The Effect of Brain Atrophy on Cerebral Hypometabolism in the Visual Variant of Alzheimer Disease FREE

ALZHEIMER disease (AD) is a neurodegenerative disorder that most often manifests itself initially through memory loss, followed by decreases in higher-order cognitive skills such as attention, language, and planning.^{1 - 2} Results from structural imaging studies, using both magnetic resonance (MR) imaging and computed tomography, have shown increases in brain atrophy consistent with the clinical and neuropathologic findings.^{3 - 6} Results from functional imaging studies using positron emission tomography (PET) have identified reductions in the regional cerebral metabolic rate for glucose (rCMRglc)⁷ and oxygen⁸ in patients with AD compared with healthy control subjects in temporal, parietal, and prefrontal cortex and in limbic structures.

In a less frequent clinical subtype of AD, the visual variant of AD (AD + VS), the initial symptoms are characterized by functional impairment in visuospatial skills in the absence of memory complaints, followed by alexia, agraphia, and visual agnosia.⁹ Postmortem findings in patients with confirmed AD + VS demonstrate a concentration of neuropathologic lesions in occipital, parietal, and posterior cingulate cortex with a relative sparing of the frontal and temporal lobes.^{9 - 12} Structural MR imaging results have shown marked posterior cortical atrophy,^{11 ,13} with relative sparing of frontal and temporal cortices. Findings on PET imaging have identified reduced rCMRglc in posterior cortical areas, with less hypometabolism in frontal and limbic regions.^{14 - 16} Thus, patterns of brain atrophy and functional hypometabolism are different between patients with AD + VS and those with typical AD.

Because of its limited spatial resolution, PET measures are subject to partial volume effects (PVE), which may result in artificially decreased values of rCMRglc. In PET studies of patients with AD, it is critical to know if reduced rCMRglc reflects a reliable in vivo index of neuronal/synaptic dysfunction or is merely a consequence of increased PVE caused by brain atrophy. Recently we showed that rCMRglc remained lower in patients with typical AD than in matched controls after PVE correction, indicating that brain atrophy alone does not account for the observed hypometabolism.¹⁷

Given that posterior brain atrophy in patients with AD + VS is generally greater than in patients with AD, and given its atypical distribution, the question arises as to what proportion of the occipital hypometabolism observed in AD + VS is caused by cortical atrophy. Our objective was to determine whether rCMRglc abnormalities observed in patients with AD + VS are a true index of neuronal/synaptic dysfunction, or simply a reflection of brain atrophy. We compared rCMRglc in patients with AD + VS with that in age-matched healthy control subjects, before and after rCMRglc was corrected for brain atrophy by means of segmented structural MR images. To examine whether the distinct distribution of rCMRglc abnormalities in patients with AD + VS compared with patients with typical AD is caused by different brain atrophy patterns, we compared these patients with AD + VS with a group of patients with typical AD from a previously reported study.¹⁷

We used 4 different groups: 2 patient groups and 2 healthy control groups that have already been described.^{14 ,17} Clinical and demographic data of the 4 groups are presented in Table 1.

The first 2 groups consisted of 10 patients with AD + VS and 18 age-matched healthy control subjects, respectively. A detailed description of the clinical, neuropsychological, and neuro-ophthalmologic examinations of the patients with AD + VS is provided elsewhere.^{14 ,18} Briefly, all patients met the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA) criteria¹⁹ for probable AD and had a history of early and prominent visual and visuospatial symptoms.^{14 ,18} Nine of the 10 patients with AD + VS were included in the original reports^{14 ,18} ; 1 was recruited subsequently. The 18 healthy control subjects were part of the original control group of 25.^{14 ,18} One patient and 7 control subjects from the previous study were excluded because adequate MR images were unavailable.

Data from the 15 patients with typical AD and their 19 age-matched healthy control subjects have been published previously.¹⁷ Briefly, all patients fulfilled NINCDS-ADRDA criteria for probable AD. All the subjects in the original report were included in this study.

Criteria for inclusion of patients in the AD + VS group are presented in detail elsewhere.^{14 ,18} Clinical and psychological evidence¹⁸ demonstrate that patients with AD + VS had more deficits in visuoconstructional abilities and visuospatial attention than did the patients with typical AD. In addition, the patients with AD + VS had better verbal memory than did those with typical AD. Both patient groups differed from healthy controls in all measures of cognitive function, including memory, visuospatial ability, and language.

