Comparisons Between Alzheimer Disease, Frontotemporal Lobar ... : Topics in Magnetic Resonance Imaging (original) (raw)
Volumetric magnetic resonance imaging (MRI) has become increasingly important in the study of neurodegenerative dementias, particularly as an aid to diagnosis and as a biomarker of disease progression. Much research has concentrated on Alzheimer disease (AD) because this is the most common cause of dementia. It affects an estimated 4.5 million people in the United States and is growing in incidence and prevalence.1 Current projections have suggested that by 2050, the number of individuals with AD could range from 11.3 million to 16 million.1 The disease is characterized by the insidious onset of episodic memory deficits, which gradually spread to other cognitive domains and ultimately lead to global cognitive impairment. Behavioral changes are also not uncommon but usually occur later in the disease course. However, over the last decade, other dementias have been increasingly recognized. Frontotemporal lobar degeneration (FTLD) is the second most common dementia in the under 65 age group with a prevalence of approximately 15 per 100,000 in the 45 to 64 year age range.2,3 In contrast to AD, it is characterized by the insidious onset of behavioral and language deficits, with relatively preserved memory until later in the disease course. Three common syndromic variants of FTLD have been described: frontotemporal dementia (FTD), which has also been described as the frontal or behavioral variant because of the early alteration in personality and social conduct; and semantic dementia (SD) and progressive nonfluent aphasia (PNFA), that both involve early language deficits.4 Subjects with SD typically present with early impairments in semantic memory,5,6 often presenting with the complaint of loss of memory for words. They show impaired naming, word comprehension, and visual object recognition, yet still have relatively fluent speech. In contrast, subjects with PNFA present with complaints of speech dysfluency.7,8 They have distorted hesitant speech characterized by reduced phrase length and diminished use of grammatical elements. The FTD syndromic variant has been reported to be the most common, accounting for between approximately 60% to 80% of FTLD cases, with SD and PNFA only accounting for between 10% to 40% of cases.6,9,10
Diagnostic criteria have been defined for both AD and FTLD;4,11 however, the differential diagnosis between them can be very difficult in the early stages of the disease. Both are common in patients younger than 65 years old, have an insidious onset, and produce a progressive dementia syndrome that can, in some cases, have overlapping features including executive dysfunction, language impairment, and can cause alterations in behavior. In fact, recent clinicopathological studies have identified cases of FTLD with prominent early episodic memory loss,12 whereas others have suggested the existence of a frontal variant of AD.13 This overlap in syndromic features increases the need to have biomarkers to aid in the differentiation of these 2 syndromes, particularly with the availability of therapies for the treatment of AD. As a result, there has been interest in using MRI to better characterize these diseases and to aid in diagnosis.
Numerous studies have applied a region of interest (ROI) approach to studying patterns of brain atrophy in these diseases. This typically involves drawing around a structure of interest on a number of slices and then summing the area measurements and multiplying by the distance between slices (Cavalieri principle) to calculate a volume. Most studies measure volumes manually using a mouse-driven cursor and often use intensity thresholds to help in the identification of cerebrospinal fluid (CSF) from brain tissue, although more automated approaches have been developed. Studies have tended to focus mainly on the temporal lobes and have demonstrated atrophy particularly of the hippocampus and entorhinal cortex in subjects with AD.14-20 Hippocampal atrophy has also been shown to correlate to cognitive decline21 and have a sensitivity of 83% and specificity of 80% in the differentiation of AD from controls.20 Temporal lobe atrophy is also present in subjects with FTLD. Some studies have shown that the hippocampus and amygdala are smaller in AD than FTLD,22,23 although others have shown more severe atrophy of these structures in FTLD (particularly the SD variant).24-26 The SD variant of FTLD has particularly been shown to be associated with severe atrophy of the left temporal lobe. However, ROI studies have generally found poor discrimination between AD and FTLD.22,27 These regional measures also have a number of disadvantages. They are very labor intensive and time-consuming which limits the number of potential measurements that can be performed in a study. They are subject to wide interrater and intrarater variabilities, and the neuroanatomical boundaries and definitions vary between studies. They also require a priori decisions concerning which structures to assess and therefore do not allow an unbiased assessment of patterns of atrophy across the whole brain. There has therefore been an increasing need for more automated techniques that can assess changes across the whole brain. One such technique that will be described in detail is voxel-based morphometry (VBM).
VOXEL-BASED MORPHOMETRY
Voxel-based morphometry involves a voxel-wise statistical comparison of the local concentration of gray matter between 2 groups of subjects.28,29 It is automated and unbiased in that it looks throughout the whole brain and does not require any a priori assumptions concerning which structures to assess, giving it a significant advantage over ROI-based methods. It was developed as a modification of the statistical parametric mapping (SPM) methodology that was originally developed to analyze functional imaging data30 and has grown in popularity over the last 7 years because of the fact that it is relatively quick and easy to apply (Fig. 1). It has been used to assess patterns of atrophy in a number of diverse disorders, ranging from neurodegenerative diseases such as Huntington disease31 and dementia with Lewy Bodies,32 to other neurological conditions such as multiple sclerosis,33 migraine,34 and schizophrenia.35
Histogram showing the number of publications listed on Pubmed that have used VBM each year since 1998.
To perform accurate statistical analyses across sets of images from different individuals, the images need to first undergo a series of preprocessing steps (Fig. 2). The original method of VBM included 3 processes: spatial normalization, segmentation, and smoothing.29 The aim of spatial normalization is to transform all the subjects' data into the same stereotactic space. It is generally achieved by registering each image onto the Montreal Neurological Institute (MNI) template image,36 which is in approximate Talairach space.37 An initial affine 12-parameter transformation, matching the whole head, including scalp, is followed by a nonlinear transformation. This method does not attempt to match every cortical feature exactly but merely corrects for global brain shape differences. If the spatial normalization was exact, then all the segmented images would appear identical, and no significant differences would be detected. After normalization, the scans are segmented into gray matter, white matter, and CSF. The SPM algorithm uses a Bayesian approach where tissue classification is based upon both the voxel intensity within the image and a priori knowledge of the spatial distribution of these tissues in healthy subjects, derived from prior probability maps.29 The segmentation also incorporates an image intensity nonuniformity correction that appears to improve tissue classification.29 Finally, the images are smoothed by convolving with an isotropic Gaussian kernel.29,38 The intensity of each voxel is basically replaced by the weighted average of the surrounding voxels. The size of this region is defined by the size of the smoothing kernel. This makes the data conform more closely to the Gaussian field model that is an important assumption of SPM, renders the data more normally distributed, increasing the validity of the parametric tests, and reduces intersubject variability.29,39 It also helps to compensate for the inexact nature of the normalization.
Schematic diagram showing the standard VBM processing steps (in black). Possible optimizations to the process are shown in color and include the use of customized templates and priors (orange), gray matter optimization (red), masking (green), and modulation (blue). Processes are outlined by an oval, and images are outlined by a rectangle.
The smoothed data are then analyzed using SPM.30 This uses the general linear model to identify differences in tissue density between subject groups and generates SPMs showing regions of significant change. Standard parametric statistical tests (t tests and F tests) are performed at every voxel in the image and the significance of any differences is ascertained using the theory of Gaussian random fields. Because a large number of statistical tests are performed, it is important to correct for multiple comparisons to prevent the occurrence of false positives. The traditional multiple comparison correction within SPM is a modification of a Bonferroni correction, controlling the chance of any false positives over the entire volume (familywise error rate). However, this is a very harsh statistical criterion. It was originally developed for functional imaging studies in which each individual is sampled many times during a time series and may not be appropriate for structural data. A more lenient approach recently developed is the false discovery rate correction that controls the expected proportion of false positives among suprathreshold voxels (http://www.fil.ion.ucl.ac.uk).
Voxel-based Morphometry Optimizations
A series of optimizations to the original VBM preprocessing techniques have been described38 (Fig. 2):
- Customized templates and prior probability maps. Customized templates are often created to improve the normalization process. The standard MNI template is created from the brains of young control subjects. However, the brain structure of older subjects is likely to differ greatly from that in young controls and therefore may not register well to the MNI template. This in turn may affect subsequent processes such as the accuracy of gray matter segmentation. The solution to this problem is to create a customized template consisting of age-matched control subjects. Furthermore, systematic differences may be introduced if there is a difference in normalization accuracy between 2 groups being studied.40,41 This is likely to be a particular problem when there is extensive atrophy due to neurodegeneration, as in both AD and FTLD. Customized templates consisting of both control and disease subjects should minimize any normalization bias between the subject groups and allow any regions of change to be attributed to the disease process rather than registration accuracy. Similarly, the use of customized prior probability maps for the segmentation procedure will prevent any systematic bias in segmentation quality between controls and disease subjects. Errors in segmentation are introduced when there is mismatch in the registration with the prior probability maps.29 Ventricular enlargement may alter the value of the a priori knowledge of tissue distribution (priors). There is some evidence to suggest that the use of customized prior probability maps, based on controls and the disease group, could improve segmentation.32,42,43
- Masking. Segmentation errors could also be reduced by using expert manual editing to "clean-up" errors in the automated segmentation. This could be achieved by masking each image with a previously segmented brain region.38,43 The disadvantage however is that creating the mask segmentations is often time consuming. There has also been a suggestion that using brain extracted images as the input to normalization will improve the quality of the normalizations.44
- Gray matter optimization. Spatial normalization of each individual volume is based on matching the initial gray matter segmentation with the gray matter prior,38 whereas in standard VBM spatial normalization is performed on the volume proper. Therefore, each scan is registered to the template using a 12dof affine registration, the scan is segmented, and then, the gray matter image is normalized to the gray matter prior using a nonlinear normalization. The parameters from this normalization are then applied to normalize the original whole head image. This optimization aims to improve the quality of the normalizations and subsequent segmentations in the assessment of gray matter volume changes.
- Modulation. Modulation aims to correct for volume change during the spatial normalization step.38 For example, volume differences due to disease may be lost when an individual is matched onto a larger, nonatrophic template image. Without modulation, a brain which had atrophy affecting gray and white matter equally would appear to be identical to a normal brain after spatial normalization. To recover atrophy and preserve the volume of a particular tissue (gray or white matter or CSF), intensities within the image are divided by the Jacobian values from the registration. Regions that have undergone extensive expansion will consequently show reduced intensities, reflecting a reduction in density of tissue. In effect, an analysis of modulated data tests for regional differences in the absolute amount (volume) of gray matter, as opposed to regional difference in concentration of gray matter in unmodulated data.
