Twenty AD patients (mean age and standard deviation: 63.1 ± 6.9 yrs) with a Mini-Mental State Examination (MMSE) [³⁵] score of on average 21.9 ± 5.6, 20 patients with FTD (age: 60.7 ± 9.9 yrs; MMSE: 23.1 ± 5.6) and 21 cognitively normal (CN) subjects (age: 61.9 ± 9.6 yrs; MMSE: 29.6 ± 0.5) were included in this cross-sectional MRI study. A summary of the subject demographics and relevant clinical information are listed in Table 1. The patients with FTD and AD were recruited from the Memory and Aging Center of the University of California, San Francisco. All patients were diagnosed based on information obtained from an extensive clinical history and physical examination. The MR images were used to rule out other major neuropathologies such as tumors, strokes, or inflammation but not to diagnose dementia. The subjects were included in the study if they were between 30–80 years old and without history of brain trauma, brain tumor, stroke, epilepsy, alcoholism, psychiatric illness, or other systemic diseases that affect brain function. FTD was diagnosed according to the consensus criteria established by Neary et al. [³⁶]. All FTD patients were diagnosed with the frontal variant subtype, two of which had combined motor neuron-related symptoms. AD patients were diagnosed according to the criteria of the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer's Disease and Related Disorders Association (NINCDS/ADRDA) [³⁷]. All subjects received a standard battery of neuropsychological tests including assessment of global cognitive impairment using MMSE and global functional impairment using the Clinical Dementia Rating (CDR) scale [³⁸]. Fifty-seven out of all 61subjects had blood drawn for APOE genotyping. Reliable information about the age of onset of symptoms was available from 12 out of 20 AD patients and from all 20 FTD patients. Because it is not unusual for subjects in this age group to have WM signal hyperintensities (WMSH) on MRI, subjects with WMSH were included. An experienced radiologist reviewed all MRI data, and the scores of WMSH were used as covariates in the analysis. The severity of WMSH was classified as mild (deep white matter lesions ≤3 mm, and periventricular hyperintensities <5 mm thickness), moderate (deep white matter lesions between 4–10 mm, or periventricular hyperintensities between 6–10 mm thickness), or severe (deep white matter lesions >10 mm, or periventricular hyperintensities >10 mm thickness), according to the Scheltens' rating scale [³⁹]. All subjects or their legal guardians gave written informed consent before participating in the study, which was approved by the Committees of Human Research at the University of California and the VA Medical Center at San Francisco.

All scans were preformed on a 4 Tesla (Bruker/Siemens) MRI system with a single housing birdcage transmit and 8-channel receive coil. T1-weighted images were obtained using a 3D volumetric magnetization prepared rapid gradient echo (MPRAGE) sequence with TR/TE/TI = 2300/3/950 ms, 7-degree flip angle, 1.0 × 1.0 × 1.0 mm³ resolution, 157 continuous sagittal slices. In addition, FLAIR (fluid attenuated inversion recovery) images with timing TR/TE/TI = 5000/355/1900 ms were acquired to facilitate the evaluation of WMSH. Perfusion images were acquired using a continuous arterial spin labeling (cASL) sequence [⁴⁰] with a single-shot echo-planar imaging (EPI) part to map the perfusion signal. cASL-MRI was performed with TR/TE = 5200/9 ms with 2-second long labeling pulses and a one-second postlabeling delay. Sixteen slices with 5 mm slice thickness and 1.2 mm interslice gap, 3.75 × 3.75 mm² in-plane resolution were acquired. DTI was acquired based on a dual spin-echo EPI sequence supplemented with two-fold parallel imaging acceleration (GRAPPA) [⁴¹] to reduce susceptibility distortions. Other imaging parameters were TR/TE = 6000/77 ms, field of view 256 × 224 cm, 128 × 112 matrix size, yielding 2 × 2 mm² in-plane resolution, 40 slices each 3 mm thick. One reference image (b = 0) and six diffusion-weighted images (b = 800 s/mm², along 6 noncollinear directions) were acquired.

The assessment of brain volume changes were performed using SPM8 software (http://www.fil.ion.ucl.ac.uk/spm/software/spm8/) based on an “optimized” VBM procedure described by Ashburner and Friston [⁴²]. The procedure included several steps. (1) Tissue segmentation: An expectation maximization segmentation (EMS) algorithm [⁴³] was applied to obtain probabilistic maps of GM, WM, and CSF from the T1-weighted MRI data. (2) Spatial normalization: first, customized GM and WM prior images were created by transforming GM and WM probabilistic maps of all subjects into the standard MNI (Montreal Neurological Institute) space [⁴²]. The segmented GM probabilistic maps in their native spaces were then spatially normalized again to the customized GM prior image using a nonlinear transformation with 16 interactions. (3) Jacobian modulation: the spatially normalized GM images were multiplied by the Jacobian determinants of the transformation (modulation) to obtain volume differences. (4) Smoothing: the modulated GM images were smoothed with an 8 mm full-width-at-half-maximum (FWHM) isotropic Gaussian kernel to reduce variations from misregistrations and to perform voxelwise image statistics.

