Applying Functional MR Imaging to Brain-behavioral Research: Can We Do Better than Simple Clinical Measures?

In the impressive brain-behavior multidisciplinary study in this issue of the AJNR (page 621), Capizzano et al evaluated four study groups: demented patients with subcortical ischemic vascular disease (SIVD)(n = 11), demented patients with probable Alzheimer's disease (n = 18), a mildly cognitively impaired group with white matter disease (n = 14), and a healthy control group (n = 20).

The investigators derived imaging measures of brain structure including segmentation of brain into volumes of gray and white matter, CSF and white matter abnormalities, functional measures using ¹HMRSI, including derivation of N-acetylaspartate (NAA) and NAA/Creatine (Cr) ratios corrected for atrophy and tissue composition, and behavioral measures including the Clinical Dementia Rating and Mini Mental Status Exam.

Statistical methods confirmed several hypotheses. In the SIVD group, NAA and NAA/Cr in the cerebral cortex and white matter, but not in the hippocampus, were reduced independent of atrophy and tissue composition when compared with the control group, suggesting neuron loss or metabolic impairment in these regions. The presumed Alzheimer's group, but not the vascular group, showed decreased NAA/Cr in the hippocampus. Cortical NAA measures were inversely correlated with the number of lacunes and with the volume of white matter disease.

This study underscores how very far radiologic research has come but also how very far it needs to go. Consider the following points:

1) The significant results in this article consisted of reductions in brain metabolites in the 10.25% to 12.64% ranges for the subcortical vascular dementia group. Similarly, the presumed Alzheimer's group showed 10.33% reduction in NAA/Cr when compared with the control group, and a similar reduction when compared with the subcortical vascular dementia group. Only NAA/Cr was significantly reduced in the Alzheimer's group, not NAA. Thus, the order of magnitude of the functional differences in this article, and in similar studies in the literature, is quite small.

2) Despite state-of-the-art measuring techniques, there were no significant correlations found between MMSE scores and either structural or metabolic changes for any of the groups studied. This is evidence of the low sensitivity of our measures.

3) Reduced NAA or NAA/Cr has also been reported in the frontal lobes of schizophrenic patients, the white matter of multiple sclerosis, St. Louis encephalitis, and traumatic brain injury patients, and the hippocampi of schizophrenic and epileptic patients. This is evidence of the low specificity of our measures.

4) Clinical measures alone, such as the widely used DSM-III-R criteria, have shown only 51% sensitivity, 66% accuracy, and high (97%) specificity in autopsy studies (1). In other words, clinical criteria apply to healthy subjects with great reliability, but for the diagnosis of demented patients can have the sensitivity of a coin toss. Thus, the typical radiologic experiment that attempts only to predict group membership (ie, is the scan that of a patient or a control subject) is further confounded by insensitive and inaccurate clinical measures.

Despite the above reservations, Capizzano et al have shown quite remarkable results that support, but do not prove, the hypothesis that in subcortical vascular dementia, the white matter lesions, notably lacunes, disconnect the cortex from the subcortical white matter. This accounts for the metabolite cortical deficits in the SIVD group, and not in the presumed Alzheimer's group. Nevertheless, the presence of coexisting Alzheimer's disease remains an issue the investigators acknowledge. We would like to propose the following study to address this question.

The presumed etiology of subcortical vascular dementia is small vessel occlusive disease. Functional radioisotope studies such as positron emission tomography and single-photon emission CT (SPECT), and functional MR imaging reports such as the current study, have shown decreases in cerebral blood flow, oxygen consumption, NAA, and NAA/Cr in subcortical vascular dementia. It would be very interesting to evaluate cortical function by using a cerebrovascular vasodilatory agent such as acetozolamide (Diamox) or CO₂ as a provocative test to evaluate cortical vascular reserve. In patients with SIVD, these vasoreactive agents may unmask the underlying hemodynamic reserve by showing decreased perfusion parameters when compared with a normal brain. In normal subjects and in patients with Alzheimer's disease, there is increased blood flow after acetazolamide or CO₂ challenge. The effect of acetozolamide on regional cerebral blood flow was shown by Bonte et al (2), who demonstrated improved temporoparietal perfusion in presumed Alzheimer's disease patients following acetozolamide-challenged SPECT imaging; those with SIVD showed no change or decreased cerebral blood flow. This study could be performed using contrast-enhanced or unenhanced perfusion-weighted MR sequences such as dynamic susceptibility contrast imaging or arterial spin labeling techniques.

Potentially, functional MR imaging may prove to be useful because improved perfusion results in improved metabolite activity and NAA or NAA/Cr measures. Regardless of the method of measurement, acetozolamide-challenged MR imaging may provide valuable insights into the study of subcortical diseases such as SIVD and cortical diseases such as Alzheimer's disease.