A recent study compared a number of different VBM processing techniques and showed that change in the image processing chain noticeably influences the results of comparisons between AD subjects and controls.43 They showed that the gray matter optimization procedure and the use of customized templates and priors (consisting of both disease and control scans) improved the plausibility of the results. The addition of a masking step did not significantly affect group comparisons.
However, there are a number of limitations inherent to the technique of VBM that should be considered. Voxel-based morphometry is greatly affected by variability. Variability among individuals both due to heterogeneity within the sample and errors introduced by the preprocessing steps reduce sensitivity for the detection of group differences. For example, the power to detect a difference in a particular region is particularly dependent on the accuracy of the normalization.45 Voxel-based morphometry cannot differentiate changes in tissue content from local misregistration of images. This may be a particular problem for small structures such as the amygdala or hippocampus that not only have a complex shape but may also be highly variable between subjects.46 A number of the optimizations described above, such as the use of customized templates, attempt to minimize the effects of misregistration. Other voxel-based techniques, such as tensor-based morphometry,47,48 that aim to warp images together exactly and then assess volumetric changes from the deformation fields, may also address some of the issues surrounding inadequate normalization. The segmentation step provides an additional source of error: the misclassification of tissue is especially likely in atrophic brains, both because there is a greater potential for partial volume effects between gray matter and CSF, and because tissue pathology may be associated with reduced gray/white matter contrast.46 Errors in segmentation can also occur because of displacement of tissues. This issue will be discussed in greater detail later in the review. In addition, the smoothing step involves a trade-off: whereas high levels of smoothing increase the ability of VBM to detect gray matter differences by reducing the variance, excess smoothing diminishes its ability to accurately localize change.
There are often significant variations across studies both in the VBM preprocessing steps, the size of the smoothing kernel, which often varies between 8 to 12-mm (see Tables 1 and 2), and in the level of significance applied, with some studies correcting for multiple comparisons and others not. These variations make it difficult to compare results directly across studies. Nevertheless, numerous studies have applied this technique to subjects with AD and FTLD and have found largely complimentary results. A number of studies have also used AD as a control group to compare with other neurodegenerative conditions.32,49,50 This review will focus only on those studies in which AD or FTLD was the primary focus of investigation. Most of these studies concentrate on assessing differences in gray matter volume, with only a few looking at white matter or CSF; therefore, only gray matter results will be discussed in this review.
A Summary of Papers That Have Assessed Patterns of Atrophy on VBM in AD or MCI Subjects Compared With Controls
A Summary of Papers That Have Assessed Patterns of Atrophy on VBM in FTLD Subjects Compared With Controls
STUDIES COMPARING AD WITH CONTROLS
Rombouts et al (2000)51 was the first to apply the technique of VBM to study patterns of atrophy in mild to moderate AD. They compared 7 subjects with clinically probable AD to 7 control subjects using the original nonoptimized version of VBM (see Table 1) and found the greatest degree of gray matter loss in the hippocampus. There was also some involvement of the head of the caudate nucleus, insula, middle temporal gyrus, fusiform gyrus, precentral gyrus, postcentral gyrus, and the parietal lobe. The finding of hippocampal atrophy is consistent with both pathological and ROI studies showing that the hippocampus is one of the earliest structures involved in AD.15,52,53
However, larger studies have since been performed and have demonstrated more widespread patterns of loss. Baron et al (2001)54 studied a group of 19 subjects with mild to moderate AD, and although they found that the gray matter loss focused on the medial temporal lobe (including the hippocampus, amygdala and entorhinal cortex), they also showed significant loss in the posterior cingulate and precuneus, insula, temporoparietal association neocortex, and prefrontal gyri. Structures in the central gray matter were also involved, including the caudate, putamen, thalamus, and hypothalamus. These results survived correction for multiple comparisons, most likely reflecting the larger subject numbers. There was however a notable sparing of the motor and sensory cortex and cerebellum. These patterns fit with the proposed pathological staging scheme in AD in which neurofibrillary pathology starts in the entorhinal cortex and hippocampus before spreading into the isocortical association areas.52 Posterior cingulate and parietal lobe involvement has also been demonstrated in functional imaging studies.55 A number of VBM studies have largely confirmed these findings, demonstrating atrophy of the medial temporal lobes, posterior cingulate, parietal lobes and frontal lobes in AD.46,56 However, significant variability does still exist between studies. Although the hippocampi are consistently implicated, opinions differ on which other medial temporal lobe structures are affected, with some implicating the amygdala,54,57-59 entorhinal cortex,60 or parahippocampal gyrus.57,60 In fact, 2 studies have failed to find any gray matter loss in the hippocampus.61,62 The first of these studies showed medial temporal atrophy of the amygdala, entorhinal cortex, and parahippocampal gyrus,61 whereas the other only found atrophy of the temporoparietal neocortex.62 Other studies have shown the focus of loss to be in the lateral temporal lobe rather than the medial temporal lobe.46,63 Subtle differences across studies in the temporal lobe may be due to the inherent difficulties in localization in VBM: smoothing the data reduces the ability to localize regional changes. A couple of studies have also shown involvement of the anterior cingulate, rather than the posterior cingulate, in AD,57,64 and whereas most show a sparing of the sensorimotor cortex,46,54,65 others have found involvement of these regions.61 Although it is important to note that the later study did not correct the results for multiple comparisons and so may be reporting some false-positive results.
The involvement of the subcortical gray nuclei in AD is somewhat controversial. Tissue classification errors often occur around enlarged ventricles during segmentation producing an artificial rim of periventricular gray matter that is often misclassified as atrophy of the central gray matter. Karas et al (2003)65 attempted to overcome this problem by using gray scale erosion followed by dilation around the brain-CSF boundaries with the aim of eliminating the thin rim of tissue around the ventricle. They then performed a comparison between AD and control subjects and confirmed previous reports of gray matter loss of the caudate and thalamus but failed to find involvement of the putamen. Other VBM studies have similarly implicated the thalamus46,56 and caudate nucleus46 to be involved in AD. In fact, atrophy of the caudate nucleus has also been confirmed to be present in AD in a ROI study.66 This raises the possibility that the finding of putamen atrophy in previous studies could have been an artifact of ventricular expansion, especially because these studies did not use customized templates or priors and so would have increased the chance of misregistration or segmentation in atrophic subjects.54,62 However, the subjects in these studies were all defined clinically; it is therefore possible that some of the cases may have had a pathological diagnosis of a non-AD dementia (eg, dementia with Lewy Bodies) which may be contributing to these findings.
Although VBM has not been developed to formally test for differences between different regions of the brain, the VBM results have suggested asymmetry in the brains of subjects with AD. One region could be said to be more severe than another if the voxels show a higher significance (T score) or if the region of atrophy is larger. A number of studies have shown bilateral patterns of medial temporal involvement in AD, although with a slightly greater involvement of the right side.51,54,58,59,61,64 Yet conversely other studies have shown left greater than right involvement of the medial temporal lobe,65,67 lateral temporal lobe,61 and temporoparietal association cortex,54,56,62,67 whereas others have found symmetric patterns of loss.46 Differences between studies are likely to be due to differences in the clinical cohort, for example subject age, disease duration or clinical severity, and differences in the VBM methods.
Affects of Age at Onset
Two studies have also shown that the patterns of gray matter loss identified on VBM differ depending on the age of onset of the AD subjects. Studies by Frisoni et al (2005)68 and Ishii et al (2005)58 both divided AD subjects into early onset (age at onset ≤65 years, EOAD) versus late onset (age at onset >65, LOAD), and then compared them with age-matched young and old control groups, respectively. The demographics of the subject groups are shown in Table 1. There were a number of differences between the studies in terms of the VBM processing and in the subject demographics, with the LOAD subjects relatively older in the Frisoni et al (2005)68 study, yet they showed strikingly similar results. Both studies showed a relatively restricted pattern of loss in the LOAD subjects involving the medial temporal lobes when compared with the old controls, yet the patterns of loss in the EOAD subjects showed additional involvement of the parietal lobes, precuneus, middle temporal gyrus, fusiform gyrus, and the frontal lobes when compared with the young controls. The Ishii et al58 study performed direct statistical comparisons between the 2 groups and showed that the EOAD group had significantly greater involvement of the precuneus, parietal lobe, right middle temporal gyrus, and left fusiform gyrus than the LOAD group. Visual inspection of the results in both studies suggested greater medial temporal involvement in the LOAD group than the EOAD group, although this difference did not reach significance in a direct comparison. These results concord with previous studies that have shown greater memory impairments in LOAD subjects.69
These studies have therefore demonstrated striking differences between early and late onset AD. The LOAD subjects show a relatively restricted pattern of loss involving the medial temporal lobes, whereas the EOAD subjects show a more widespread pattern involving the medial temporal lobes but also the precuneus and temporoparietal association neocortex. It is likely that the early onset group may have genetic underpinnings,70,71 whereas environmental factors as well as genetics (eg, APOE genotype) may contribute to the late onset group. However, no genetic mutation information was available on subjects in either study, so this remains speculative. The authors suggest that the temporoparietal neocortex may be genetically controlled, whereas the medial temporal lobes may show greater age-associated changes which reduces the functional reserve and predisposes older patients to the development of symptoms related to medial temporal lobe damage.68 These results further suggest that these groups should be considered separate in future studies of AD. Previous reports of anatomical variability in AD may have been the result of including both early and late onset groups.
