Comparison of Regional Brain Volume and Glucose Metabolism Between Patients with Mild Dementia with Lewy Bodies and Those with Mild Alzheimer's Disease

MRI Procedure

The MR scanner was a 1.5-T Signa Horizon (GE Healthcare). Three-dimensional spoiled gradient echo imaging (repetition time, 14 ms; echo time, 3 ms; flip angle, 20°; thickness, 1.5 mm; 124 slices) was performed. MR images were obtained 0–28 d before the PET examinations.

PET Procedure

The PET procedure is described in detail elsewhere (21). In brief, ¹⁸F-FDG PET images were obtained using a Headtome IV scanner (Shimadzu Corp.). Before the PET scans, all subjects received MR examinations for PET positioning. All subjects fasted for at least 4 h before PET. Subjects were studied in the resting condition, with eyes closed but without ears plugs. After obtaining a transmission scan, a 12-min emission scan was started 60 min after intravenous injection of 185–370 MBq of ¹⁸F-FDG. Data were collected in 128 × 128 matrices. The slice interval was 6.5 mm when the z-motion mode was used.

Data Analysis

Volumetry and Relative Glucose Metabolism.

Volumetry and measuring ¹⁸F-FDG accumulation were performed with our modified automatic volumetric measurement of segmented brain images system (AVSIS) (26). In brief, before this analysis, voxels of interest (VOIs) templates for intracranial, hippocampi, striata, and occipital lobes were prepared. Each regional VOI template was produced on a digital phantom: the Simulated Brain Database ([SBD] http://www.bic.mni.mcgill.ca/brainweb/) in standard Montreal Neurological Institute (MNI) space, delineating the contours of each structure manually. The occipital lobe's VOI template included Brodmann areas 17, 18, and 19. The process of modified AVSIS is as follows: first, each individual MR image and PET image were coregistered, and then each subject's MR image was segmented into GM, white matter (WM), and cerebrospinal fluid (CSF) with the SPM5 segmentation program. The GM template image derived from the SBD was then spatially transformed into the individual subject's GM image, and a normalization parameter was produced. This is the same as a reversed parameter produced in anatomic normalization of individual brain to standard brain. Using this parameter, the intracranial, hippocampal, striatal, and occipital VOI templates were transformed to the individual subject's space. The total intracranial volume (TIV) area was adjusted with an image derived from the segmented GM, WM, and CSF images. The segmented images were calculated as follows: WM and GM areas were calculated with the voxels in the TIV VOI template, GM volumes in the hippocampi and striata were calculated using the transformed hippocampal and striatal VOI templates, whereas volumes of GM and WM in the occipital lobes were calculated using the transformed occipital VOI template for each subject. Each regional relative volume was calculated by normalization relative to the TIV. Next, hippocampal, striatal, and occipital ¹⁸F-FDG uptake counts were calculated by determining the regions of coregistered ¹⁸F-FDG images on each extracted area using the described volumetric procedure (Fig. 1). Because pontine and sensorimotor glucose metabolisms are relatively preserved in DLB (1), we used these regions for reference regions for calculating the relative regional cerebral glucose metabolism. Beforehand, ¹⁸F-FDG PET images were transformed into the MNI space with SPM5, and then pontine and sensorimotor ¹⁸F-FDG uptake counts were determined in the ROIs placed on the pons below 20–22 mm and the sensorimotor area over 38–42 mm from the anterior commissure–posterior commissure plane on the transformed MNI space. Then the extracted regional counts were normalized to the pontine counts or the sensorimotor counts and shown as relative glucose metabolism. Additionally, we calculated the affected region ratio of DLB and AD: for MRI, striatal volume/hippocampal volume (Str/Hipp), and for PET, occipital/hippocampal counts (Occ/Hipp). Differences between groups were assessed using ANOVA followed by the post hoc Scheffé test to correct for multiple comparisons. All statistical tests were regarded as significant at P < 0.05.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Algorithm for calculating regional structural volume and glucose metabolism An example of calculating hippocampal volume and metabolism is shown. Hippocampal VOI template is transformed to individual brain and, using this VOI template on individual segmented MR image, hippocampi are extracted and hippocampal volume is calculated. Then, using extracted hippocampi as a hippocampal VOI for individual coregistered PET image, hippocampal counts representing glucose metabolism are determined.

Voxel-Based Comparison

In the MRI VBM study, the gray matter densities of the left parahippocampal gyrus and the hippocampal complex in the AD group were significantly lower than those in the NC group (P < 0.05, corrected), whereas the gray matter densities of the left putamen and caudate head in the DLB group were significantly lower than those in the NC group (P < 0.05, corrected) (Table 1; Fig. 2A).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

Voxels with significantly lower GM density or glucose metabolism are shown in axial plane of standard brain. (A) Left caudate and putamen GM densities in DLB group were significantly lower than those of NC group (upper left row). Left hippocampal and parahippocampal GM densities in AD group were significantly lower than those of NC group (lower left row). Bilateral occipital, temporoparietal, and frontal glucose metabolisms in DLB group were significantly lower than those in NC group (upper right row), and parietal and posterior cingulated metabolisms were also decreased (not shown). Bilateral posterior cingulate glucose metabolism in AD group was significantly lower than that in NC group (lower right row) (P < 0.05, corrected). (B) Left caudate GM densities in DLB group were significantly lower than those of AD group (upper row). Bilateral occipital glucose metabolism was significantly decreased in DLB compared with that in AD (lower row), and glucose metabolism in bilateral frontal, parietotemporal association cortices and left posterior cingulate gyrus, was also significantly reduced in DLB (not shown) (P < 0.001, uncorrected).

