|Home | About | Journals | Submit | Contact Us | Français|
Accumulating evidence suggests that altered cerebral white matter (WM) influences normal aging, and further that WM degeneration may modulate the clinical expression of Alzheimer’s disease (AD). Here we conducted a study of differences in WM volume across the adult age span and in AD employing a newly developed, automated method for regional parcellation of the subcortical WM that uses curvature landmarks and gray matter (GM)/WM surface boundary information. This procedure measures the volume of gyral WM, utilizing a distance constraint to limit the measurements from extending into the centrum semiovale. Regional estimates were first established to be reliable across two scan sessions in 20 young healthy individuals. Next, the method was applied to a large clinically-characterized sample of 299 individuals including 73 normal older adults and 91 age-matched participants with very mild to mild AD. The majority of measured regions showed a decline in volume with increasing age, with strong effects found in bilateral fusiform, lateral orbitofrontal, superior frontal, medial orbital frontal, inferior temporal, and middle temporal WM. The association between WM volume and age was quadratic in many regions suggesting that WM volume loss accelerates in advanced aging. A number of WM regions were further reduced in AD with parahippocampal, entorhinal, inferior parietal and rostral middle frontal WM showing the strongest AD-associated reductions. There were minimal sex effects after correction for intracranial volume, and there were associations between ventricular volume and regional WM volumes in the older adults and AD that were not apparent in the younger adults. Certain results, such as the loss of WM in the fusiform region with aging, were unexpected and provide novel insight into patterns of age associated neural and cognitive decline. Overall, these results demonstrate the utility of automated regional WM measures in revealing the distinct patterns of age and AD associated volume loss that may contribute to cognitive decline.
Accumulating evidence suggests that processes associated with nondemented aging as well as the degenerative processes of Alzheimer’s disease (AD) have a negative effect on cerebral white matter (WM). The regional predilection for WM degeneration and the contribution of this tissue loss to cognitive and neural dysfunction is of considerable interest, but has been difficult to quantify. The anatomic organization of brain WM is highly complex and not visible on a standard T1 weighted MRI scan. However, WM proximal to a cortical region preferentially contains afferent and efferent fibers associated with that cortical region, and thus, the condition of this adjacent tissue provides information about the integrity of specific neural systems. In the present paper we apply a novel, automated method to explore local WM integrity in normal aging and AD.
Quantification of age and AD associated alterations in WM volume has been a goal in postmortem (Anderson et al., 1983; Hubbard and Anderson, 1981; Meier-Ruge et al., 1992; Miller et al., 1980) as well as early imaging studies using CT (de Leon et al., 1989; Gado et al., 1983; Obara et al., 1994). MRI studies have examined whole brain WM volume, volume of selective WM regions, and abnormal WM signal (Christiansen et al., 1994; de Leeuw et al., 2001; DeCarli et al., 1996; DeCarli et al., 1995; Head et al., 2005; Jernigan et al., 1991a; Jernigan et al., 1991b; Mungas et al., 2005; Mungas et al., 2002; Pfefferbaum et al., 1994; Raz et al., 1997; Rusinek et al., 1991; Salat et al., 1999; Sullivan et al., 1995; Udaka et al., 2002; Wahlund et al., 1994). The majority of these studies generally demonstrate that there is a significant loss of WM volume and an increase in abnormal WM signal with aging (Guttmann et al., 1998; Jernigan et al., 1991a; Raz et al., 1997; Resnick et al., 2000), however, several studies utilizing differing procedures have also demonstrated relatively little change in WM volume in aging and associated disease (Double et al., 1996; Jernigan et al., 1991a; Jernigan et al., 1991b; Obara et al., 1994; Rusinek et al., 1991; Sullivan et al., 1995; Tanabe et al., 1997). These disparate findings could be due to the fact that WM volume loss may have a highly specific temporal and spatial pattern. For example, WM volume is reported to be relatively stable until approximately the fifth decade of life in healthy adults (Bartzokis et al., 2001; Pfefferbaum et al., 1994). Controversially, WM changes are reported to contribute in a complex manner (Burns et al., 2005; DeCarli et al., 1995; Fotenos et al., 2005; Hirono et al., 2000a; Hirono et al., 2000b; Smith et al., 2000; Stout et al., 1996; Tanabe et al., 1997; Tohgi et al., 1998; Waldemar et al., 1994) or not at all (Rusinek et al., 1991; Wahlund et al., 1994) to the clinical expression of AD. For example, Smith and colleagues reported an association between normal WM volume and dementia which was not true for periventricular hyperintensities (Smith et al., 2000). Other studies have demonstrated a significant role for WM signal abnormalities in AD dementia but are consistent with a modulatory effect of symptoms rather than a direct contribution to AD (Burns et al., 2005). Wu et al. (2002), based on analysis of two MR and PET studies, concluded that changes associated with AD and commonly observed WM changes linked to vascular disease are synergistic contributors to dementia (Wu et al., 2002).
