Patients
This was a retrospective study that used data obtained at a single medical centre.
All the subjects with AD were included in previously reported paper from our institute [14]. This previous study was approved by the National Center of Neurology and Psychiatry Ethics Committee for Clinical Research, and informed consent was obtained from all the subjects. For our study, public notification was applied according to the requirement of the ethical committee. Then, the ethical committee approved our retrospective study using preliminary obtained images.
At first, 20 of the PiB-positive patients who fulfil the criteria of probable AD with a high level of evidence of the AD pathophysiological process, with both atrophy of the medial temporal lobe and decrease of the brain perfusion by over two standard deviations based on the Z-scores, were observed as described in this paper [14]. The [11C]PiB amyloid PET results were confirmed by two board-certified nuclear medicine physicians. After reinvestigation of clinical data and MRI, three patients were excluded: one who was first diagnosed as DLB, another had infarction in the brain stem on the T2-weighted image, and the other had complication of brain amyloid angiopathy. The average lag time between PET and SPECT acquisition was 1.40 ± 2.33 (mean ± SD) months.
For DLB, from chart screening, among all accessible patients who underwent brain perfusion SPECT, those fulfil the probable DLB on the basis of the criteria proposed in the third consortium on DLB international workshop [15] were selected. Finally, 18 patients with probable DLB (M/F = 10:8; age, 73.9 ± 6.8 years) and 17 patients with AD (M/F = 6:11; age, 73.6 ± 8.9 years) were studied.
Brain perfusion SPECT
In advance, all patients received an intravenous line and an intravenous injection of 740 MBq [99mTc]ECD (Fujifilm RI Pharma, Tokyo, Japan) was administered while lying down with eyes closed in dark, quiet surroundings. The global CBF was noninvasively measured using graphic analysis as described previously [16–18], without blood sampling. The passage of tracer from the aortic arch to the brain was monitored for 100 s at 1-s intervals. Regions of interest (ROIs) were hand-drawn over the aortic arch (ROIaorta) and both brain hemispheres (ROIbrain). A hemispheric brain perfusion index (BPI) [16] was determined before the start of the initial back diffusion of the tracer from the brain to the blood as follows:
$$ \mathsf{B}\mathsf{P}\mathsf{I}=\mathsf{100}\times \mathsf{k}\mathsf{u}\frac{\mathsf{10}\times {\mathsf{ROI}}_{\mathsf{aorta}}\mathsf{size}}{{\mathsf{ROI}}_{\mathsf{brain}}\mathsf{size}}, $$
(1)
where ku is the unidirectional influx rate for the tracer from the blood to the brain, determined by the slope of the line in graphic analysis within the first 30 s after injection. Then, BPI (x) was converted to global CBF values (y) obtained by 133Xe inhalation SPECT studies (y = 2.60x + 19.8) [16].
Ten minutes later, SPECT imaging was performed on a two-head gamma camera and six-slice CT system (Symbia T6; Siemens, Erlangen, Germany) equipped with low-energy, high-resolution, and parallel-hole collimators. Ninety views were obtained continuously throughout 360° of rotation (4°/step, 128 × 128 matrix, zoom 1.45). The voxel size was 3.3 × 3.3 × 3.3 mm. To reconstruct the SPECT image, a combination of Fourier rebinning followed by ordered subset expectation maximization (iteration number 8 and subset 10) and a 7-mm full width at half maximum Gaussian filter was used. Attenuation correction was performed using the CT data and Chang’s method [19]. CT attenuation-corrected images were used for measuring global CBF, and Chang’s attenuation-corrected images were used for Z-score analysis because the database included in the software was reconstructed using Chang’s attenuation-correction methods. To calculate CBF and to correct for incomplete retention of [99mTc]ECD in the brain, the following linearization algorithm [20] of a curve-linear relationship between the brain activity and blood flow was applied:
$$ \mathsf{F}\mathsf{i}=\mathsf{F}\mathsf{r}\times \frac{\mathit{\mathsf{a}}\times \left(\raisebox{1ex}{$\mathsf{C}\mathsf{i}$}\!\left/ \!\raisebox{-1ex}{$\mathsf{C}\mathsf{r}$}\right.\right)}{\left[\mathsf{1}+\mathit{\mathsf{a}}-\left(\raisebox{1ex}{$\mathsf{C}\mathsf{i}$}\!\left/ \!\raisebox{-1ex}{$\mathsf{C}\mathsf{r}$}\right.\right)\right]}, $$
(2)
where Fi and Fr represent CBF values for a region I and a reference region, respectively, and Ci and Cr are the SPECT counts for the region i and the reference region, respectively. The cerebral hemisphere was used as the reference region, and global CBF obtained from the graphic analysis was substituted for Fr. The linearization factor a was set to 2.59, which was a proposed value by Friberg et al. [20].
Image preprocessing
Z-score maps of the obtained SPECT images were converted using the easy Z-score imaging system (eZIS) analysis software (Fujifilm RI Pharma Co., Ltd., Tokyo, Japan). It included spatial normalization parameters in statistical parametric mapping (SPM)2 (http://www.fil.ion.ucl.ac.uk/spm/) and a [99mTc]ECD brain template in the same space as the Montreal Neurological Institute (MNI) standard brain template. Normal databases are included in this software, and inter-institutional differences can be corrected. That is, correction can be made for data obtained in the institute where the database was built, using previously scanned Hoffman 3-D Brain Phantom™ data.
After the inter-institutional correction, specially normalized [99mTc]ECD SPECT images from each patient were compared with normal images from the age-matched database: ECD60-69y DB and ECD70y DB, using voxel-by-voxel Z-score analysis after pixel normalization to the global mean values [Z-score = ([control mean] − [individual value])/(control SD)] as previously reported by Minoshima et al. [21]. The VOIs in areas of significant perfusion reduction were included in this software; these areas were identified in patients with AD following a group comparison with cognitively healthy individuals [22]. Among these preset VOIs, we chose VOIs that were set within the bilateral posterior cingulate to the precunei area; we then used the border between the Cingulum_VOIs and the Precuneus_VOIs in the Automated Anatomical Labeling (AAL) atlas to split the VOIs into two parts, the bilateral PCG_AD_VOIs and the Precuneus_AD_VOIs (Fig. 1). We also assembled VOIs in the medial and lateral occipital areas in this atlas and constructed the Medial_Occipital_VOI and the Lateral_Occipital_VOI. Additionally, we constructed the Whole_Occipital_VOI as the sum of these two VOIs. Then, we summated the positive Z-scores within each VOI.
For the CIS, we first divided each value: the sum of all the positive Z-scores in the PCG_AD_VOI by that in the Precuneus_AD_VOI and named it as CISpreC. Second, in order to compare the value of cingulate preservation to the value of occipital hypoperfusion, we divided the PCG_AD_VOI by the Medial_Occipital_VOI, the Lateral_Occipital_VOI, or the Whole_Occipital_VOI. We named their three values as CISmedO, CISlatO, and CISwO; CISmedO = PCG_AD_VOI/Medial_Occipital_VOI, CISlatO = PCG_AD_VOI/Lateral_Occipital_VOI, and CISwO = PCG_AD_VOI/Whole_Occipital_VOI.
The area under the receiver operating characteristic (ROC) curve (AUC) was obtained by thresholding each of these values for all VOIs, CISpreC, CISmedO, CISlatO, and CISwO. Finally, the AUCs were statistically compared [23] (Table 2).
We also added group comparison between AD and DLB subjects using SPM2. We added the results as Additional file 1 (Fig. 2). VOIs used for analyses were integrated in the images.