All studies of schizophrenia in our laboratory were conducted with an MRS imaging (MRSI) technique initially developed by Duyn et al.6 and subsequently modified by van der Veen et al.1 to allow the simultaneous acquisition of the water signal. It consisted of a spin echo technique (TR 2300 ms and TE 280 ms) from which 4 slices were acquired during approximately 30 min of examination. Outer volume suppression pulses were used to null the lipid signal from the skull. All of our human and animal studies were conducted on 1.5T GE systems (GE Medical Systems, Milwaukee, WI), but more recently we have been using a 3T system from the same manufacturer. At 1.5T, the nominal resolution of the voxels was 0.84 cc, with a slice thickness of 15 mm. Due to the increased susceptibility artifacts at 3T causing spectral line broadening, in the context of improved signal, the slice thickness at 3T was halved (voxel resolution of 0.42 cc), resulting in lines of the same width or narrower than at 1.5T for the same duration of exam. An example of spectra at 1.5T and 3T is shown in Figure 1. The analysis of the spectra consisted of integrating the values under the peaks corresponding to NAA (around 2 ppm), creatine (Cre: around 3 ppm) and choline (Cho: around 3.2 ppm) containing compounds. The values for each metabolite were then averaged over several voxels included in a region of interest (ROI). The ROIs were drawn on structural MRI images (T1 weighted) co-registered to the MRSI. Several ROIs were drawn on the MRSI scans, but our focus has been principally on the medial temporal lobe and the dorso-lateral prefrontal cortex (DLPFC), areas that have been implicated in the pathophysiology of schizophrenia by multiple lines of converging evidence.8,9 We have used ratios (NAA/Cre, NAA/Cho and Cho/Cre) as our main outcome measures. Although methods for absolute quantification of NAA are available, they are based on multiple assumptions and on the measurement of several quantities (e.g. the proportion of contamination with cerebrospinal fluid (CSF), the t2 of water in a certain tissue, etc) that allow for considerable uncertainty due to error propagation. The degree of error involved for these procedures has not been well characterized. All these technical issues are not a reason for concern when measuring ratios. However, the assumption underlying the use of ratios is that the denominator (in this case Cre) does not vary regionally or with pathological conditions. This seems to be generally true in the case of schizophrenia, although direct positive experimental evidence of this is lacking (on the other hand there is no available evidence to the contrary, to our knowledge).
For primate studies, a similar sequence to the one described above was used on the same 1.5T scanner equipment used for humans, but with a special surface coil for data acquisition. Only one slice was acquired through the frontal lobes.
Single voxel methodology was used for the study of rat models of a developmental lesion of the medial temporal lobe (described below). A 4.7T Varian Inova scanner was used with a 20 mm diameter circular surface coil. Spectra were acquired from a voxel of 36 mm3 nominal volume with a stimulated echo acquisition sequence (STEAM: TR=3000 ms; TE=12 ms).
Representative spectrum for a 1.5T MRSI acquisition
Was this article helpful?