Patients (or holders of durable power of attorney) and healthy control subjects gave their written informed consent after a full explanation of the procedures and risks of the study.

Details regarding imaging procedures have been published previously.¹⁴ Briefly, for the patients with AD + VS and their control subjects, a Scanditronix PC-1024-7B tomograph (Scanditronix, Uppsala, Sweden) with an in-plane resolution of 6.0 mm at full-width half-maximum (FWHM) and axial resolution of 10 mm (FWHM) was used to measure rCMRglc following an intravenous injection of 185 MBq of [¹⁸F] fluorodeoxyglucose. Forty-five minutes after injection of the radiotracer, 2 interleaved sets of 7 images offset by 6.9 mm were obtained parallel to and 10 to 100 mm above the inferior orbitomeatal line. Subjects were studied at rest with eyes covered and ears occluded. Arterial blood samples were collected for measurement of plasma radioactivity and glucose concentration. Values of rCMRglc were calculated in units of milligrams of glucose per 100 g of tissue per minute.^{20 - 21} Images were reconstructed and then interpolated to 48 slices for a voxel size of 2 × 2 × 2 mm.

The PET scanner used for the patients with typical AD and their control subjects was a Scanditronix PC-2048 tomograph (Scanditronix) that acquires 15 slices with an in-plane resolution of 6.5 mm FWHM and 6.5 mm between slice centers. Glucose metabolism was measured following an intravenous bolus injection of 157.25 MBq of [¹⁸F] fluorodeoxyglucose, according to the same procedure as above. The images were reconstructed and then interpolated to 43 slices with a 2 × 2 × 2-mm voxel size.

The MR images were acquired in patients with AD + VS and their control subjects with a Picker 0.5-T scanner (Picker, Cleveland, Ohio) using both axial and coronal T1-weighted images (repetition time, 2000 milliseconds; echo time, 20 milliseconds). The field of view was 25 cm, with voxel sizes of 0.976 × 0.976 × 6 mm for the coronal images and 0.976 × 0.976 × 7 mm for the axial images. In the AD + VS group, 3 subjects had axial images and 7 had coronal images. In the healthy control group, 8 subjects had axial images and 10 had coronal images. The MR images were taken within 2 months of the PET scans for each patient with AD + VS, except 1 whose MR images were obtained 1 year later. For the healthy controls, 9 scans were within 1½ years of the PET scan, and the rest were within 6 years.

The MR images for the typical AD group and their control subjects were acquired within 1 year of the PET scan. Two types of scans were used for the atrophy correction. For 7 patients with AD and 11 controls, a Picker 0.5-T scanner (Picker) was used, and for 8 patients with AD and 4 controls, a GE Signa 1.5-T scanner (General Electric Co, Milwaukee, Wis) was used. The structural images were T1 weighted (repetition time, 24 milliseconds; echo time, 5 milliseconds; flip angle, 45°) with a field of view of 26 cm for the Picker scanner and 24 cm for the GE scanner. Voxel sizes of the images were 1.02 × 1.02 × 2 mm for the Picker scanner and 0.938 × 0.938 × 2.0 mm for the GE scanner.

Computations were performed on a Sun SPARC computer (Sun Microsystems, Palo Alto, Calif) with ANALYZE version 7.5.4 software (BRU; Mayo Foundation, Rochester, Minn), and also with algorithms implemented using MATLAB version 4.2 software (MathWorks Inc, Natick, Mass) and the C computer language.

The atrophy correction algorithm uses an individual's MR image to correct for PVE in the PET scan.^{17 ,22} Scans were edited manually for nonbrain and were segmented into brain and cerebrospinal fluid compartments. For the AD + VS group and its healthy control group, the 2 interleaved PET images were registered to one another.²³ The MR images were then registered to the PET images using a rigid model body transformation,²⁴ and the same transformation was applied to the segmented MR volume. A mask was made based on the brain tissue compartment from the segmentation. The dispersion coefficient was calculated by the convolution of the brain mask with the point spread function of the PET camera. The original PET volume was then divided by the dispersion coefficients to obtain the atrophy-corrected volume. Each PET volume was corrected separately. The algorithm limits the corrected values to the maximum measured value before PVE correction.¹⁷

