Quantitative perfusion obtained with cASL MRI in patients with autopsy or CSF biomarkers revealed doubly dissociated patterns of hypoperfusion in AD and FTLD. Significantly reduced frontal CBF was seen in FTLD, and significant temporal-parietal hypoperfusion was found in AD. Moreover, we observed distinct areas of significant hyperperfusion in FTLD and AD. These observations underline the potential usefulness of ASL in distinguishing between FTLD and AD.
Direct comparisons of ASL with 15
O-PET show that MRI and PET methods for measuring CBF are highly comparable.25
Moreover, an MRI-based blood flow technique like ASL is more widely available than PET because it can be obtained with standard clinical MRI equipment, ASL is less expensive, and ASL avoids exposure to ionizing radiation. Importantly, CBF techniques seem to validly reflect neuronal glucose metabolism as measured by FDG-PET because FDG-PET and 15
O-PET yield similar results in healthy seniors and in patients with neurodegenerative conditions.26
Using ASL to measure CBF, we found significantly reduced perfusion in inferior, dorsolateral, medial, and insular regions of the frontal lobe in FTLD. These areas correspond to regions of known disease in FTLD, based on regional autopsy findings.4
A similar anatomical distribution of hypoperfusion was seen in the subgroup of patients with FTLD with autopsy-confirmed disease. One previous study of FTLD with ASL demonstrated superior frontal hypoperfusion,10
partially corresponding to our findings. The discrepancies between this study and our findings may have been due in part to partial coverage of the brain and use of a less sensitive pulsed ASL sequence in prior work. The regions of hypoperfusion in both studies correspond to areas of significant atrophy in structural MRI studies of FTLD.4,5
In the present study, we used 2 approaches to correct for morphometric effects on regional CBF. The PVE correction adjusted for the partial volume of white matter in gray matter voxels10
and the Jacobian-based spatial-transformation correction adjusted for the compression or expansion of brain regions due to normalization to the local template.18
The latter step was reported in a study that found more significant hyperperfusion in hippocampus in patients with mild AD after correction.23
However, similar effects but with a slightly reduced significance level were obtained if the second Jacobian-based correction was not applied. Accordingly, the observed hypoperfusion likely reflects functional changes beyond those that can be attributed to atrophy alone. Furthermore, PVE cannot easily explain the observed regions of hyperperfusion.
Patients with FTLD showed significant hyperperfusion in parietal cortex and precuneus. These areas tend to show significant disease only late in the course of FTLD, whereas the patients participating in this study were relatively early in their disease course. Hyperperfusion may be possible because of preserved connectivity in unaffected intrinsic connectivity networks in FTLD.27
The significance of observed hyperperfusion remains to be established. One possibility is that there is a dissociation between vascular perfusion measured by ASL and neuronal metabolic consumption of substrate. Partial perfusion-metabolic uncoupling can occur if there is a change in vascular diameter or a mitochondrial disorder of neuronal metabolism, but neither is associated with FTLD. Another possibility is related to the concept of “reserve” or “compensation” in neurodegenerative conditions like FTLD.28
From this perspective, neuronal integrity in these areas is beginning to be stressed by the incipient accumulation of abnormal neuronal histopathologic inclusions, and up-regulation of regional cortical activity may attempt to compensate in part for these early challenges to regional neuronal functioning. A related account implicates increased activity in response to partial deafferentation as a result of disease in frontal areas that compromises projections to these hyperperfused regions. This would be consistent with imaging observations of diffusion tensor imaging defects in WM projections in FTLD.29
It is difficult to assess these hypotheses because direct tissue evidence of modest disease burden in hyperperfused areas is difficult to obtain. Indirect evidence can be acquired in future work that assesses the fate of CBF in hyperperfused areas in longitudinal ASL observations.
Areas of significant hypoperfusion in AD involved a different anatomical distribution than areas showing significantly reduced CBF in FTLD. Patients with AD thus showed significant hypoperfusion in posterior brain regions, including parietal, temporal, and precuneus regions. This corresponds to areas of significant histopathologic burden in autopsy-confirmed AD,30
and hypoperfusion was observed in these areas in autopsy-confirmed cases of AD. Previous ASL studies of AD have shown a similar anatomical pattern of hypoperfusion.10,31–34
We did not observe medial temporal hypoperfusion, as reported in some of this work. Cross-sectional studies thus have suggested that patients with very mild AD and mild cognitive impairment (MCI) may have hippocampal hyperperfusion.32,33
This would be consistent with other work showing increased medial temporal activation during performance of memory-related tasks in MCI35
and hippocampal neuronal hypertrophy in asymptomatic AD.36
One longitudinal study related hippocampal activation to subsequent cognitive decline in AD.37
There may have been some averaging across cases with hypoperfusion and hyperperfusion of the medial temporal region in our series. Moreover, because some patients classified as AD in our series had clinical features more consistent with bvFTD or PPA, there may have been little change in medial temporal perfusion early in the course of disease. Additional work is needed in larger groups of patients to establish more precisely the fate of medial temporal perfusion in mild AD.
We also observed areas of significant cortical hyperperfusion in AD. This included areas of the frontal lobe and lateral temporal lobe, regions where disease emerges later in the course of AD than the medial temporal lobe.30
Cortical hyperperfusion has been described in several previous studies.32,33
This may be related in part to a similar mechanism implicated in the neuronal hypertrophy observed in the anterior cingulate of asymptomatic AD.36
We also cannot rule out the contribution of partial deafferentation due to degraded WM projections from diseased areas.38,39
We observed a double dissociation between AD and FTLD, emphasizing the potential usefulness of ASL as a diagnostic modality. PET studies have repeatedly demonstrated frontal hypometabolism in bvFTD and parietal hypometabolism in AD, as we observed, although comparative studies such as ours are rare. Moreover, one recent PET study described parietal-occipital hypermetabolism in bvFTD and frontotemporal hypermetabolism in AD when adjusting for global glucose metabolism,40
consistent with our findings using ASL.
Several limitations in our study should be kept in mind. Regional perfusion patterns may depend in part on disease duration, and the slightly longer disease duration in FTLD may have confounded our analysis. CSF AD biomarkers were used as surrogate markers of pathology rather than true neuropathologic analysis for many patients. Despite the high sensitivity and specificity of CSF AD biomarkers, there is currently no biomarker that positively identifies FTLD pathology, and patients may have been misclassified by CSF biomarkers alone. A definitive study is needed in the future to determine the usefulness of ASL as a biomarker using pathologically confirmed cases of AD and FTLD. With these caveats in mind, the double dissociation between AD and FTLD that we observed is consistent with the hypothesis that simultaneous analysis of susceptible and resistant brain regions can improve the differential diagnosis of underlying pathology in AD and FTLD clinical syndromes.