PET SPM analysis demonstrated that a metabolic reduction was present in the bilateral posterior cingulate gyri in the AD group compared with that of the NC group (P < 0.05, corrected). Furthermore, in the DLB group, occipital, temporal, parietal, and frontal metabolisms were significantly reduced compared with those in the NC group (P < 0.05, corrected) (Table 1; Fig. 2A).

In direct comparison between the DLB and AD groups, VBM demonstrated a reduced left caudate volume in the DLB group compared with that in the AD group but did not show significantly reduced regions in the AD group compared with those in the DLB group (Table 1; Fig. 2B) (P < 0.001, uncorrected). SPM analysis for PET showed significantly reduced metabolism in the bilateral frontal, parietotemporal association cortices, left posterior cingulate gyrus, and bilateral occipital lobes in the DLB group compared with that in the AD group but did not show significantly reduced regions in the AD group compared with that in the DLB group (Table 1; Fig. 2B) (P < 0.001, uncorrected).

Volumetry and Relative Glucose Metabolism

Mean regional volumes in each group are presented in Table 2. TIV, whole brain (WB), and hippocampal volumes in the AD group were significantly smaller than those in the NC group. The AD group's WB volume normalized to TIV did not differ significantly from that of the NC group, but the AD group's hippocampal volume normalized to TIV was significantly smaller than that in the NC group. There were no significant differences in occipital volumes among the DLB, AD, and NC groups for either the absolute or the relative values normalized to TIV. Absolute and relative striatal volumes (Str/TIV) in the DLB group were significantly smaller than those in the NC group, and striatal volumes (Str/TIV) in the DLB group were significantly smaller than those in the AD group.

Occipital glucose metabolism normalized to that of the pons or sensorimotor area was significantly reduced in the DLB compared with that of the AD and NC groups. Hippocampal metabolism normalized to that of the pons or sensorimotor area was also significantly reduced in the AD group compared with that of the DLB and NC groups (Table 2).

The affected region ratios of MRI and PET are shown in Table 3. The MRI Str/Hipp ratio in the AD group was significantly higher than that of the DLB and NC groups, and the PET Occ/Hipp ratio in the DLB group was significantly lower than that of the AD and NC groups.

ROC Analysis for Discriminating DLB from AD

Accuracies and areas under the ROC curve for discriminating DLB from AD by each value are shown in Table 4. The areas under the ROC curve and the accuracy for PET Occ/Hipp were the highest among those of other values, though there were no significant differences in the partial area index for ROC curves among these parameters. Figure 3 shows the ROC curves of the MRI Hipp/TIV, Str/TIV, PET Hipp/pons, and Occ/pons and the affected regional ratio of MRI Str/Hipp and PET Occ/Hipp for discriminating DLB from AD.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

ROC curves for PET and MRI diagnostic performances in discriminating DLB from AD. Area under ROC curve of Occ/Hipp by PET (area under the curve [A_z] = 0.941) was the largest among all values. (A) Relative regional volumes of hippocampi (Hipp/TIV) and striata (Str/TIV) by MRI. (B) Relative regional glucose metabolism in hippocampi (Hipp/pons) and occipital lobes (Occ/pons) by ¹⁸F-FDG PET. (C) Affected regional ratio obtained by PET (Occ/Hipp) and MRI (Str/Hipp). TPF = true-positive fraction = sensitivity; FPF = false-positive fraction = 1 − specificity.

Voxel-Based Analysis

VBM analysis showed significant striate atrophy in DLB and hippocampal atrophy in AD, even in the mild stages of these diseases, though only on the left side. SPM analysis of ¹⁸F-FDG PET findings revealed significant glucose metabolic reductions in the occipital, temporoparietal, and frontal association cortices in the DLB. Thus, even in the mild stage of DLB, glucose metabolic disturbances are widespread compared with those of healthy aged subjects. SPM analysis of mild AD brains also demonstrated posterior cingulate glucose metabolic reduction, an observation that supports previous reports (10,15,16). Our study demonstrated in the DLB brain not only previously identified posterior hypometabolism, including the posterior cingulate and occipital regions, but also frontal hypometabolism. The metabolic reduction pattern of DLB is relatively similar to that of AD, with association of occipital metabolic reduction; however, the degree of metabolic reduction in those areas is larger than that of AD, which has been reported in moderate DLB (1). Our study showed the same pathophysiology even in mild DLB. We emphasized that mild DLB shows abnormalities involving vast portions of the entire cortex compared with mild AD. Thus, it appears that the process of neurodegeneration may be more advanced in the mild DLB group compared with the mild AD group, in spite of comparable neuropsychologic results. There are 2 different aspects: (a) better cognitive reserve in the DLB group or (b) insufficient sensitivity of the applied neuropsychologic testing—MMSE to detect the true extent of neurodegeneration in DLB. In any case, it cannot be excluded that the observed findings may in part be due to different stages of disease in using MMSE scores.