The assessment of perfusion changes included the following steps. (1) Cerebral blood flow (CBF) image calculation: a perfusion weighted (PWI) image was created by pairwise subtraction of coregistered labeled from unlabeled ASL images. The PWI images were then scaled to obtain a quantitative CBF image based on a single compartment perfusion model [⁴⁴]. (2) Intermodality coregistration: ALS perfusion and the corresponding T2- and T1-weighted image were coregistered using an affine alignment to establish anatomical correspondence between CBF and segmented GM images and to compute partial volume-corrected CBF in GM. (3) Partial volume correction: partial volume correction of CBF images was performed by rescaling CBF in each voxel proportionately to the GM and WM content, assuming that perfusion of white matter is only 25% of that of GM, as detailed by Du et al. [¹⁶]. (4) Spatial normalization: the partial volume-corrected CBF images in GM were spatially normalized to the customized GM prior image that was created from the previous processing of VBM, using the same nonlinear transformation and smoothing parameters.

The assessment of white matter integrity was performed based on DTI and computation of fractional anisotropy (FA) maps, using the dTV.II software [⁴⁵] and Volume-one software package (URL: http://www.ut-radiology.umin.jp/people/masutani/dTV.htm), supplemented by automatic image denoising and eddy-current corrections. SPM8 software was used for voxelwise analysis of the FA images as outlined in detail elsewhere [²⁵]. In brief, the FA images in the native space underwent the following procedures. (1) Creation of a customized FA template: an averaged FA image was first created from all subjects' FA images that initially transformed to the EPI-derived MNI template in SPM. This averaged FA image was further co-registered to the WM prior image using affine alignment and manual adjustment [⁴⁶] to archive a customized FA template that accurately correspond to the anatomical information in the WM prior image. (2) Spatial normalization and smoothing: the FA image of each subject in the native space was recursively normalized to the customized FA template using the same nonlinear transformation and smoothing parameters that were applied in the previous VBM procedure.

Paired group differences were evaluated voxel-by-voxel using a general linear model with diagnosis as main contrast and age, sex, and the years of education as covariates. In addition, total intracranial volume (TIV) was used as a covariate in test for volume differences and global mean perfusion was used as a covariate in test for perfusion differences. A threshold of at least 85% GM volume fraction in voxels was applied to restrict GM volume and perfusion analyses to voxels containing predominantly GM. Similarly, a threshold of at least 80% WM volume fraction was applied to restrict FA analysis to voxels containing predominantly WM. The statistical significance for main effects of abnormalities, which was only the prestep for joint analyses, was set to an uncorrected voxel-level P value of  .001.

To test if FTD and AD differ with respect to WM and GM abnormalities, we started by determining the number of voxels representing “abnormal” values for each patient and each MRI modality. This was conducted by recording voxelwise differences between an individual's image value (GM volume, GM perfusion, or WM FA) and the respective mean value of the control group. The difference was then normalized to the standard error of the control mean value to express abnormality as a T-score [⁴⁷]. High T-scores represent high abnormalities beyond the normal ranges. Next, the number of “abnormal” voxels above a T-score threshold was recorded and normalized to the total number of GM or WM voxels to account for each patient and each modality to quantify the extent of brain abnormality. We termed the number of normalized “abnormal” voxels load. Differences in loads between AD and FTD patients were evaluated statistically via permutation test with 1000-fold completely random resampling of the diagnostic labels (AD and FTD), using the R project (http://www.r-project.org/). The statistics on the loads were evaluated for T-score values ranging from −2 (indicating a mild abnormality level) to −6 (indicating a severe abnormality level). Using the permutation test, we also assessed whether the load from a specific MRI modality (e.g., WM FA) relative to another (e.g., GM loss) modality, termed conditional load, differed within and across the diagnosis groups. The level of significance for permutation tests was set at P = .05.

As shown in Table 1, there were no significant differences in age, sex, years of education, and severity of WMSH between each patient group (AD or FTD) and controls. AD and FTD patients had significant lower MMSE (P < .001, by ANOVA test) scores compared with controls, as expected. Furthermore, AD patients had a greater proportion of APOE-ε4 carriers than CN (P = .04 by χ² test) but the FTD patients did not (P = .35 by χ² test) when compared to CN. AD and FTD patients did not differ significantly with respect to age, sex, years of education, MMSE, CDR, age of onset, symptom duration, and severity of WMSH. To avoid further reductions in sample size due to the missing values, we did not perform the joint analysis including the age of onset or the APOE-ε4 genotyping as covariates across all MRI modalities, although symptom duration was associated with GM volume loss within the FTD group.

Figures 1(a)–1(d) depict the regional distributions of GM loss (in warm color), GM hypoperfusion (in green color), and reduced WM FA (in blue color) in AD and FTD, compared to CN, respectively, as well as the direct comparisons between AD and FTD, based on voxel-wise tests prior to the joint analysis. For better visualization of regional relations, the various distributions are overlaid on each other in Figure 1(d).