More recently developed imaging techniques, such as diffusion imaging, magnetization transfer imaging, spectroscopy, and other advanced MR procedures have contributed additional information about the vulnerability of WM to degenerative processes (Armstrong et al., 2004; Bartzokis et al., 2003; Choi et al., 2005; Head et al., 2004; Madden et al., 2004; Pfefferbaum and Sullivan, 2003; Pfefferbaum et al., 2000; Rose et al., 2000; Rose et al., 2006; Salat et al., 2005a; Salat et al., 2005b; Sandson et al., 1999; Sullivan et al., 2001) and complement studies utilizing histological analysis to understand pathologic mechanisms of WM degeneration (Braak and Braak, 1989; Hashimoto et al., 2003; Hyman et al., 1984; Kovari et al., 2007; Roher et al., 2002; Thal et al., 1998; van de Nes et al., 2002; Webster et al., 2006; Xuereb et al., 2000; Yang et al., 2005; Zhukareva et al., 2002). These studies support the important role of WM deterioration in brain aging and dementia, with some specificity to the regional patterns of changes in WM integrity. For example, DTI studies have clarified an anterior to posterior gradient of age effects on brain WM microstructure (Head et al., 2004; Pfefferbaum et al., 2005; Salat et al., 2005a) and altered diffusion measures in temporal or temporal stem WM with AD, with additional effects reported in frontal, parietal, and callosal regions (Bozzali et al., 2002; Hanyu et al., 1998; Head et al., 2004; Kantarci et al., 2001; Salat et al., 2008).
Volumetric studies of WM have been mostly limited in measuring the entire cerebral WM at a coarse spatial scale; regional variation in morphometry has been less accessible. Relatively few studies have examined alterations in WM volume on a regional or lobar basis but notable exceptions exist (Bartzokis et al., 2001; Bigler et al., 2002; Raz et al., 1997; Salat et al., 1999). These studies have demonstrated regional variation in WM volume loss underscoring the importance of more selective regional volumetric measures. Similarly, voxel based morphometry studies reinforce the fact that WM changes regionally accelerated in particular areas including frontal WM and anterior callosum in normal aging (Brickman et al., 2007; Good et al., 2001) and parahippocampal and temporal in preclinical or clinically diagnosed AD (Li et al., 2008; Stoub et al., 2006). However, there is still relatively limited information on the temporal and regional basis of WM volume loss in these conditions. Thus, a procedure to divide cerebral WM into homologous volumetric regions across individuals on T1 images, which are widely available in MR studies including retrospective studies, would be of high value for addressing questions about age and dementia associated WM degeneration.
The current study utilized detailed morphometric procedures to obtain regional WM volume measurements to determine whether WM degeneration is a global or regionally selective process, and the pattern of that degeneration across the adult age span. We employed a newly developed method for automated parcellation of brain WM using morphometric landmarks and GM/WM boundary surface information. This procedure builds on our prior described and validated cortical parcellation technique (Desikan et al., 2006; Fischl et al., 2004), and extends this labeling to the subcortical WM directly underlying the cortical parcellation. We first applied this novel procedure to a sample of individuals that were imaged on two different occasions within a short time period to examine the test-retest reliability of the regional measurements. We next utilized the technique to examine regional volumetric differences in a large sample of healthy younger (YNG) and nondemented older (OLD) adults, and a group of individuals with AD. These initial descriptive results demonstrate that cerebral WM volume is affected in a regionally-specific manner with nondemented aging as well as in AD, and that WM changes in AD could contribute to the clinical expression of dementia.