Medial Temporal Lobe

The medial temporal area is histopathologically affected in the early stages of AD. This medial temporal change has been described in previous reports using MRI (7,8) and was confirmed in the present study of patients with very mild AD. Morphologic imaging by MRI clearly demonstrates macropathologic changes in mild AD. However, previous PET studies have shown that medial temporal glucose metabolism and blood flow are preserved in mild-to-moderate AD patients (11,28). Recently, Nestor et al. in 2003 (14) and Mosconi et al. in 2005 (15) suggested that hippocampal metabolism in patients with mild cognitive impairment and AD was reduced, based on results obtained using their technique of drawing ROIs on MR images, and described the voxel-based analysis using SPM as failing to show the hippocampal metabolic reduction. Herein, we verified their findings using our new VOI warping method combined with MRI. In mild AD patients, SPM did not demonstrate hippocampal metabolic reduction, whereas our method showed significantly reduced hippocampal metabolism. However, in mild DLB, neither SPM nor our method demonstrated this reduction in hippocampal metabolism. As to structural change, no hippocampal volume reduction was demonstrated in DLB using either SPM or our method. Therefore, in mild AD, but not mild DLB, both metabolic and structural alterations are present. These findings support previous results obtained by studying functional and morphologic changes separately. There was a 5% reduction in the Hipp/SM ratio in AD compared with NC but there was a 6% increase in this ratio when DLB was compared with NC (Table 1). It seems highly unlikely that DLB is associated with hippocampal hypermetabolism. This suggests that the increase in the ratio is actually due to a decrease in the denominator—that is, there is a degree of SM hypometabolism in DLB consequently due to a widespread metabolic reduction even in the mild DLB brain. Therefore, as the Hipp/pons ratio showed, hippocampal metabolism in mild DLB seems not increased but relatively preserved.

Striatum

In the striatum, dopamine transporter imaging is reportedly useful for distinguishing between DLB and AD because, in the former, dopamine transporter uptake in the striatum is low (29,30). Though low dopamine transporter uptake has been shown in the DLB striatum, to our knowledge, there are no reports indicating that striatal metabolism or perfusion is altered. Glucose metabolism in the striatum was reported to be preserved in DLB, based on measuring absolute glucose metabolism (1). Our study verified that relative striatal metabolism was also preserved in the mild DLB brain. We showed reduced striatal volume in DLB brains compared with those affected by AD and in NC subjects. Barber et al. reported that volumetric MRI analysis of the caudate nucleus does not discriminate among patients with DLB, AD, and vascular dementia (31). Almeida et al. also found that caudate atrophy is not region specific but, rather, is part of generalized brain volume loss (32). On the contrary, Cousins et al. showed that patients with DLB had significantly smaller ratios of putamen volume to TIV than both control subjects and patients with AD and, furthermore, that the AD patients did not differ from control subjects in any measure of the putamen volume (33). Our study supports their findings indicating that striatal volume in the DLB brain is significantly smaller than those in AD and NC brains. It is interesting that, in spite of volume loss of striatum, there is no glucose metabolic reduction in DLB. The pathophysiology of this situation is unclear. This may be due to upregulation of the low dopamine uptake.

Occipital Lobe

In the DLB brain, occipital glucose metabolism is affected despite the absence of morphologic change, though the metabolic and morphologic features have been studied separately. Gross structural changes in the occipital lobe reportedly do not occur in patients with mild-to-moderate DLB or AD, based on results obtained using MRI volumetric techniques (34). Our study demonstrates these findings in individual subjects. On the contrary, using VBM, Burton et al. reported that occipital atrophy in Parkinson disease dementia (PDD) was the only difference between patients with Parkinson's disease and a PDD group (35). Although PDD is thought to be in the same spectrum as DLB, no previous reports have demonstrated significant occipital atrophy in DLB. Our VBM study and volumetric study confirmed the latter findings. Burton et al. also noted that uncertainty remains with regard to the relationships between atrophy, neuropathologic changes, and perfusion deficits in Parkinson's disease with and without dementia (35). Our study confirmed morphologic and functional discrepancies in the same individual DLB brains.

Differential Diagnosis for AD and DLB

For routine clinical examination, conventional ROI placement or a voxel-based technique, identifying reduced occipital glucose metabolism, may be worthwhile because, as shown in the present study, these techniques are moderately useful for distinguishing mild DLB from mild AD. Moreover, our ROC analysis demonstrated that the PET Occ/Hipp value—which was amplified with combination of each DLB's and AD's affected value obtained with ¹⁸F-FDG PET—has higher accuracy than other values. Herein, we used software combining ¹⁸F-FDG PET and MRI with extraction of regional structures such as the hippocampus.