Voxel-based Morphometry in Mild AD
A couple of studies have also looked particularly at the use of VBM in very mild cases of AD. Detecting the disease at very mild stages will ultimately be important for the early treatment of subjects. Frisoni et al (2002)64 specifically selected 3 of their AD subjects that were in the highest decile of the mini-mental state examination (MMSE) distribution (MMSE = 26, 27, and 27) and compared them with 26 controls. The main regions of gray matter loss were identified in the medial temporal lobe, including the hippocampus and amygdala. Smaller regions were also identified in the temporal and frontal gyri, although only the hippocampus and amygdala region survived after a correction for multiple comparisons. Busatto et al (2003)61 similarly performed a subanalysis on cases with mild AD but with an MMSE more than or equal to 20 (n = 11). They identified gray matter loss bilaterally in the posterior parahippocampal gyrus, the right amygdala/entorhinal cortex complex, the left inferior and middle temporal gyri, the fusiform gyrus, and the superior temporal gyrus. The left posterior inferior temporal gyrus and fusiform gyrus survived after a correction for multiple comparisons. Therefore, both studies have shown that VBM can detect atrophy in subjects with mild AD, although the results from the 2 studies differed with the Frisoni et al (2002)64 study showing a focus of loss in the amygdala and hippocampus but the Busatto et al (2003)61 study showing a focus of loss in the inferior temporal and fusiform gyrus with no hippocampal involvement. The Busatto et al (2003)61 study also showed a more widespread pattern of loss which is most likely because of their subjects were more severe clinically, and the subject group was larger. These findings therefore suggest that both the medial temporal lobe and inferior temporal regions are preferential sites for early pathological involvement in AD.52
A study by Matsuda et al (2002)63 attempted to address the question of how these regions of atrophy change over time by assessing patterns of atrophy in AD over multiple time points. Fifteen subjects with AD were compared with controls at 3 time points each approximately 1 year apart. The MMSE score was a mild 26 at baseline, and then 22 and 21 at each respective time point. At baseline, the AD subjects showed loss of the temporal lobe, including the superior and inferior temporal gyrus, amygdala, hippocampus and parahippocampal gyrus, and also the frontal lobe, posterior cingulate and precuneus, compared with controls. At follow-up, these regions of loss spread to surrounding areas with further atrophy of the thalamus, head of the caudate nucleus, anterior cingulate, and the parietal lobe. These results therefore suggest early involvement of the medial and lateral temporal lobes and the posterior cingulate/precuneus in AD, whereas atrophy of the anterior cingulate, parietal lobe and central gray matter structures occur later in the disease course.
The question then arising from these group studies is how applicable are these patterns of atrophy to diagnosis in subjects with AD, especially subjects in very early stages of the disease? Kawachi et al (2006)57 attempted to answer this question by assessing patterns of atrophy in a group of subjects with mild AD (MMSE 22, range, 21-23) and then investigating the diagnostic use of the results (z score maps) in a group of very mild AD subjects (MMSE 26, range 24-28). The mild AD subjects showed loss of the amygdala/hippocampal complex and the bilateral temporal and frontal gyri compared with controls. The ROI maps were defined from this first comparison, and Z scores were calculated for each segmented MRI in the very mild AD group by comparison with the mean and SD of the segmented MRI in the first healthy control group. Receiver operator characteristic (ROC) analysis with the maximum Z scores in the ROIs showed that in very mild AD subjects, the accuracy of VBM diagnosis was very high, at 83%.
Voxel-based Morphometry in Mild Cognitive Impairment
Voxel-based morphometry has also been used to investigate patterns of atrophy in subjects with amnestic mild cognitive impairment (MCI). This syndrome is considered as a transitional period between normal aging and a diagnosis of AD, with subjects showing early deficits in memory but not fulfilling current criteria for dementia.72 Most of the VBM studies have demonstrated gray matter loss predominantly involving the medial temporal lobe, including the hippocampus, parahippocampal gyrus, and amygdala, in amnestic MCI.73-77 Some have shown symmetric involvement,76,77 whereas others have shown greater involvement of the right temporal lobe.73,75 A study by Hirata et al (2005)76 used a method similar to that by Kawachi et al (2006)57 (described above) and showed that atrophy of parahippocampal gyrus has an accuracy of 87.8% in discrimination of amnestic MCI subjects from controls. Lateral regions of the temporal lobe, including the inferior, middle, and superior temporal gyri, have also been implicated.73-75,77 A couple of studies have also shown severe loss of the thalamus in MCI.73-75 Although the thalamus has been shown to be atrophic in AD (see above), these results do not concord with those by Matsuda et al (2002)63 that showed thalamic involvement later in the disease course in AD. Findings concerning the cingulate gyrus are variable, with some studies showing loss of the posterior73 and anterior cingulate gyrus,73,75 whereas others failed to find any involvement in subjects with MCI.74,76,77 A couple of studies have compared MCI subjects with those with a clinical diagnosis of AD.73,74 They both demonstrated greater involvement of the temporoparietal association areas in AD than MCI. Karas et al74 (2004) also found greater involvement of the medial temporal lobes in AD, although the study by Chetelat et al (2002)73 failed to find any differences in this region.
The regions implicated in MCI are therefore largely consistent with previous studies on mild AD, suggesting that the disease starts at a focus in the medial temporal lobes. Often, by the time subjects are scanned, the disease has also spread into the basal and lateral temporal neocortex. The MCI studies have extended previous findings by showing that these structures are affected even before subjects are given a clinical diagnosis of dementia. However, varying degrees of gray matter loss have been demonstrated in MCI, with some studies showing more widespread involvement of the frontal and parietal lobes,73,74,77 whereas others show patterns of loss relatively restricted to the temporal lobes.76 Variability across studies could reflect differences in the clinical criteria used for the diagnosis of MCI, in the severity of subjects, the group size, and in the VBM techniques used (see Table 1). Although it is very likely that amnestic MCI subjects will progress to AD,78 it is also possible that some subjects may progress to other non-AD dementias or not progress clinically. Chetelat et al (2005)79 showed greater atrophy of the medial and lateral temporal lobes in those MCI subjects that do go on to progress to AD compared with those that do not. There is also some suggestion that the patterns of atrophy vary dependent upon the number of cognitive domains showing impairment, with subjects that show deficits in multiple domains showing a more widespread pattern of atrophy.77
STUDIES COMPARING FTLD WITH CONTROLS
Studies that have applied VBM to groups of subjects with FTLD are summarized in Table 2. Patterns of regional atrophy in FTLD generally involve the anterior temporal and frontal lobes, with some involvement of subcortical gray matter structures,67,80 in keeping with the findings from pathological studies.81 However, as discussed in the introduction, FTLD consists of 3 different syndromic variants which likely have different patterns of regional involvement. Voxel-based morphometry studies in FTLD have tended to concentrate mainly on the language variants of FTLD, especially SD. This review will therefore discuss each syndromic variant in turn, beginning with SD, and will then focus on studies that have performed comparisons firstly between the 3 variants and also between each variant and AD.
Semantic Dementia
Mummery et al (2000)82 were the first to apply VBM to subjects with SD. They performed essentially a single subject VBM analysis by comparing gray matter density for each of 6 subjects with SD with an age-matched control group (n = 14). A conjunction analysis was subsequently performed to determine which regions were common to all 6 subjects. Regions that were common to all subjects were found predominantly in the left temporal lobe, particularly involving the temporal pole. The loss also extended medially to involve the amygdala, posteriorly to the fusiform gyrus, and laterally to involve the middle and inferior temporal gyrus. The right temporal pole and amygdala were also involved but to a lesser degree. It was also notable that the temporal lobe showed an anterior-posterior gradient of loss, with a relative sparing of posterior regions of the temporal lobe. Gray matter loss was also identified outside the temporal lobes in the left insula and the ventromedial frontal cortex. This pattern showed a striking left-greater-than-right (L > R) asymmetry. On an individual basis, 4 of the 6 subjects showed L > R asymmetry, whereas one showed no asymmetry, and one had greater atrophy on the right side. There was generally a high correspondence between subjects, although the authors noticed that the cases that were more severe clinically showed a pattern of temporal lobe loss extending further back along the middle and inferior temporal gyri suggesting that the disease begins in the anterior temporal lobe and then spreads back in the temporal lobe. Rosen et al (2002)83 similarly used VBM to investigate patterns of gray matter atrophy in a single case of SD and found complementary results with the focus of loss identified in the left anterior temporal lobe with further involvement of lateral and medial structures. They also looked at the pattern of loss 15 months later and showed more extensive loss in the left temporal lobe as well as increasing involvement of the right temporal lobe. These studies therefore suggest that VBM is sensitive to change in subjects with SD, even at the single-subject level. However, VBM is designed for group studies and has not been optimized fully for single-subject comparisons. Single-subject comparisons appear to be less robust to violations of normality.39 A significant number of false positives can arise if adequate smoothing is not performed,39 and there is also the risk of false negatives.