High-resolution structural MR scans were obtained from two sets of participants. Twenty young healthy individuals were imaged on two occasions for assessing test-retest reliability of the parcellation method (TRT; n = 8M/12F, mean age = 23.4) and 299 participants for the examination of volumetric alterations in aging and AD. The main sample of 299 individuals was grouped as younger adults (YNG; individuals younger than 60 years; n = 62M/73F, mean age = 26.5), older adults (OLD; individuals 60 years and older; n = 26M/47F, mean age = 76.8), and individuals with AD (matched to OLD; n = 37M/54F, mean age = 77.6, clinical dementia rating [CDR] (Morris, 1993); 63 = 0.5 and 28 = 1; Table 1). The association between age and regional volume was examined across all nondemented individuals, as well as through group comparisons using 60 years of age as the cutoff. This grouping of YNG and OLD was based on prior literature suggesting that WM volume is relatively preserved until late aging and to match the OLD to the AD group.
All OLD and AD were recruited and clinically evaluated through the Washington University Alzheimer’s Disease Research Center (ADRC) as reported previously (Berg et al., 1998; Morris, 1993). The YNG adults were recruited from the community as part of ongoing cognitive studies. Individuals were excluded if they exhibited any neurological or psychiatric conditions that could contribute to dementia. OLD individuals were all clinically-screened to show no signs of even mild cognitive impairment (all CDR 0). Fotenos et al (2005) describe the recruitment characteristics of this sample in detail. Participants consented in accordance with guidelines of the Washington University Human Studies Committee.
These data are openly available to the community via the OASIS project (www.oasis.org;(Marcus et al., 2007)) thanks in part to resources from the Washington University ADRC. This sample overlaps with prior studies examining aging and AD (Andrews-Hanna et al., 2007; Buckner et al., 2005; Dickerson et al., 2008; Fotenos et al., 2008; Fotenos et al., 2005; He et al., 2008; Head et al., 2004; Head et al., 2005; Salat et al., 2004).
Two to five T1-weighted MP-RAGE scans were motion corrected and averaged per participant for image processing on a single scanner (Siemens 1.5-T Vision System, resolution 1 × 1 × 1.25 mm, TR = 9.7 ms, TE = 4 ms, FA = 10°, TI = 20 ms, TD = 500 ms) to create high signal/contrast to noise volumes for each participant. Acquisition parameters were empirically optimized to increase GM/WM/cerebrospinal fluid contrast. Imaging of participants from each group was distributed across time, and there was no overrepresentation of a particular group at any time during imaging.
Cortical surface models were created as described previously (Dale et al., 1999; Desikan et al., 2006; Fischl et al., 1999a; Fischl et al., 2004; Salat et al., 2004). This method utilized intensity and continuity information from the entire three dimensional MR volumes in segmentation and deformation procedures to produce accurate representations of the cortical mantle. The tissue boundaries are determined using spatial intensity gradients and are therefore not simply reliant on absolute signal intensity. After the creation of topologically correct (Fischl et al., 2001) surface models, spherical coordinate procedures (Fischl et al., 1999b) were utilized in cortical surface parcellation which was validated against manual measurements (Desikan et al., 2006; Fischl et al., 2004).
These parcellations were subsequently used to assign a label to the underlying WM by the construction of a Voronoi diagram in the WM voxels based on distance to the nearest cortical parcellation label (Figure 1). Each Voronoi polygon then inherited the label of the parcellation unit, yielding a complete labeling of the cerebral WM. A distance constraint was applied, halting the label expansion after 5mm to result in labeling of the gyral WM and to avoid inclusion of WM from the centrum semiovale and periventricular regions (although certain regions in close proximity to the ventricles did contact this structure), and WM voxels beyond this were unlabeled resulting in ‘cortically associated’ gyral WM volumes. WM signal abnormalities (WMSA; i.e. T1 hypointensities within the WM) were labeled utilizing a probabilistic procedure (Fischl et al., 2002) and included in the total regional WM volume to examine volume changes independent of WMSA burden. However, total WMSA volumes were obtained (as described in (Burns et al., 2005) and the association between this classic imaging measure of WM abnormality and the regional WM volumes was examined for the current study. All regional volumes were corrected for estimated total intracranial volume (eTIV) utilizing an atlas scaling and covariance approach as previously described (Buckner et al., 2004). This method was validated against manual measurements of intracranial volume in prior work (Buckner et al., 2004).
In order to assess the reliability of the WM parcellation procedure, we examined regional volumes from each of the twenty young healthy TRT individuals over two imaging sessions separated by a brief interval during which minimal true biological changes are likely to have occurred. There was a mean delay of 21 days between test and retest sessions (range 1–89 days).