Compared to CN, AD patients showed widespread GM loss in bilateral parietal and temporal lobes. The left temporoparietal lobes had the most prominent GM loss. Other regions of GM loss in AD included the posterior cingulate gyrus, thalamus, and bilateral occipital lobes. Compared to CN, FTD patients showed GM loss predominantly in bilateral frontal and temporal lobes. The right frontoinsular gyrus showed the most prominent GM loss. Other regions of GM loss included limbic lobes such as bilateral anterior cingulate gyrus, uncus, subcortical nuclei including the bilateral caudate and the thalamus, and the lateral parietal lobes. Comparing FTD and AD directly, patients with AD showed more GM loss than FTD in bilateral occipital gyrus, left precuneus whereas FTD showed more GM loss in bilateral frontal lobes, including the orbital gyrus, inferior and medial frontal gyrus, and anterior cingulate gyrus.

Regarding perfusion, AD patients showed reductions relative to CN in bilateral temporoparietal lobes, including superior temporal gyrus, precuneus and posterior cingulate gyrus. Hypoperfusion in AD was most pronounced in the left temporal gyrus. Compared to CN, FTD patients showed reduced perfusion in bilateral frontal lobes, including inferior, medial, and superior frontal gyrus, anterior cingulate gyrus, and thalamus. Hypoperfusion in FTD was most pronounced in the right inferior frontal gyrus. Compared to FTD, AD patients showed significant hypoperfusion in the left superior temporal gyrus, claustrum; whereas compared to AD, FTD patients had significant hypoperfusion in right superior, middle, and bilateral medial frontal gyrus.

Regarding WM FA, AD patients had FA reductions relative to CN bilaterally in WM regions in parietal, temporal, and some frontal lobe regions. The periventricular deep WM, posterior corpus callosum and the left posterior cingulum exhibited the most prominent FA reductions. In contrast, FTD patients had widespread FA reductions relative to CN bilaterally in frontal and temporal lobes, and the anterior corpus callosum, and bilateral anterior cingulum were prominently involved. Compared to AD, FTD patients had lower FA values bilaterally in frontal deep WM, anterior corpus callosum and bilateral anterior cingulum, whereas AD patients showed no region with significantly lower FA values when compared to the FTD group.

We first tested if loads differed between AD and FTD. Figure 2(a) displays the loads of GM loss (2A-I), GM hypoperfusion (2A-II), and WM FA (2A-III) reduction, respectively, in AD and FTD over a range of T-score levels (deviation from normal). The significance of differences in the loads between AD and FTD as a function of T-scores is plotted in Figure 2(b), and separately for each type (GM Vol, GM Perf, WM FA) of load. Note, the loads and the P values are plotted on a logarithmic scale and increasing negative T-scores indicate increasing deviation from normal values. This demonstrates that there is a significantly greater load of WM FA reduction in FTD compared to AD at mild abnormality levels (up to T-scores of about −2.3) while the significance gradually vanishes at more severe abnormality levels. There is a trend (P = .07 to  .1) towards more GM loss in FTD compared to AD at moderate to severe abnormality levels (T-scores < −4). However, the load of GM hypoperfusion does not differ significantly between AD and FTD across the range of T-scores.

Figure 3 shows the conditional loads (the load in one MRI modality relative to another) over a range of abnormality levels (T-scores) separately for AD (Figure 3(a)) and for FTD (Figure 3(b)). Note, the plots in Figures 3(a) and 3(b) are on a logarithmic scale. Accordingly, a conditional load larger than 1 indicates that the load of a particular modality is greater than the load of the reference modality, whereas a value smaller than 1 indicates that the load of the reference is greater than particular modality. The extent to which differences in one load (on particular modality) relative to another (on reference modality) are above chance is depicted in Figures 3(c) and 3(d) for AD and FTD, respectively. Note again, P-values in Figures 3(c) and 3(d) are plotted on a logarithmic scale of the base-10 logarithm. The loads of GM volume loss relative to GM hypoperfusion as a function of severity (T-scores) are displayed in brown lines, reduced WM FA relative to GM hypoperfusion in purple lines, and GM volume loss relative to WM FA in red lines.

Figures 3(a) and 3(c) show that in AD, the loads of the different modalities are not significantly different compared to each other across the level of abnormalities (T-scores) levels. In FTD, by contrast (Figures 3(b) and 3(d)), the load of GM volume loss as well as the load of reduced WM FA are each significantly greater relative to the load of GM hypoperfusion at lower levels of abnormality (up to T-scores = −4). However, as abnormality increases (T-scores < −4), the significance of the difference in the load of GM volume loss or WM FA reductions relative to GM hypoperfusion gradually disappears. There are no significant differences between the load of GM loss and the load of WM FA reductions across all ranges of abnormalities levels in FTD patients.

Finally, we also tested if AD and FTD differed with regard to their conditional loads from a particular modality relative to that from reference modality but found no significant difference between the groups (P > .1) across all ranges of the abnormalities (T-score) levels.