A number of group studies have since been performed comparing patterns of gray matter atrophy in subjects with SD with healthy controls. Good et al (2002)46 studied 10 subjects with SD and compared them with 10 age- and sex-matched control subjects. As in the single-subject studies, they found significant loss in the amygdala, fusiform gyrus, inferior and middle temporal gyrus, and the temporal pole, with a greater involvement in the left than the right temporal lobe. However, in addition, they also demonstrated loss in the hippocampus, entorhinal cortex, and superior temporal gyrus. Region of interest studies have similarly confirmed the presence of hippocampal atrophy in SD,24,25,46 suggesting that the lack of hippocampal atrophy in the Mummery et al (2000)82 study was a false-negative result because of reduced statistical power. In addition, the Good et al (2002)46 study used customized templates and prior probability maps which may have improved normalization.46
More recent VBM group studies have generally confirmed the results of these early studies, showing severe atrophy of anterior inferior temporal regions in subjects with SD. These temporal regions have been shown to be critically involved in language production, including the formulation of language, and semantic processing. A number of studies have also confirmed the presence of hippocampal atrophy in SD,56,84,85 in fact some have demonstrated that the hippocampus and amygdala are the most severely affected structures.85 As well as these temporal lobe regions, some VBM studies have shown loss in ventromedial and superior frontal regions in SD.48,56,82,84,85 Atrophy of the frontal lobes is consistent with the fact that SD subjects often develop behavioral deficits as the disease progresses. However, some studies have failed to find any involvement of the frontal lobes,67,86 possibly reflecting an earlier disease stage in these subjects. Posterior changes have also been reported in SD, although less severe than the more anterior temporal and frontal lobe changes.67
A couple of studies have also shown involvement of subcortical gray matter structures in SD, including the left dorsomedial thalamus56 and the caudate nucleus.85 As discussed previously, it is important to be cautious in the interpretation of atrophy in these structures given the problems with misclassification around the ventricular rim. This may have been a particular problem in these studies because subjects with SD show a large amount of asymmetric ventricular enlargement, the authors did not use a customized template consisting of both disease and control subjects, and they also did not customize their prior probability maps. However, caudate atrophy has been demonstrated to be present in FTLD using linear measurements and shown to correlate to the presence of simple compulsive behaviors.87
A pattern of left-sided asymmetry has been found in a number of studies,46,67,82,88 although others have shown a more symmetrical pattern.56,84,85 At post mortem, the atrophy similarly appears either asymmetric,89,90 or symmetric,90,91 with marked involvement of the temporal lobes.89-91 A number of studies have suggested that the earliest site of change in subjects with SD is the left temporal lobe, but that as the disease progresses, the atrophy spreads to the right temporal lobe.83,92 Differences in asymmetry across studies may therefore be reflecting a difference in disease stage. However, it has recently become apparent that subjects with SD can show more severe involvement of the right than the left temporal lobe, providing almost the mirror image of the patterns typically observed in SD.93 These subjects showed social awkwardness and inappropriateness, behavioral abnormalities, loss of insight, and difficulty with person identification,94-97 whereas the word-finding difficulties and reduced comprehension typical of SD were found insubjects with predominantly left-sided temporal lobe atrophy.93 Voxel-based morphometry has demonstrated significant atrophy in the right amygdala/hippocampal complex and right insula in a case of right-temporal lobe variant of SD.97
Progressive Nonfluent Aphasia
Fewer studies have used VBM to assess the pattern of atrophy in subjects with PNFA. One of the earliest studies by Nestor et al (2003)98 examined 10 subjects with PNFA and compared them with 10 control subjects. They also performed a secondary analysis of "pure" cases of PNFA after removing 3 subjects which had a more extensive dementia. However, both the whole group and the "pure" PNFA group only revealed 1 small region of gray matter loss in the left perisylvian region, even at a low and uncorrected threshold of P < 0.001. Previous reports of subject with PNFA have similarly demonstrated atrophy in the left frontal lobe and the perisylvian region on MRI67,85,86,98-101 and at postmortem99 but have also highlighted the large degree of variability present between subjects. Nevertheless, more recent studies have found larger regions of significant atrophy in subjects with PNFA. Gorno-Tempini et al (2004)85 found gray matter loss in the left inferior frontal gyrus including Brocas area, the left insula, the inferior precentral gyrus, and the middle frontal gyrus in 11 subjects with PNFA compared with controls. They also found some involvement of the caudate nucleus and putamen although at a lower statistical threshold. Although this study showed a more widespread pattern of loss than the study by Nestor et al (2003),98 the atrophy still focused predominantly around the left perisylvian fissure. However, temporal lobe regions have also been implicated. Grossman et al (2004)67 studied 7 subjects with PNFA and showed loss in the left anterior and ventral temporal lobe, as well as in the left insula, premotor cortex, dorsolateral prefrontal cortex, and inferior frontal lobes. In addition, Zahn et al (2005)62 found an almost exclusively temporal pattern of involvement in PNFA, involving the inferior and middle temporal gyri, as well as the putamen. Inferior regions of the parietal lobe have also been implicated in PNFA.86,102
These studies therefore highlight the fact that patterns of atrophy in PNFA are relatively heterogeneous. However, there were a few trends throughout the studies, namely that the atrophy was restricted to the left hemisphere and that it involved the frontal and temporal lobes, with a particular preference for the regions of the brain surrounding the perisylvian fissure. Studies have suggested that some of these regions play a role in motor speech and syntax processing.103 Although the VBM methods did vary across studies, it is likely that the observed variability reflects clinical heterogeneity in the patient cohorts rather than inaccuracies introduced by VBM. Clinical definitions of PNFA vary across studies.62 In fact, other categorizations of PNFA exist, a recent clinicopathological study has split subjects into a number of different language categories and shown different patterns of atrophy in each group.104 For example, they suggest that atrophy of the premotor cortex and supplemental motor area is associated with a specific motor speech disorder, apraxia of speech.104 It is hard to compare categories across studies. Subjects also vary clinically; although language remains the only deficit for a number of years, patients often develop behavioral problems and more widespread cognitive decline later in the disease course. For example, Rosen et al (2002)83 used VBM to investigate the patterns of atrophy in 2 subjects with PNFA. Case 1 showed loss predominantly in the frontal lobes and presented with some behavioral abnormalities. In contrast, case 2 showed no behavioral deficits but did show evidence of a more global cognitive dysfunction more typical of AD and consequently showed a more temporoparietal pattern of atrophy. Similarly, another single-case study investigated serial VBM in a subject that presented as PNFA but later developed features of corticobasal degeneration (CBD).102 At baseline, the subject showed loss of the inferior frontal gyrus, insula and frontal and temporal lobes, compared with controls. As the disease progressed, the thalamus and parietal lobe became involved, and then after 4 years when the subject began to show clinical signs of CBD, they found additional involvement of the left prefrontal regions and the supplementary motor cortex. These cases illustrate how the patterns of anatomical involvement vary depending upon clinical presentation. The pathology underlying PNFA is also highly heterogeneous105-107 and may influence the patterns of atrophy. In fact, subjects with PNFA often have AD and CBD on pathology.105,107,108 The anatomical correlates of pathology will be discussed in more detail later.
Frontotemporal Dementia
A few studies have used VBM to assess the patterns of gray matter atrophy in the frontal variant of FTLD. One recent study by Boccardi et al (2005)109 studied 9 subjects with FTD and compared them with 26 control subjects. They found that the areas of loss correlated to the rostral limbic system, including the anterior cingulate, ventromedial frontal cortex, right ventral striatum, amygdala, anterior insula, and the periaqueductal gray. Small regions of loss were also found scattered in the left frontal gyrus, including Brocas area, and in the left inferior temporal gyrus. The rostral limbic system is important in the appropriate tuning of behavior and therefore damage to this system may be responsible for the behavioral disturbances observed in subjects with FTD. A couple of other VBM studies have similarly found loss in the anterior insula, the anterior cingulate and the frontal lobes in FTD,67,84 although loss has also been observed in the left anterior and medial temporal lobe.67 In contrast to the results in both SD and PNFA, the patterns of loss were relatively symmetric in each of the FTD studies, consistent with previous MRI studies.101 There were, however, a number of methodological differences between the studies, including the size of the smoothing kernel and the type of template used (Table 2). It will be important to replicate these findings in larger group studies.
Comparisons Between Syndromic Variants
All the studies discussed so far have compared the different syndromic variants of FTLD with a control population. However, a couple of studies have also attempted to directly assess the differences and similarities between the different FTLD subtypes. Rosen et al (2002)84 studied 12 subjects with SD and 8 subjects with FTD and assessed regions of loss that were common to both groups and regions that were different using a series of masking procedures. For example, to assess regions where significant atrophy was present in both FTD and SD relative to controls, they compared all 20 subjects with a control group of 20 subjects and then inclusively masked the results such that only those voxels with significant gray matter tissue loss in the FTD-versus-controls and SD-versus-controls comparisons were included. They found gray matter loss in the ventromedial cortex, insula, and anterior cingulate in both the FTD and SD subjects. This led them to hypothesize that atrophy of these structures may be contribute to behavioral deficits that are common to both SD and FTD subjects, such as changes in eating behavior, obsessive-compulsive behaviors, apathy, mental rigidity, disinhibition, and aggression.110,111 Regions that showed greater loss in the SD group compared with the FTD group and controls were found in the anterior inferior temporal gyrus, anterior superior temporal gyrus, the amygdala/hippocampal region, and the ventromedial frontal cortex. In contrast, regions that showed greater loss in the FTD group compared with the SD group and controls were found in the anterior insula, anterior cingulate, premotor cortex, and the frontal lobes (middle frontal gyrus and superior frontal gyrus).
Similarly, Grossman et al (2004)67 also studied differences and similarities between FTLD subgroups by comparing pairs of comparisons involving each patient group and healthy controls. They studied subjects with SD, FTD, and PNFA and found left anterior temporal cortex loss in all 3 groups. In contrast to the results by Rosen et al (2002),84 the SD group showed greater loss in the left ventral temporal lobe and the occipital lobe relative to the FTD subjects. However, the results in the FTD versus SD comparison were remarkably similar to the Rosen et al (2002)84 study showing greater involvement of the anterior cingulate, insula, and the dorsolateral prefrontal cortex in FTD. The PNFA group showed greater loss in the frontal lobes compared with SD and greater atrophy in the temporal and occipital lobes compared with FTD. It is however difficult to interpret the findings of greater occipital atrophy in PNFA because this region has not been implicated in any studies that have compared PNFA with controls. It may be a false-positive result because the data were not corrected for multiple comparisons.
STUDIES COMPARING FTLD WITH AD
A number of VBM studies have compared patterns of atrophy in AD with FTLD in the hope of elucidating differences that may aid in the differential diagnosis of the 2 dementias. As before, most studies have concentrated on the different syndromic variants of FTLD, although 1 study has compared AD with an umbrella group of FTLD subjects.67 They demonstrated greater gray matter loss in the hippocampus, medial temporal lobe, and temporoparietal regions in AD than FTLD,67 and greater loss of the left medial frontal lobe, the left striatum, and prefrontal regions, in FTLD than AD subjects.