The associations between age and regional WM volume were examined by Pearson’s correlation and simple and polynomial regression to examine whether curvilinear fits significantly explained the data. Group differences in regional WM volume between OLD and AD were examined by analysis of covariance with regional WM volumes as dependent factors and age as a cofactor. Given the statistical power, we used a moderate statistical threshold and only results with p ≤ 0.01 were considered. Interactions with p > 0.01 were removed from the model. Secondary analyses were performed splitting the groups by sex to determine whether observed effects were replicable across men and women. Additional analyses examined the association between ventricular and regional WM volumes to determine whether WM volume was affected independently of this classical metric of neural integrity.
The results for the test-retest error for each structure are presented in Table 2. Test-retest assessment was based on automated measures without additional corrections or processing. Each measure came from a distinct MR session and thus the error estimates includes real-world sources of variance linked to head positioning and scanner instability. Measurements were generally reliable, with the majority of structures within approximately 5% of the total structure. A small number of structures showed greater measurement variability and were thus examined only as exploratory analyses for the current study. The highlighted results come from regions that show good to excellent test-retest reliability. Certain regions, such as the entorhinal WM, increased in reliability when examined as a combined metric with parahippocampal WM. However, we choose to analyze this structure independently from the parahippocampal WM because the effects measured were reliable across independent samples, demonstrating the ability to detect valid changes regardless of the cross-session reliability.
There were strong associations between WM volume and age in most regions examined in the nondemented individuals with effects in bilateral fusiform, superior frontal, medial orbital frontal, inferior temporal, and middle temporal WM (Table 3; Figure 2). Of note, an absence of age-associated reduction was found in bilateral caudal anterior cingulate and transverse temporal WM. The majority of regions were significantly fit by a quadratic function. This fit generally exhibited a preservation or rise in volume until approximately the end of the fifth decade, and then, subsequently, a precipitous decline in volume (Table 3; Figure 2). However a variety of patterns of age-associated change were apparent in the different regions.
Effects of AD that exceeded volume reductions linked to aging were found in several regions but were somewhat more selective than those found with the comparison of OLD to YNG. In general, effects of AD tended to be in regions where cortical degeneration has been reported including parahippocampal, entorhinal, precuneus and inferior parietal WM, as well as rostral middle frontal WM (Table 4; Figure 2). The majority of the WM volume differences were significantly different from nondemented control participants (OLD) when limiting the AD sample to individuals with CDR 0.5 only indicating that WM differences could be detected in the very earliest stages of the disease (Table 4).
We next examined the association between total WM signal abnormalities and regional WM volumes, controlling for age in the OLD and AD combined. There were modest associations between abnormalities and WM volume in the left pars opercularis (t = 2.21, p = 0.030), left parsorbitalis (t = −2.10, p = 0.038), left and right postcentral (t = 2.83, p = 0.006; t = 2.94, p = 0.004, respectively), left precentral (t = 2.84, p = 0.006), left and right superior parietal (t = 2.80, p = 0.006; t = 2.88, p = 0.005, respectively), left transverse temporal (t = 3.20, p = 0.002), right banks of the superior temporal sulcus (t = 2.08, p = 0.040), right superior temporal (t = 2.32, p = 0.023), and right supramarginal WM (t = 2.42, p = 0.018).
The primary age and AD effects on WM were apparent in both men and women when splitting the groups by sex, demonstrating a replication of the study results in most regions (Figure 3). Sex differences were apparent in a small number of regions and were generally of minor effect size as contrast to the major effects that generalized across male and female participants. Specifically, we examined whether there were sex differences in any region in YNG, OLD, and AD, with each group first examined individually, and secondarily, examined for group interactions. YNG men had greater WM volume in left superior temporal and right temporal pole and pars orbitalis, and less volume in left frontal pole, right caudal anterior cingulate, lingual, and transverse temporal WM compared to YNG women (all ps < 0.05). OLD men had greater WM volume than OLD women in left posterior cingulate, right entorhinal, and bilateral temporal pole, and less volume in left frontal pole, right superior parietal, and bilateral cuneus (all ps < 0.05). AD men had greater WM volume than AD women in left and less volume in left lateral orbitofrontal, supramarginal, and right superior temporal, right caudal middle frontal, right middle temporal, and right pars opercularis (all ps < 0.05). Only the left lateral orbitofrontal, left posterior cingulate, and bilateral cuneus showed a significant groups by sex interaction overall (all ps < 0.05).
