves give additional deterioration of the voxel definition because of truncation. Although the hardware, gradients and field strength are not directly comparable to ours, we should bear in mind that signal of adjacent tissue outside the nominal VOI may have contributed to the resulting signal. We later studied the volume selection by looking at the correlation between the MR signal and the nominal volume for different sizes. Good linearity was found, indicating that side-loops are not a major problem (145). Another problem is the chemical-shift-dependent localization of the metabolites under study because one may be comparing metabolite levels in different volumes. In our situation (VI an VII), using slice selective gradients of 1 mT/m at 1.5 T, the bandwidth corresponds to an uncertainty in the position of the VOI of less than 8 mm, i.e.+4mm, for a chemical shift range of about 5 ppm i.e., 320 Hz. Increasing the slice-selection gradients will reduce this problem. Other sources of errors have been mentioned in the physics section. It should be noted that a rigorous physical treatment of the stimulated echo sequence is only possible by a quantum mechanic description method.