A study by Boxer et al (2003)56 performed a comparison between AD (n = 11) and age-, sex-, education- and MMSE-matched SD subjects (n = 11). They performed direct statistical comparisons between the 2 groups and showed that the AD group had significantly greater loss in the left parietal lobe and the posterior cingulate/precuneus bilaterally than the SD group. The left parietal lobe showed the most significant loss. In addition, they analyzed the gray matter signal of a voxel in the left parietal lobe in each subject and found that 10 of the 11 AD subjects had less parietal lobe gray matter than subjects with SD. The reverse comparison showed that the SD group had significantly greater loss bilaterally in the amygdala, hippocampi and anterior temporal lobes, the right middle temporal gyrus, and the left temporal pole, than the AD group. A discriminant function analysis showed that the combination of gray matter tissue concentration at voxels in the right posterior cingulate, left parietal lobe, right amygdala, and right anterior temporal lobe was best able to differentiate the 2 subject groups and controls. The discriminant function analysis correctly classified 96% of the patients. These results highlight the use of cinguloparietal measurements in the differential diagnosis of AD, and show that severe patterns of temporal lobe atrophy suggest a diagnosis of SD. Similarly, another study found greater loss in temporal lobe structures in SD than in AD on direct statistical comparison.46 Region of interest studies have also found significantly smaller volumes of all left temporal lobe structures in SD compared with AD.24,25
A couple of studies have also compared AD with subjects with PNFA. One such study demonstrated greater loss of the right hippocampus, the posterior cingulate/precuneus, and the right posterior temporoparietal region in AD than PNFA.62 This again highlights the role of the posterior cingulate/precuneus and parietal lobes in the diagnosis of AD. In the reverse comparison, they found that the PNFA subjects showed greater loss in the left anterior temporal lobe and right inferior temporal gyrus than AD subjects. They also attempted to look at the diagnostic value of these results in single cases. A VBM map was created for each case compared with a control group, and they looked at the presence of abnormalities in regions of interest placed over the medial temporal lobe, posterior parietal lobe, posterior cingulate/precuneus, and the left anterior temporal lobe. They defined a set of criteria for the diagnosis of each group (AD, evidence for medial temporal or posterior cingulate/precuneus abnormality; PNFA, evidence for left anterior lateral temporal abnormality with no involvement of the medial temporal lobe or posterior cingulate/precuneus). Using these criteria, the diagnostic classification confirmed the clinical diagnosis in 12 (80%) of the total 15 cases suggesting again that the patterns of atrophy identified on VBM could be diagnostically useful. However, the other study that compared PNFA and AD subjects failed to identify any regions that showed greater loss in PNFA than in AD,98 suggesting that the diagnostic rules devised for PNFA in the previous study may not be applicable across studies.
STRUCTURE-FUNCTION CORRELATIONS WITH VBM
Voxel-based morphometry has also been used to directly correlate behavioral and cognitive deficits with regional patterns of gray matter loss in both AD and FTLD subjects. The standard VBM studies discussed above have implied associations between anatomical damage and functional deficits. Neuropsychological testing is helpful in characterizing the cognitive profile of subjects with AD and FTLD and provides quantitative scores that can be used to correlate to gray matter atrophy. Similarly, behavioral deficits can be quantified using caregiver questionnaires such as the Neuropsychiatric Inventory (NPI).112 Studies have used 2 basic mechanisms to assess correlations. The first involves entering the cognitive/behavioral scores into VBM and performing a regression analysis between the scores and the gray matter density values. A second less direct method is to derive gray matter density values for a particular voxel from each subject and then perform a regression of these values and the cognitive scores outside of VBM. The advantage of the first technique is that it looks for associations across the whole brain volume, whereas the second technique will only assess correlations at a few specific and selected voxels.
A number of studies have investigated anatomical correlations of semantic memory deficits (ie, loss of memory for words) in subjects with SD. Mummery et al (2000)82 looked for correlations between semantic deficit and loss of either the anterior temporal lobe or ventromedial frontal cortex in subjects with SD. They selected these regions based the comparison between the SD subjects and controls. The relative gray matter density values for the peak voxel in these 2 regions were derived for each of their 6 SD subjects and correlated to test performance using a Spearman rank test. They found that the degree of loss in the left temporal lobe correlated with semantic performance, but loss of the ventromedial cortex did not, leading them to conclude that this region plays an important role in the semantic deficits present in this patient group. Another study used the same method to correlate performance on a number of neuropsychological tests to the peak voxel of each significant cluster identified in the group analysis of disease subjects to controls.84 They found a correlation between a semantic memory task and the inferior temporal gyrus in a group of subjects with SD. A more recent study has attempted to correlate semantic deficits to specific regions of brain atrophy in subjects with FTLD by performing a regression analysis within VBM.80 The analysis will therefore look for correlations in each voxel of the brain and does not require any a priori assumptions concerning which structures to assess. They found that, as in the previous 2 studies, semantic deficits correlated to loss in the left anterior temporal lobe, particularly the temporal pole, parahippocampal gyrus, fusiform gyrus, amygdala, and the superior and middle temporal gyrus suggesting that all these regions may contribute to semantic processing.
Various other studies have also assessed anatomical correlations of confrontation naming impairments.67,113 Confrontation naming deficits could be associated with deficits in lexical retrieval (ie, retrieving and expressing the name of an object), semantic or visual components of naming. These studies performed regression analyses within VBM and demonstrated correlations between the left anterior temporal cortex and naming deficits in both AD and FTLD. They concluded that this region may therefore play a role in lexical retrieval because this deficit was present in all subjects. However, within the AD and FTLD groups different regions correlated to naming deficits suggesting a different basis for naming deficits in the 2 diseases.67,113 Some studies have however failed to find a correlation between naming and regional atrophy.86
A couple of studies have also investigated anatomical correlates of abnormal behaviors in FTLD.80,114 Behavioral deficits were assessed using the NPI. In the first study, a summary score was calculated which summarized a number of abnormal behaviors, including apathy, aberrant motor behavior, and abnormal eating behaviors, for a group of subjects with FTD. The summary score was then entered into the VBM analysis as a covariate of interest and was found to correlate to a region of gray matter loss in the mesial frontal lobe. Although it is difficult to attribute this region to a particular behavior, it suggests that frontal atrophy plays an important role in the behavioral deficits observed in FTD.80 A more recent study by Rosen et al (2005)114 then looked in more detail at specific behavioral abnormalities. The NPI score from a number of different behaviors was correlated to gray matter loss in a group of subjects with dementia. The results showed that apathy correlated with tissue loss in the ventromedial superior frontal gyrus, disinhibition correlated with the subgenual cingulate gyrus, and aberrant motor behavior correlated with tissue loss in the dorsal anterior cingulate cortex and premotor region.114 Damage to these specific regions of the frontal lobe may therefore contribute to these different behavioral abnormalities in subjects with dementia. Emotional impairments are also a common feature of both FTLD and AD, although they tend to be more common in FTLD.115 The same group of investigators have used similar methods and showed that impairment in the recognition of negative emotions correlated to tissue loss in the right inferior and middle temporal gyri.116 All these results suggest that VBM is a useful way of mapping brain atrophy and correlating anatomical and behavioral or emotional changes in subjects with dementia.
The MMSE provides an overall index of neuropsychological impairment and has also been correlated to regional gray matter atrophy in subjects with AD64 and MCI.77 The MMSE scores were entered into the analysis as covariates of interest and were found to correlate to loss in the temporoparietal cortex, mainly involving superior and inferior temporal gyri, parietal lobe, posterior cingulate and precuneus in AD,64 and the entorhinal cortex and inferior frontal gyri in MCI.77
In addition, studies have used VBM to correlate gray matter atrophy with other disease biomarkers. For example, levels of tau in the CSF can often be useful in distinguishing FTLD from AD, and 1 study has shown that CSF tau levels correlate to right frontal and left temporal cortical loss in subjects with FTLD.117 Another study has examined the effect of the apoliloprotein (APOE) ε4 allele on the presence of cortical atrophy.118 APOE ε4 is associated with an increased risk of AD119 and has previously been shown to correlate to hippocampal volume in subjects with a mild cognitive impairment.120 The study found that AD subjects carrying the e4 allele had greater loss of amygdala and head of the hippocampus than those without the allele. In addition, FTD subjects carrying the allele had greater loss in right frontotemporal regions and striatum than those without. Therefore, greater amounts of cerebral atrophy were observed in subjects with the APOE e4 allele, although there were only 2 subjects in the FTD carrier group. These results are consistent with the hypothesis that APOE has a role in modulating the morphological expression of degenerative dementias.
PATHOLOGICAL CORRELATIONS IN VBM
The studies described so far have concentrated on examining patterns of atrophy in clinically defined AD and FTLD. Pathological diagnosis is however the only gold standard. The pathological features of AD include the presence of sufficient numbers (age-related) of amyloid plaques and neurofibrillary tangles in brain tissue. There is a relatively good correspondence between the clinical and pathological diagnosis in AD;121,122 however, the pathological diagnosis of FTLD is heterogeneous, including a number of conditions characterized by the presence of intracellular inclusions formed by abnormal cytoskeletal components both in neurons and glial cells. Current classification of FTLD divides FTLD into tau-positive and tau-negative diseases,107,123 dependent upon the presence of tau protein. The tau-positive group includes 4 forms of tauopathy: Pick's disease (PiD), progressive supranuclear palsy, CBD, and frontotemporal dementia and parkinsonism linked to chromosome 17 (FTDP-17). The tau-negative group includes 1 form with tau-negative, ubiquitin positive inclusions without motor neuron disease (FTLD-U), 1 form with tau-negative, ubiquitin positive inclusions with motor neuron disease (FTLD-MND) and 1 form with neither tau-positive nor ubiquitin-positive inclusions (dementia lacking distinctive histology124). In addition to these 3 tau-negative diseases, an additional form associated with intermediate filaments is recognized.125 These multiple pathological subtypes of FTLD do not map neatly onto the clinical syndromes, making it difficult to predict FTLD pathology during life.105-107 The identification of reliable in vivo predictors of tissue pathology would therefore be very valuable for the development, evaluation and monitoring of disease-modifying therapies in FTLD.