Ventricular enlargement is an additional potential index of WM alterations. Figure 4 demonstrates ventricular enlargement in OLD compared to YNG (left and right ps < 0.001) and in AD compared to OLD (left and right ps = 0.07 and 0.003, respectively; Figure 4). Table 5 presents the association between regional WM volumes and ventricular volume in each hemisphere. There were minimal associations between ventricular and WM volume in YNG, several regional associations in OLD, and a selective set of associations in AD which were mostly found in a subset of the regions showing associations in the OLD.
The current study demonstrated patterns of WM atrophy in normal older adults and in patients with AD utilizing a novel, automated WM parcellation procedure. These data reveal that widespread differences in WM volume are a consequence of late, nondemented aging and that additional WM volume reduction beyond the effects of age is significant, yet somewhat more regionally selective, in AD. Profound age-associated WM differences were found in regions typically thought to be affected by age such as portions of the frontal lobe including medial orbitofrontal WM, a region that has been found to show accelerated diffusion changes in an independent sample in our prior work (Salat et al., 2005a; Salat et al., 2005b). The novel procedure revealed information that has not been reported in prior work. For example, although WM atrophy was apparent in regions often described to be affected with aging, significant reductions in volume were also found in regions that have not been widely previously reported, such as fusiform, inferior temporal, and middle temporal WM. Such findings may not have been possible when examining data on a lobar basis. Certain regions such as superior frontal gyrus showed WM atrophy with age as well as cortical thinning in our prior work (Salat et al., 2004). These findings may suggest a preferential vulnerability of the superior frontal gyrus to the effects of aging although the general pattern of age-associated WM atrophy is best described as being wide-spread. Some cortical areas demonstrated evidence of preserved or even increased volume in the younger adults into the sixth decade, with a precipitous decline thereafter. These results suggest that regional WM volume is dynamically changing throughout the adult age span and may exhibit complex patterns of growth and decline.
Strong effects were apparent in some regions that have not typically been noted in prior studies of aging, such as the fusiform WM. However, recent work by Thomas and colleagues found an age associated reduction in diffusion metrics in regions within the fusiform that was associated with a decline in facial perception in older adults (Thomas et al., 2008). These findings provide converging evidence of the effects measured with the WM parcellation technique and the potential importance of such procedures for broadly exploring the brain in a regionally unbiased manner.
Effects of AD were greatest in areas associated with cortical degeneration including entorhinal, parahippocampal, and posterior cingulated/precuneus WM, and show some overlap with our prior work examining cortical degeneration in AD (Dickerson et al., 2008). These effects remained when limiting the sample to individuals with a CDR 0.5 (Table 4), suggesting that WM changes occur relatively early with regard to clinical presentation. Similarly, patterns of WM degeneration with aging showed some overlap (e.g. superior frontal gyrus) and some distinct effects (e.g. fusiform) compared to our prior work on age-associated cortical loss (Salat et al., 2004). Patterns of age and AD associated WM atrophy differed, yet there was some convergence between these conditions in the vulnerability of superior frontal, middle temporal, and inferior temporal WM.
Overall, these data demonstrate the potential role of regional WM changes in age and AD associated neural and cognitive decline. One possibility is that WM degeneration contributes to the overall reduction in integrity of cognitive networks that are affected in the earliest stages of AD. These alterations could contribute to the classically described isolation of the hippocampal formation from the neocortex (Hyman et al., 1984), as also demonstrated in our recent diffusion work (Salat et al., 2008), and to the profound memory symptoms that are evident in the earliest stages of clinically recognizable AD. Recent functional imaging studies have revealed a more extensive network important to memory function that includes the hippocampal formation and specific, distributed cortical regions linked to the hippocampal formation (Buckner et al., 2008; Vincent et al., 2006; Wagner et al., 2005). WM volume in AD was reduced throughout many of these regions consistent with AD being linked to disruption of a broad network (Buckner et al., 2008; He et al., 2008; Salat et al., 2008). Given the finding that WM volume loss is apparent with very mild dementia, it is possible that this loss could have an important impact on clinical status, and could explain a portion of the variability in clinical presentation.