The PET scans were transformed into the stereotactic space of the Talairach and Tournoux atlas^{25 - 26} and smoothed with an isotropic gaussian filter (FWHM, 10 mm). The voxel dimensions after transformation were 2 × 2 × 4 mm. Statistical parametric maps were calculated by means of SPM95 software (Wellcome Department of Cognitive Neurology, London, England; available at: http://www.fil.ion.ucl.ac.uk/spm/). The voxel values were normalized by proportional scaling to the global activity. Linear contrasts were used to estimate the differences in ratio-normalized rCMRglc, before and after PVE correction, between both patient groups and their respective controls and between the 2 patient groups.

Regions of interest (ROIs) were used to calculate changes in absolute CMRglc caused by PVE and group differences in regional brain volumes. The ROIs were located at the maxima of significant differences in the statistical parametric maps between the patients with AD + VS and their healthy controls. The ROIs consisted of 5-mm-radius cylinders and a volume of 1.2 cm³. Paired t tests were used to calculate whether differences in CMRglc in each ROI before and after PVE correction were significant. Unpaired t tests were used to compute whether group differences in regional brain volume were significant.

Before atrophy correction, compared with their healthy control subjects, the patients with AD + VS showed areas of reduced rCMRglc bilaterally in primary and association visual cortex, posterior cingulate and cuneus, and parietal and posterior temporal areas. There were no group differences in frontal regions, in medial and inferior temporal regions, or in subcortical structures. After atrophy correction, the regions with significant hypometabolism remained the same as before correction (P<.01) (Figure 1), despite larger PVE corrections in the posterior cortex in patients with AD + VS than in control subjects. The mean PVE correction in global CMRglc in the patients with AD + VS was significantly greater (P<.001) than in the healthy subjects. The increase for the AD + VS group was 19.2% (range, 5.1%-31.9%) and 12.5% (range, 6.5%-19.7%) for their healthy control group. These numbers are comparable to the increases in mean global CMRglc PVE correction for the patients with typical AD and their control subjects, 19.4% and 11.9%, respectively.¹⁷ Locations and z values for the peaks of the significant differences between the AD + VS group and the healthy controls, before and after PVE correction, are given in Table 2. Although these local maxima after atrophy correction are located more dorsally than in the uncorrected data, their anatomic position remains in the occipital and parietal lobes. Similar results, before and after atrophy correction, were obtained when the AD + VS patient group was compared with the healthy control group corresponding to the typical AD group (data not shown).

The original report by Pietrini et al¹⁴ showed hypometabolism in the sensorimotor cortex of the patients with AD + VS compared with healthy control subjects. Because we did not have MR images for all subjects superior to 40 mm above the anterior commissure–posterior commissure plane, we could not replicate this result here because the statistical parametric map was limited to brain regions where MR imaging data were available.

Each patient group had a significantly smaller total brain volume (normalized to cranial volume) than did the healthy control subjects (mean ± SD, AD + VS, 0.730 ± 0.019; AD, 0.743 ± 0.033; healthy controls, 0.799 ± 0.031; P<.001). The 2 patient groups were not significantly different from each other (P = .27).

Regional CMRglc values were calculated by means of ROIs to evaluate the percentage change in CMRglc caused by PVE correction. The ROIs were centered at the local maxima of significant differences between the AD + VS group and their healthy control group. In addition, 2 ROIs (in the frontal cortex) were located at the maxima of significant hypometabolism of the typical AD group compared with the AD + VS group. These changes are given in Table 3. In both the AD + VS patient group and the healthy control group, there were significant increases in rCMRglc because of PVE correction (P<.05).

We calculated the proportion of brain tissue in the ROIs defined above (Table 4). The proportion of tissue in each ROI was significantly different (P<.05) between the AD + VS group and the healthy control group, except in the left medial frontal cortical ROI. The correlation coefficient between the increase in rCMRglc caused by atrophy correction (Table 3) and the percentage of cerebrospinal fluid (defined as 100 minus percentage of brain tissue) in the ROI is 0.93 (P<.001).