A number of studies have used VBM to assess the patterns of gray matter atrophy in different pathological subtypes of FTLD. The first study compared patterns of atrophy in tau-positive and tau-negative FTLD.126 They found no significant difference between the 2 groups, with both showing a pattern of frontotemporal gray matter loss. This suggests that patterns of atrophy using VBM do not predict the presence or absence of tau pathology. Postmortem studies have similarly failed to establish patterns of brain atrophy that reliably distinguish diseases with and without tau pathology.127-129 However, as described above both the tau-positive and tau-negative groups are pathologically heterogeneous, consisting of a number of specific pathological substrates. More recent studies have therefore examined VBM in the specific pathological substrates of FTLD. Whitwell et al (2005)130 examined patterns of gray matter atrophy in a group of 21 pathologically or genetically confirmed subjects with FTLD; 9 with FTLD-U, 7 with PiD, and 5 familial FTLD with tau exon 10+16 mutations (tau exon 10+16). Each group was compared with a group of 20 healthy controls. The middle and inferior temporal gyri, medial temporal lobes, insula, and orbitofrontal cortex were involved in all pathological groups. However, each group also showed a specific signature of loss: FTLD-U was associated with predominant temporal lobe atrophy, whereas PiD was associated with a more severe pattern of bilateral frontal atrophy, and tau exon 10+16 was associated with a more focal pattern of atrophy involving the medial temporal lobe. These results were largely supported by visual assessments of regional atrophy. The sensitivity and specificity of the VBM signatures of pathology were estimated from the individual visual assessment scores for each pathological subgroup. The pattern of temporal lobe atrophy in FTLD-U had a sensitivity of 67% and specificity of 83% in distinguishing FTLD-U from the other pathologies. In addition, the findings in the FTLD-U group have since been reproduced in another independent study performed at a different institution.131 As before, the patterns of loss in a group of 11 subjects with FTLD-U mainly involved the temporal lobes, particular the posterior temporal lobe. It is important to note that these similar results were obtained although the clinical diagnosis of the FTLD-U subjects differed across the studies. The subjects in the first study were a mixture of FTD and SD subjects, whereas the subjects in the second study all had a clinical diagnosis of FTD. In addition, this second study showed a significant difference between subjects with FTLD-U and FTLD-MND, with the FTLD-MND group showing a pattern of loss restricted to the frontal lobe, consistent with previous PET findings.132 These studies collectively suggest that patterns of atrophy detected using VBM may help predict pathological diagnosis in FTLD. This may be especially useful in subjects with a clinical diagnosis of FTD in which the pathology is particularly variable.
LONGITUDINAL STUDIES
Most of the VBM studies in subjects with AD and FTLD have been cross-sectional, assessing patterns of brain atrophy at a single point in time. These studies are therefore limited by the large amount of between-individual variation in brain shape and volume within a normal population. This variation makes diagnosis based on just 1 scan difficult and limits the power to detect and localize group differences. Performing serial scans over a number of years allows disease progression to be tracked over time and by making within-individual comparisons reduces the problem of inter-individual variation. This review has already discussed a few studies that have performed cross-sectional VBM analyses at a number of different time points in an attempt to assess longitudinal change, although most of these studies were performed in single individuals.63,83,102 These studies basically compared the baseline scans with a group of controls, and then compared the repeat scan with the same group of controls and visually described the differences between the 2 different VBM outputs. These techniques do not however perform any statistical comparisons between the 2 time points.
One solution is to perform a nonlinear registration between the baseline and repeat scan. The nonlinear registration aims to transform the repeat scan so that it is identical to the baseline scan. The Jacobian determinants of the deformation field then contain an estimate of the volume change between both images. There are a number of ways that these Jacobian determinants can then be processed. One technique which has been applied to study subjects with MCI is to create an average of the baseline and warped repeat image to create the "baseline" data for the VBM analysis. This image is then multiplied by the Jacobian determinants to create the "follow-up" data.79 The images are then normalized, segmented, and smoothed and compared using 2 sample t tests. The study by Chetelat et al (2005)79 used this technique and demonstrated atrophy of the medial and lateral temporal lobes, posterior cingulate, and precuneus over a period of 18 months in subjects with MCI. They also demonstrated greater rates of loss in these regions in subjects that go on to convert to AD compared with those that do not.
Another method which has been applied to subjects with AD and FTLD involves the use of a fluid registration.92,133 The fluid registration is a nonlinear registration that matches the repeat scan to the baseline using a series of voxel-level deformation fields that are based on the constraints of a viscous fluid model.134,135 The Jacobian determinants of the deformation field can be used to quantify volume change at the voxel level in each individual. Two images are created for each subject, one showing voxel expansion and the other voxel contraction. These images are normalized to a template, smoothed, and analyzed using SPM. The study by Scahill et al (2003)133 applied this technique to look at patterns of regional change in presymptomatic, mild, and moderate subjects with AD. They demonstrated increased rates of hippocampal atrophy in presymptomatic subjects and mildly affected subjects. However, they observed a shift in the distribution of temporal atrophy as the disease progresses with the inferolateral regions of the temporal lobes showing the most increased rates of atrophy by the time the subjects were moderately affected. Whitwell et al (2004)92 also applied this technique to study subjects with FTLD. They found increased rates of atrophy in the frontal, temporal, and parietal regions in subjects with FTLD. However, the different syndromic variants showed different patterns of regional change. The SD subjects showed increased atrophy in the posterior left temporal lobe, the inferior frontal lobes, and the right temporal lobe, whereas the FTD and PNFA groups showed very little atrophy over time. The only region to show significantly increased rates of atrophy in the PNFA group was found in the left perisylvian region. These results suggest that in subjects with SD the disease spreads from a focus in the left anterior temporal lobe both anteriorly and posteriorly in the left hemisphere and progressively involves the right temporal lobe. These 2 studies therefore suggest that this technique provides an insight into the disease evolution in AD and FTLD.
It is however important that these different VBM-based longitudinal techniques are validated and compared. These techniques will become increasingly important in assessing the efficacy of disease-modifying treatments in future clinical trials.
REFERENCES
1. Hebert LE, Scherr PA, Bienias JL, et al. Alzheimer disease in the US population: prevalence estimates using the 2000 census. Arch Neurol. 2003;60:1119-1122.
2. Ratnavalli E, Brayne C, Dawson K, et al. The prevalence of frontotemporal dementia. Neurology. 2002;58:1615-1621.
3. Harvey RJ, Skelton-Robinson M, Rossor MN. The prevalence and causes of dementia in people under the age of 65 years. J Neurol Neurosurg Psychiatry. 2003;74:1206-1209.
4. Neary D, Snowden JS, Gustafson L, et al. Frontotemporal lobar degeneration: a consensus on clinical diagnostic criteria. Neurology. 1998;51:1546-1554.
5. Warrington EK. The selective impairment of semantic memory. Q J Exp Psychol. 1975;27:635-657.
6. Hodges JR. Frontotemporal dementia (Pick's disease): clinical features and assessment. Neurology. 2001;56:S6-10.
7. Green J, Morris JC, Sandson J, et al. Progressive aphasia: a precursor of global dementia? Neurology. 1990;40:423-429.
8. Weintraub S, Rubin NP, Mesulam MM. Primary progressive aphasia. Longitudinal course, neuropsychological profile, and language features. Arch Neurol. 1990;47:1329-1335.
9. Snowden JS, Neary D, Mann DM. Frontotemporal dementia. Br J Psychiatry. 2002;180:140-143.
10. Johnson JK, Diehl J, Mendez MF, et al. Frontotemporal lobar degeneration: demographic characteristics of 353 patients. Arch Neurol. 2005;62:925-930.
11. McKhann G, Drachman D, Folstein M, et al. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's disease. Neurology. 1984;34:939-944.
12. Graham A, Davies R, Xuereb J, et al. Pathologically proven frontotemporal dementia presenting with severe amnesia. Brain. 2005;128:597-605.
13. Johnson JK, Head E, Kim R, et al. Clinical and pathological evidence for a frontal variant of Alzheimer disease. Arch Neurol. 1999;56:1233-1239.
14. Convit A, de Leon MJ, Golomb J, et al. Hippocampal atrophy in early Alzheimer's disease: anatomic specificity and validation. Psychiatr Q. 1993;64:371-387.
15. Jack CR Jr., Petersen RC, Xu YC, et al. Medial temporal atrophy on MRI in normal aging and very mild Alzheimer's disease. Neurology. 1997;49:786-794.
16. Lehericy S, Baulac M, Chiras J, et al. Amygdalohippocampal MR volume measurements in the early stages of Alzheimer disease. AJNR Am J Neuroradiol. 1994;15:929-937.
17. Callen DJ, Black SE, Gao F, et al. Beyond the hippocampus: MRI volumetry confirms widespread limbic atrophy in AD. Neurology. 2001;57:1669-1674.
18. Cuenod CA, Denys A, Michot JL, et al. Amygdala atrophy in Alzheimer's disease. An in vivo magnetic resonance imaging study. Arch Neurol. 1993;50:941-945.
19. Laakso MP, Partanen K, Lehtovirta M, et al. MRI of amygdala fails to diagnose early Alzheimer's disease. Neuroreport. 1995;6:2414-2418.
20. Xu Y, Jack CR Jr, O'Brien PC, et al. Usefulness of MRI measures of entorhinal cortex versus hippocampus in AD. Neurology. 2000;54:1760-1767.
21. Grundman M, Jack CR Jr, Petersen RC, et al. Hippocampal volume is associated with memory but not monmemory cognitive performance in patients with mild cognitive impairment. J Mol Neurosci. 2003;20:241-248.
22. Frisoni GB, Laakso MP, Beltramello A, et al. Hippocampal and entorhinal cortex atrophy in frontotemporal dementia and Alzheimer's disease. Neurology. 1999;52:91-100.
23. Boccardi M, Pennanen C, Laakso MP, et al. Amygdaloid atrophy in frontotemporal dementia and Alzheimer's disease. Neurosci Lett. 2002;335:139-143.
24. Chan D, Fox NC, Scahill RI, et al. Patterns of temporal lobe atrophy in semantic dementia and Alzheimer's disease. Ann Neurol. 2001;49:433-442.
25. Galton CJ, Patterson K, Graham K, et al. Differing patterns of temporal atrophy in Alzheimer's disease and semantic dementia. Neurology. 2001;57:216-225.
26. Whitwell JL, Sampson EL, Watt HC, et al. A volumetric magnetic resonance imaging study of the amygdala in frontotemporal lobar degeneration and Alzheimer's disease. Dement Geriatr Cogn Disord. 2005;20:238-244.
27. Frisoni GB, Beltramello A, Geroldi C, et al. Brain atrophy in frontotemporal dementia. J Neurol Neurosurg Psychiatry. 1996;61:157-165.
28. Wright IC, McGuire PK, Poline JB, et al. A voxel-based method for the statistical analysis of gray and white matter density applied to schizophrenia. Neuroimage. 1995;2:244-252.