A number of findings uphold the utility of the presented WM parcellation method for labeling homologous regions across individuals. Test-retest reliability across two scan sessions in twenty individuals was generally good, with the majority of regions being within 5%. Variance in the volumetric measures was low across young and older adults, as well as across patients with AD. Data showed expected directional age and disease related changes. Certain findings of age and AD-related volume reduction measured using this technique are in accord with our prior studies of WM volume loss as well as DTI anisotropy studies of WM integrity (Head et al., 2004; Salat et al., 2005a; Salat et al., 2005b; Salat et al., 2008). There were associations between WMSA volume and regional WM volumes, which may suggest that volume loss is in some way associated with pathological mechanisms contributing to these lesions. However, these were modest effects, and the relative lack of associations overall may suggest that these volumetric measures provide indices of additional independent pathologic processes as has been demonstrated in prior work (Smith et al., 2000). In contrast, there were regional associations between WM volume and ventricular volume in the OLD and AD suggesting that these regional measurements of WM volume could provide further insight into the clinical implications of ventricular enlargement which has been traditionally fairly nonspecific with regard to disease pathology. Future studies will examine this in greater detail, as well as the relationship between WM atrophy and cortical thickness to better understand the relationship between these distinct markers of degeneration. Sex differences in volumetric measures were minimal in each of the three groups, and differences found were not consistent across groups or across hemispheres, suggesting that gender has a relatively minor influence on the described measures.
Certain prior imaging and post mortem studies have noted limited or total lack of change in WM volume in aging or AD (Double et al., 1996; Jernigan et al., 1991a; Jernigan et al., 1991b; Obara et al., 1994; Rusinek et al., 1991; Sullivan et al., 1995; Tanabe et al., 1997). We find preservation of WM volume in the early portion of the adult age span, and only selective regional loss of WM volume with AD. Additionally, the data suggest that younger AD may show greater regional WM volume differences compared to age matched controls compared to older AD (see Figure 2). Such an effect could make the detection of WM volume loss more difficult in smaller participant samples or with examination of a limited portion of the age span. These data demonstrate the importance of a full adult agespan study of regional effects due to the complex regional patterns measured.
It should be noted that the present WM parcellation scheme is an extension of cortical parcellation procedures. In this sense, any variation in the methods for cortical parcellation could affect the overall WM measures. For example, changes in surface area of a cortical region would affect the amount of WM labeled. It is also possible that changes in tissue composition could affect surface placement and volume measurements. A portion of the temporal lobe regions showed reduced cross-session reliability, particularly in the smaller regions. We note that that the effects were generally reliable across regions. That is, most of the effects reported were replicated in the two samples examined (men and women) across groups. Additionally, combining across regions, for example, combining the entorhinal and parahippocampal regions, improves reliability without changing the interpretation of the results. It is therefore likely that the results presented are not substantially altered by the cross-session reliability. However, future work examining the longitudinal nature of WM volume loss will require further development to enhance this reliability. The procedures for surface generation utilize local intensity gradient information (Dale et al., 1999; Fischl and Dale, 2000; Fischl et al., 2001; Fischl et al., 1999a), and thus, should not be greatly affected by absolute intensity changes as long as contrast between the two tissue classes (GM and WM) still remains. The participant sample overlaps partially with our prior work, and the finding of decreases in WM volume in certain regions that also showed cortical thinning, such as the superior frontal gyrus, suggest that the WM volume reduction measured is not a result of surface misplacement in the cortical reconstruction. We also note that the specific technical procedures used here may influence the measurements. For example, the cortical parcellation regions designate the labeling of the WM, and it is possible that different anatomical schemes would produce varying results. The Freesurfer tools for the parcellation procedure are freely available for download (ude.dravrah.hgm.rmn.refrus), and therefore, investigators could examine the reproducibility of the results in independent datasets such as the Alzheimer’s Disease Neuroimaging Initiative (ADNI; www.adni-info.org/) and also against alternative parcellation schemes. The data examined here are also freely available through the OASIS data release (Marcus et al., 2007; http://www.oasis-brains.org/) allowing future investigators to contrast and improve upon all aspects of the present study. The present study demonstrates differential patterns of WM degeneration in aging and dementia and it is possible that WM degeneration may contribute to cognitive decline and dementia severity in AD.
This work was supported by NCRR P41RR14075, the MIND Institute, the Athinoula A. Martinos Center for Biomedical Imaging, NINR NR010827, NIA AG024898, AG05886, AG05681, HHMI, HBP NS39581 and RR14075. We thank the Washington University ADRC for making MRI data available.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.