We have two main findings. First, FTD patients exhibited more WM damage than AD patients at mild abnormality levels (i.e., small T-scores). The difference vanished gradually for more severe abnormality levels (i.e., larger T-scores). Second, FTD patients had greater GM loss and WM damage relative to GM hypoperfusion, although the differences of damage between modalities gradually vanished with increasing levels of abnormality. In contrast to FTD, AD patients had equal damage in all three modalities, irrespective of the level of abnormality. Taken together, the results suggest that FTD and AD differ in amount of WM and GM structural and functional damages, in addition to their characteristic regional patterns of brain alterations than healthy controls.
Our first main finding of greater WM damage in FTD than AD at mild levels of abnormality suggests that WM may be more sensitive to the pathology of FTD than to the pathology of AD at their early disease stages. The hallmark of FTD is tauopathies or ubiquitin immunoreactive inclusions by the presences of neuronal and glial inclusions [48
] in gray and white matter. WM pathologies in FTD have been reported with astrocytic gliosis and oligodendroglial apoptosis, which may ultimately result in axonal degeneration [50
]. A recent study [54
] also reported that oligodendroglial pathology can be predominant in FTD despite severe GM damage. On the other hand, degeneration of frontostriatal networks, which is a characteristic feature of FTD [26
], suggests that WM denegation in anterior brain could also be a primary FTD pathology. In contrast to FTD, the hallmark of AD is the deposition of amyloid plaques and neurofibrillary tangles that are associated with loss of neurons and synapses [55
]. WM pathology in AD has been suggested to occur secondarily to GM pathology and may include reduction of myelin, axons, and oligodendrocytes [57
]. A vascular origin of WM pathology in AD has also been suggested [59
]. However, WM pathologies in AD are usually considered mild and potentially reversible [60
]. Interestingly, our data showed that differences in WM damage between AD and FTD disappeared at higher abnormality levels. It is possible to suggest that AD and FTD undergo similar WM pathologies at a severe brain damage level. One explanation is that severe WM abnormalities are an outcome of irreversible vascular damage, such as appearance of WMSH, which affects AD and FTD similarly [61
]. Another possible explanation is that WM changes resulting from degeneration of corticocortical connections may inevitably occur in both AD and FTD.
In addition, we found a trend of more GM loss in FTD than AD at moderate to severe levels of abnormality. This observation is consistent with histopathological studies showing substantial loss of spindle neurons in the cortex in FTD but not in AD [62
]. The finding is also in agreement with several MRI studies [29
] which compared FTD and AD directly, showing regionally greater GM loss in FTD when compared to AD but no greater GM loss in AD when compared to FTD. Similarly, several longitudinal studies reported greater rates of GM atrophy in FTD when compared to AD [65
]. However, some MRI studies [26
] reported greater GM loss in AD compared to FTD, though the extent of the GM losses varied regionally. The discrepancy may result from notorious difficulties to adequately match impairment severity in FTD and AD given substantial differences in symptomatology, although our current study attempted to match the severities (such as the measures of CDR, MMSE) of AD and FTD groups as closely as possible.
The second major finding that GM loss and WM damage exceed perfusion damage in FTD may be explained by loss of brain tissue other than neurons contributing to GM atrophy in FTD. Information about differential loss of various types of brain tissue may be particularly relevant in earlier and potentially reversible stages of the disease. Our findings are supported by several histopathological studies demonstrating that the earliest cellular changes in FTD occur in astrocytes without neuron loss and that neuronal damage becomes more prominent in later stages of the disease [50
]. Therefore, at mild abnormality levels of FTD, the surviving neurons may still function normally as reflected by the normal levels of perfusion. Our findings are consistent with other studies carried out in our laboratory [69
] and on different cohorts of FTD and AD patients which have implied dissociation between GM atrophy and perfusion. Specifically, these studies indicated that FTD was associated with greater GM atrophy in absence of significant reduction of perfusion. These results contrast the findings in AD, where the abnormalities across the three modalities were all similar.
Our voxelwise analysis demonstrates the regional patterns of GM loss, hypoperfusion, and reduced WM FA in this sample of AD and FTD patients are consistent with the disease-specific patterns that have been reported in previous MRI studies of brain structure [14
], perfusion [5
], and DTI [25
]. Furthermore, the similarity between the regional patterns of GM and WM alterations which appear in the same lobe together for each disease, implies that WM degradation mirrors that of GM damages and is consistent with previous reports [74
]. Taken together, these results can demonstrate the well-documented pathological bases of AD [56
] and FTD [68
The new joint analysis approach allows for the investigation of different abnormalities across multiple MRI modalities such as structural, perfusion and diffusion MRI between groups. The approach can be used to test whether groups differ with respect to a single MRI measure as well as multiple measures. The approach was augmented by nonparametric statistical tests via permutations and carried out across a range of T-scores to reduce measurement bias toward the various brain conditions. The concept can be expanded in principle to conduct a voxelwise joint analysis of multiple MRI measures to determine regional variations in GM loss and hypoperfusion. Other statistical methods for joint analyses of multiple image modalities such as joint independent component analysis (jICA) [75
] may provide alternative solutions. The findings with multimodal MRI could potentially be useful to improve the design of AD and FTD clinical trials involving MRI. First, correlations across the MRI measures, potentially boosting sensitivity and specificity, could lead to reduced sample sizes. Second, the finding revealed that FTD presents more white matter involvement relative to AD, thus providing a new biological feature of FTD, and could be used to relax the need to match disease severity in studies recruiting AD and FTD patients.
Limitations of the current study include a small sample size that was reliant on clinical diagnoses which were not sufficiently confirmed by autopsies. Therefore, confidence in the generalization of the results is limited, and potential misdiagnosis of patients may have resulted in spurious findings. Furthermore, we cannot completely rule out that other factors than disease etiology, such as genetic profiles, duration of symptoms, and cardiovascular conditions, which contributed to MRI differences between the patients and thus contaminated the findings. Second, diffusion encoding was limited to the minimum of 6 directions at the time this protocol was initiated, although it is known that many more encoding directions improve the characterization of diffusion such as fewer ambiguities in regions of crossing fibers and better spatial invariance of the noise pattern. Therefore, fiber crossings and the DTI noise pattern may potentially mimic regional differences in FA between these groups. Third, we ignored relationships between brain regions in our joint analysis and therefore under-utilized information from the multimodal MRI data. A more powerful statistical framework [76
] that takes spatial relations between multivariate measures into account may provide more power. Finally, the data was artificially scaled to provide a uniform resolution for all MRI modalities, which may have induced a spatial bias as well as altered selectively the sensitivity of each modality. Other approaches that do not require a uniform resolution but can operate on variable spatial scales, such as information theoretic formalisms [77
], may lead to differences in results.
In conclusion, our findings suggest that FTD and AD differ regarding their impacts on WM and GM structural and functional abnormalities, in addition to differences between their characteristic regional patterns of brain alterations. Furthermore, the joint assessment of multimodal MRI measures employed in this study has potential value to improve the differential diagnosis between FTD and AD.