29. Ashburner J, Friston KJ. Voxel-based morphometry-the methods. Neuroimage. 2000;11:805-821.
30. Friston KJ, Holmes AP, Worsley KJ, et al. Statistical parametric maps in functional imaging: a general linear approach. Hum Brain Mapp. 1995;2:189-210.
31. Kassubek J, Juengling FD, Kioschies T, et al. Topography of cerebral atrophy in early Huntington's disease: a voxel based morphometric MRI study. J Neurol Neurosurg Psychiatry. 2004;75:213-220.
32. Burton EJ, Karas G, Paling SM, et al. Patterns of cerebral atrophy in dementia with Lewy bodies using voxel-based morphometry. Neuroimage. 2002;17:618-630.
33. Prinster A, Quarantelli M, Orefice G, et al. Grey matter loss in relapsing-remitting multiple sclerosis: a voxel-based morphometry study. Neuroimage. 2006;29:859-867.
34. Matharu MS, Good CD, May A, et al. No change in the structure of the brain in migraine: a voxel-based morphometric study. Eur J Neurol. 2003;10:53-57.
35. Job DE, Whalley HC, McConnell S, et al. Structural gray matter differences between first-episode schizophrenics and normal controls using voxel-based morphometry. Neuroimage. 2002;17:880-889.
36. Mazziotta JC, Toga AW, Evans A, et al. A probabilistic atlas of the human brain: theory and rationale for its development. The International Consortium for Brain Mapping (ICBM). Neuroimage. 1995;2:89-101.
37. Talairach J, Tournoux P. Co-planar Stereotaxic Atlas of the Human Brain. Stuttgart: Thieme Verlag; 1988.
38. Good CD, Johnsrude IS, Ashburner J, et al. A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage. 2001;14:21-36.
39. Salmond CH, Ashburner J, Vargha-Khadem F, et al. Distributional assumptions in voxel-based morphometry. Neuroimage. 2002;17:1027-1030.
40. Gitelman DR, Ashburner J, Friston KJ, et al. Voxel-based morphometry of herpes simplex encephalitis. Neuroimage. 2001;13:623-631.
41. Muzik O, Chugani DC, Juhasz C, et al. Statistical parametric mapping: assessment of application in children. Neuroimage. 2000;12:538-549.
42. Scahill R. MRI in Alzheimer's disease and related disorders: the application of statistical tests of difference to serial and cross-sectional imaging to improve diagnosis and progression measurement. Institute of Neurology. London: University College London; 2003:385.
43. Senjem ML, Gunter JL, Shiung MM, et al. Comparison of different methodological implementations of voxel-based morphometry in neurodegenerative disease. Neuroimage. 2005;26:600-608.
44. Fein G, Landman B, Tran H, et al. Statistical parametric mapping of brain morphology: sensitivity is dramatically increased by using brain-extracted images as inputs. Neuroimage. 2006;30:1187-1195.
45. Boaokstein FL. "Voxel-based morphometry" should not be used with imperfectly registered images. Neuroimage. 2001;14:1454-1462.
46. Good CD, Scahill RI, Fox NC, et al. Automatic differentiation of anatomical patterns in the human brain: validation with studies of degenerative dementias. Neuroimage. 2002;17:29-46.
47. Thompson PM, Giedd JN, Woods RP, et al. Growth patterns in the developing brain detected by using continuum mechanical tensor maps. Nature. 2000;404:190-193.
48. Studholme C, Cardenas V, Blumenfeld R, et al. Deformation tensor morphometry of semantic dementia with quantitative validation. Neuroimage. 2004;21:1387-1398.
49. Brenneis C, Wenning GK, Egger KE, et al. Basal forebrain atrophy is a distinctive pattern in dementia with Lewy bodies. Neuroreport. 2004;15:1711-1714.
50. Burton EJ, McKeith IG, Burn DJ, et al. Cerebral atrophy in Parkinson's disease with and without dementia: a comparison with Alzheimer's disease, dementia with Lewy bodies and controls. Brain. 2004;127:791-800.
51. Rombouts SA, Barkhof F, Witter MP, et al. Unbiased whole-brain analysis of gray matter loss in Alzheimer's disease. Neurosci Lett. 2000;285:231-233.
52. Braak H, Braak E. Evolution of the neuropathology of Alzheimer's disease. Acta Neurol Scand Suppl. 1996;165:3-12.
53. Fox NC, Warrington EK, Freeborough PA, et al. Presymptomatic hippocampal atrophy in Alzheimer's disease. A longitudinal MRI study. Brain. 1996;119(pt 6):2001-2007.
54. Baron JC, Chetelat G, Desgranges B, et al. In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer's disease. Neuroimage. 2001;14:298-309.
55. Minoshima S, Giordani B, Berent S, et al. Metabolic reduction in the posterior cingulate cortex in very early Alzheimer's disease. Ann Neurol. 1997;42:85-94.
56. Boxer AL, Rankin KP, Miller BL, et al. Cinguloparietal atrophy distinguishes Alzheimer disease from semantic dementia. Arch Neurol. 2003;60:949-956.
57. Kawachi T, Ishii K, Sakamoto S, et al. Comparison of the diagnostic performance of FDG-PET and VBM-MRI in very mild Alzheimer's disease. Eur J Nucl Med Mol Imaging. 2006;33:801-809.
58. Ishii K, Kawachi T, Sasaki H, et al. Voxel-based morphometric comparison between early-and late-onset mild Alzheimer's disease and assessment of diagnostic performance of z score images. AJNR Am J Neuroradiol. 2005;26:333-340.
59. Ishii K, Sasaki H, Kono AK, et al. Comparison of gray matter and metabolic reduction in mild Alzheimer's disease using FDG-PET and voxel-based morphometric MR studies. Eur J Nucl Med Mol Imaging. 2005;32:959-963.
60. Ohnishi T, Matsuda H, Tabira T, et al. Changes in brain morphology in Alzheimer disease and normal aging: is Alzheimer disease an exaggerated aging process? AJNR Am J Neuroradiol. 2001;22:1680-1685.
61. Busatto GF, Garrido GE, Almeida OP, et al. A voxel-based morphometry study of temporal lobe gray matter reductions in Alzheimer's disease. Neurobiol Aging. 2003;24:221-231.
62. Zahn R, Buechert M, Overmans J, et al. Mapping of temporal and parietal cortex in progressive nonfluent aphasia and Alzheimer's disease using chemical shift imaging, voxel-based morphometry and positron emission tomography. Psychiatry Res. 2005;140:115-131.
63. Matsuda H, Kitayama N, Ohnishi T, et al. Longitudinal evaluation of both morphologic and functional changes in the same individuals with Alzheimer's disease. J Nucl Med. 2002;43:304-311.
64. Frisoni GB, Testa C, Zorzan A, et al. Detection of grey matter loss in mild Alzheimer's disease with voxel based morphometry. J Neurol Neurosurg Psychiatry. 2002;73:657-664.
65. Karas GB, Burton EJ, Rombouts SA, et al. A comprehensive study of gray matter loss in patients with Alzheimer's disease using optimized voxel-based morphometry. Neuroimage. 2003;18:895-907.
66. Barber R, McKeith I, Ballard C, et al. Volumetric MRI study of the caudate nucleus in patients with dementia with Lewy bodies, Alzheimer's disease, and vascular dementia. J Neurol Neurosurg Psychiatry. 2002;72:406-407.
67. Grossman M, McMillan C, Moore P, et al. What's in a name: voxel-based morphometric analyses of MRI and naming difficulty in Alzheimer's disease, frontotemporal dementia and corticobasal degeneration. Brain. 2004;127:628-649.
68. Frisoni GB, Testa C, Sabattoli F, et al. Structural correlates of early and late onset Alzheimer's disease: voxel based morphometric study. J Neurol Neurosurg Psychiatry. 2005;76:112-114.
69. Jacobs D, Sano M, Marder K, et al. Age at onset of Alzheimer's disease: relation to pattern of cognitive dysfunction and rate of decline. Neurology. 1994;44:1215-1220.
70. Harvey RJ, Rossor MN. Does early-onset Alzheimer disease constitute a distinct subtype? The contribution of molecular genetics. Alzheimer Dis Assoc Disord. 1995;9(suppl 1):S7-S13.
71. Janssen JC, Beck JA, Campbell TA, et al. Early onset familial Alzheimer's disease: mutation frequency in 31 families. Neurology. 2003;60:235-239.
72. Petersen RC. Mild cognitive impairment as a diagnostic entity. J Intern Med. 2004;256:183-194.
73. Chetelat G, Desgranges B, De La Sayette V, et al. Mapping gray matter loss with voxel-based morphometry in mild cognitive impairment. Neuroreport. 2002;13:1939-1943.
74. Karas GB, Scheltens P, Rombouts SA, et al. Global and local gray matter loss in mild cognitive impairment and Alzheimer's disease. Neuroimage. 2004;23:708-716.
75. Pennanen C, Testa C, Laakso MP, et al. A voxel based morphometry study on mild cognitive impairment. J Neurol Neurosurg Psychiatry. 2005;76:11-14.
76. Hirata Y, Matsuda H, Nemoto K, et al. Voxel-based morphometry to discriminate early Alzheimer's disease from controls. Neurosci Lett. 2005;382:269-274.
77. Bell-McGinty S, Lopez OL, Meltzer CC, et al. Differential cortical atrophy in subgroups of mild cognitive impairment. Arch Neurol. 2005;62:1393-1397.
78. Petersen RC, Parisi JE, Dickson DW, et al. Neuropathologic features of amnestic mild cognitive impairment. Arch Neurol. 2006;63:665-672.
79. Chetelat G, Landeau B, Eustache F, et al. Using voxel-based morphometry to map the structural changes associated with rapid conversion in MCI: a longitudinal MRI study. Neuroimage. 2005;27:934-946.
80. Williams GB, Nestor PJ, Hodges JR. Neural correlates of semantic and behavioural deficits in frontotemporal dementia. Neuroimage. 2005;24:1042-1051.
81. Pick A. Uber Einen Weiteren Symptomenkomplex In Rahmen Der Dementia Senilis Bedingt Durch Umschirebene Starkere Hirnatrophie (gemischte Apraxie). Monatsschrift Fur Psychiatrie Und Neurologie. 1906;19:97-108.