The AD + VS group was significantly hypometabolic compared with the typical AD group in primary and extrastriate visual areas (Figure 2), as previously reported in other studies^{14 - 15} that did not correct for atrophy. After correction for atrophy, the AD + VS areas remained hypometabolic relative to typical AD in the visual association areas. The location of the hypometabolism maxima of the AD + VS group compared with the AD group is given in Table 5.

Regions in the frontal cortex were hypometabolic in the patients with AD compared with the AD + VS group (Figure 3) before atrophy correction and remained hypometabolic after atrophy correction. The location of the maxima for hypometabolism of the AD group compared with the AD + VS group is given in Table 6.

In this article we have examined differences in rCMRglc between patients with AD + VS and age-matched healthy control subjects, with and without PVE correction. Previous studies have applied this correction to typical AD patient groups.^{17 ,27 - 28} Here, we studied a distinct AD subgroup with atypical clinical manifestations to examine the effects of brain atrophy on calculated rCMRglc. Values of rCMRglc before PVE correction are per unit volume in a PET image, whereas after PVE correction the values are per unit volume of tissue. The differences in rCMRglc between patients and controls are focused in the posterior cortical regions and are consistent with previous PET studies.^{14 - 15} Thus, the increased brain atrophy in the hypometabolic areas does not account for the measured decrease in rCMRglc in the AD + VS patient group compared with the control group. These PET results are consistent with the hypothesis that glucose hypometabolism measured in AD represents an in vivo biochemical index of neuronal/synaptic dysfunction.²⁹ This supports the validity of the PET fluorodeoxyglucose method to assess synaptic viability in the brain during different stages of dementia severity and in relation to brain cognitive and sensory stimulation or pharmacologic modulation.^{30 - 32}

As was true for patients with typical AD,¹⁷ the increase in rCMRglc caused by PVE correction is larger in the AD + VS patient group than in the healthy control group because of the greater brain atrophy in the AD + VS patient group. In addition, the increases in rCMRglc are localized in regions of greatest brain atrophy.

We found that, before PVE correction, the AD + VS group was hypometabolic in posterior cortex when compared with the AD group, whereas patients with AD were hypometabolic in frontal cortex when compared with the AD + VS group. Correcting for brain atrophy did not alter this pattern of differences. Our results show that this subgroup of patients with AD with early and prominent visual dysfunction have a different pattern of rCMRglc hypometabolism from that observed in patients with typical AD after the effects of atrophy caused by neuronal loss have been accounted for by application of PVE correction. Therefore, this difference in the pattern of hypometabolism is not caused merely by different brain atrophy patterns, but rather reflects different patterns of neuronal/synaptic dysfunction caused by distinct developments of the AD neuropathologic process.

Various technical factors can affect our results. Perhaps the primary one is the relative simplicity of the algorithm used for atrophy correction. Correcting for PVE is an active area of research,³³ but at present there is no consensus on the most appropriate method. All atrophy correction algorithms assume uniform loss within gray and white matter compartments, thus ignoring differential neuronal loss in the different cortical laminas. In AD, neuronal loss is greatest in layers 3 and 5.³ The effect of this type of nonuniform neuronal loss on PVE correction is unknown.

Note that the differences observed between the AD + VS and the typical AD patient groups are unlikely to be due to use of different PET cameras. A comparison was made between each patient group and both healthy control groups to see whether differences in PET camera resolution would affect the statistical results. The findings (not shown) obtained with either healthy control group were similar, and thus we believe that the differences in rCMRglc between the 2 patient groups are not confounded by the use of different PET cameras. Likewise, the use of different MR imagers also is unlikely to change the PVE correction factors. The degree of global atrophy in the AD + VS group compared with the typical AD group was not significantly different. The increase caused by PVE correction was the same for both patient groups, as would be expected for similar degrees of brain atrophy, thus indicating that the different resolution of the MR imager is not affecting the correction coefficients.

In summary, hypometabolism measured by PET in patients with AD + VS represents a real decrease in metabolism per gram of tissue and is not an artificial decrease caused solely by increased brain atrophy.

Accepted for publication July 11, 2000.

This work was supported by the National Institute on Aging Intramural Research Program, Bethesda, Md.

Corresponding author and reprints: Barry Horwitz, PhD, Language Section, National Institute on Deafness and Other Communication Disorders, NIH, Bldg 10, Room 6C420, Bethesda, MD 20892 (e-mail: horwitz@helix.nih.gov).