82. Mummery CJ, Patterson K, Price CJ, et al. A voxel-based morphometry study of semantic dementia: relationship between temporal lobe atrophy and semantic memory. Ann Neurol. 2000;47:36-45.
83. Rosen HJ, Kramer JH, Gorno-Tempini ML, et al. Patterns of cerebral atrophy in primary progressive aphasia. Am J Geriatr Psychiatry. 2002;10:89-97.
84. Rosen HJ, Gorno-Tempini ML, Goldman WP, et al. Patterns of brain atrophy in frontotemporal dementia and semantic dementia. Neurology. 2002;58:198-208.
85. Gorno-Tempini ML, Dronkers NF, Rankin KP, et al. Cognition and anatomy in three variants of primary progressive aphasia. Ann Neurol. 2004;55:335-346.
86. Sonty SP, Mesulam MM, Thompson CK, et al. Primary progressive aphasia: PPA and the language network. Ann Neurol. 2003;53:35-49.
87. Rosso SM, Roks G, Stevens M, et al. Complex compulsive behaviour in the temporal variant of frontotemporal dementia. J Neurol. 2001;248:965-970.
88. Halpern CH, Glosser G, Clark R, et al. Dissociation of numbers and objects in corticobasal degeneration and semantic dementia. Neurology. 2004;62:1163-1169.
89. Graff-Radford NR, Damasio AR, Hyman BT, et al. Progressive aphasia in a patient with Pick's disease: a neuropsychological, radiologic, and anatomic study. Neurology. 1990;40:620-626.
90. Rossor MN, Revesz T, Lantos PL, et al. Semantic dementia with ubiquitin-positive tau-negative inclusion bodies. Brain. 2000;123(pt 2):267-276.
91. Harasty JA, Halliday GM, Code C, et al. Quantification of cortical atrophy in a case of progressive fluent aphasia. Brain. 1996;119(pt 1):181-190.
92. Whitwell JL, Anderson VM, Scahill RI, et al. Longitudinal patterns of regional change on volumetric MRI in frontotemporal lobar degeneration. Dement Geriatr Cogn Disord. 2004;17:307-310.
93. Thompson SA, Patterson K, Hodges JR. Left/right asymmetry of atrophy in semantic dementia: behavioral-cognitive implications. Neurology. 2003;61:1196-1203.
94. Tyrrell PJ, Warrington EK, Frackowiak RS, et al. Heterogeneity in progressive aphasia due to focal cortical atrophy. A clinical and PET study. Brain. 1990;113(pt 5):1321-1336.
95. Evans JJ, Heggs AJ, Antoun N, et al. Progressive prosopagnosia associated with selective right temporal lobe atrophy. A new syndrome? Brain. 1995;118(pt 1):1-13.
96. Edwards-Lee T, Miller BL, Benson DF, et al. The temporal variant of frontotemporal dementia. Brain. 1997;120(pt 6):1027-1040.
97. Gorno-Tempini ML, Rankin KP, Woolley JD, et al. Cognitive and behavioral profile in a case of right anterior temporal lobe neurodegeneration. Cortex. 2004;40:631-644.
98. Nestor PJ, Graham NL, Fryer TD, et al. Progressive non-fluent aphasia is associated with hypometabolism centred on the left anterior insula. Brain. 2003;126:2406-2418.
99. Turner RS, Kenyon LC, Trojanowski JQ, et al. Clinical, neuroimaging, and pathologic features of progressive nonfluent aphasia. Ann Neurol. 1996;39:166-173.
100. Abe K, Ukita H, Yanagihara T. Imaging in primary progressive aphasia. Neuroradiology. 1997;39:556-559.
101. Fukui T, Kertesz A. Volumetric study of lobar atrophy in Pick complex and Alzheimer's disease. J Neurol Sci. 2000;174:111-121.
102. Gorno-Tempini ML, Murray RC, Rankin KP, et al. Clinical, cognitive and anatomical evolution from nonfluent progressive aphasia to corticobasal syndrome: a case report. Neurocase. 2004;10:426-436.
103. Dronkers NF. A new brain region for coordinating speech articulation. Nature. 1996;384:159-161.
104. Josephs KA, Duffy JR, Strand EA, et al. Clinicopathological and imaging correlates of progressive aphasia and apraxia of speech. Brain. 2006;129:1385-1398.
105. Hodges JR, Davies RR, Xuereb JH, et al. Clinicopathological correlates in frontotemporal dementia. Ann Neurol. 2004;56:399-406.
106. Kertesz A, McMonagle P, Blair M, et al. The evolution and pathology of frontotemporal dementia. Brain. 2005;128:1996-2005.
107. Josephs KA, Petersen RC, Knopman DS, et al. Clinicopathologic analysis of frontotemporal and corticobasal degenerations and PSP. Neurology. 2006;66:41-48.
108. Galton CJ, Patterson K, Xuereb JH, et al. Atypical and typical presentations of Alzheimer's disease: a clinical, neuropsychological, neuroimaging and pathological study of 13 cases. Brain. 2000;123(pt 3):484-498.
109. Boccardi M, Sabattoli F, Laakso MP, et al. Frontotemporal dementia as a neural system disease. Neurobiol Aging. 2005;26:37-44.
110. Snowden JS, Bathgate D, Varma A, et al. Distinct behavioural profiles in frontotemporal dementia and semantic dementia. J Neurol Neurosurg Psychiatry. 2001;70:323-332.
111. Liu W, Miller BL, Kramer JH, et al. Behavioral disorders in the frontal and temporal variants of frontotemporal dementia. Neurology. 2004;62:742-748.
112. Cummings JL, Mega M, Gray K, et al. The Neuropsychiatric Inventory: comprehensive assessment of psychopathology in dementia. Neurology. 1994;44:2308-2314.
113. Gee J, Ding L, Xie Z, et al. Alzheimer's disease and frontotemporal dementia exhibit distinct atrophy-behavior correlates: a computer-assisted imaging study. Acad Radiol. 2003;10:1392-1401.
114. Rosen HJ, Allison SC, Schauer GF, et al. Neuroanatomical correlates of behavioural disorders in dementia. Brain. 2005;128:2612-2625.
115. Lavenu I, Pasquier F, Lebert F, et al. Perception of emotion in frontotemporal dementia and Alzheimer disease. Alzheimer Dis Assoc Disord. 1999;13:96-101.
116. Rosen HJ, Wilson MR, Schauer GF, et al. Neuroanatomical correlates of impaired recognition of emotion in dementia. Neuropsychologia. 2006;44:365-373.
117. Grossman M, Farmer J, Leight S, et al. Cerebrospinal fluid profile in frontotemporal dementia and Alzheimer's disease. Ann Neurol. 2005;57:721-729.
118. Boccardi M, Sabattoli F, Testa C, et al. APOE and modulation of Alzheimer's and frontotemporal dementia. Neurosci Lett. 2004;356:167-170.
119. Apolipoprotein E genotyping in Alzheimer's disease. National Institute on Aging/Alzheimer's Association Working Group. Lancet. 1996;347:1091-1095.
120. Fleisher A, Grundman M, Jack CR Jr, et al. Sex, apolipoprotein E epsilon 4 status, and hippocampal volume in mild cognitive impairment. Arch Neurol. 2005;62:953-957.
121. Galasko D, Hansen LA, Katzman R, et al. Clinical-neuropathological correlations in Alzheimer's disease and related dementias. Arch Neurol. 1994;51:888-895.
122. Varma AR, Snowden JS, Lloyd JJ, et al. Evaluation of the NINCDS-ADRDA criteria in the differentiation of Alzheimer's disease and frontotemporal dementia. J Neurol Neurosurg Psychiatry. 1999;66:184-188.
123. Hodges JR, Davies R, Xuereb J, et al. Survival in frontotemporal dementia. Neurology. 2003;61:349-354.
124. Knopman DS, Mastri AR, Frey WH 2nd, et al. Dementia lacking distinctive histologic features: a common non-Alzheimer degenerative dementia. Neurology. 1990;40:251-256.
125. Josephs KA, Holton JL, Rossor MN, et al. Neurofilament inclusion body disease: a new proteinopathy? Brain. 2003;126:2291-2303.
126. Whitwell JL, Warren JD, Josephs KA, et al. Voxel-based morphometry in tau-positive and tau-negative frontotemporal lobar degenerations. Neurodegener Dis. 2004;1:225-230.
127. Mann DM, South PW. The topographic distribution of brain atrophy in frontal lobe dementia. Acta Neuropathol (Berl). 1993;85:334-340.
128. Broe M, Hodges JR, Schofield E, et al. Staging disease severity in pathologically confirmed cases of frontotemporal dementia. Neurology. 2003;60:1005-1011.
129. Kril JJ, Halliday GM. Clinicopathological staging of frontotemporal dementia severity: correlation with regional atrophy. Dement Geriatr Cogn Disord. 2004;17:311-315.
130. Whitwell JL, Josephs KA, Rossor MN, et al. Magnetic resonance imaging signatures of tissue pathology in frontotemporal dementia. Arch Neurol. 2005;62:1402-1408.
131. Whitwell JL, Jack CR Jr, Senjem ML, et al. Patterns of atrophy in pathologically confirmed FTLD with and without motor neuron degeneration. Neurology. 2006;66:102-104.
132. Jeong Y, Park KC, Cho SS, et al. Pattern of glucose hypometabolism in frontotemporal dementia with motor neuron disease. Neurology. 2005;64:734-736.
133. Scahill RI, Schott JM, Stevens JM, et al. Mapping the evolution of regional atrophy in Alzheimer's disease: unbiased analysis of fluid-registered serial MRI. Proc Natl Acad Sci U S A. 2002;99:4703-4707.
134. Christensen GE, Rabbitt RD, Miller MI. Deformable templates using large deformation kinematics. IEEE Trans Image Proc. 1996;5:1435-1447.
135. Freeborough PA, Fox NC. Modeling brain deformations in Alzheimer disease by fluid registration of serial 3D MR images. J Comput Assist Tomogr. 1998;22:838-843.
Keywords:
Alzheimer disease; frontotemporal lobar degeneration; magnetic resonance imaging; voxel-based morphometry
© 2005 Lippincott Williams & Wilkins, Inc.