Pattern of Altered Magnetization Transfer Rate in Alzheimer’s Disease

Abstract

Background:

Biomarkers for Alzheimer’s disease (AD) are crucial for early diagnosis and treatment monitoring once disease modifying therapies become available.

Objective:

This study aims to quantify the forward magnetization transfer rate (k_for) map from brain tissue water to macromolecular protons and use it to identify the brain regions with abnormal k_for in AD and AD progression.

Methods:

From the Cardiovascular Health Study (CHS) cognition study, magnetization transfer imaging (MTI) was acquired at baseline from 63 participants, including 20 normal controls (NC), 18 with mild cognitive impairment (MCI), and 25 AD subjects. Of those, 53 participants completed a follow-up MRI scan and were divided into four groups: 15 stable NC, 12 NC-to-MCI, 12 stable MCI, and 14 MCI/AD-to-AD subjects. k_for maps were compared across NC, MCI, and AD groups at baseline for the cross-sectional study and across four longitudinal groups for the longitudinal study.

Results:

We found a lower k_for in the frontal gray matter (GM), parietal GM, frontal corona radiata (CR) white matter (WM) tracts, frontal and parietal superior longitudinal fasciculus (SLF) WM tracts in AD relative to both NC and MCI. Further, we observed progressive decreases of k_for in the frontal GM, parietal GM, frontal and parietal CR WM tracts, and parietal SLF WM tracts in stable MCI. In the parietal GM, parietal CR WM tracts, and parietal SLF WM tracts, we found trend differences between MCI/AD-to-AD and stable NC.

Conclusion:

Forward magnetization transfer rate is a promising biomarker for AD diagnosis and progression.

Keywords

Alzheimer’s disease magnetization transfer imaging magnetization transfer rate mild cognitive impairment

INTRODUCTION

Alzheimer’s disease (AD) is the most common cause of dementia in older adults and mild cognitive impairment (MCI) is often considered as a transitional clinical phase between normal brain aging and AD [1]. Brain magnetic resonance imaging (MRI) studies have shown that AD involves abnormalities in both gray matter (GM) and white matter (WM). Structural MRI demonstrated the discriminative and predictive power of atrophy in the hippocampus and entorhinal cortex and ventricular enlargement [2 –4]. Perfusion MRI detected compromised cerebral blood flow in the posterior cingulate, precuneus, parietal, hippocampus, temporal, and frontal regions [5, 6]. Diffusion tensor imaging (DTI) assessed brain microstructural changes and reported integrity disruption of WM pathways in corona radiata, superior and inferior longitudinal fasciculus, corpus callosum for individuals with preclinical AD, such as those with subjective cognitive decline [7], positive amyloid-β protein levels [8], and clinically diagnosed AD [9 –11].

Magnetization transfer imaging (MTI) evaluates brain microstructure in both GM and WM and has been used in the detection of AD. MTI includes the techniques of chemical exchange saturation transfer (CEST), T₁ρ, and magnetization transfer (MT) weighted MRI [12, 13]. MT results from dipolar coupling and ¹H exchange between tissue water and macromolecular proton compartments [14]. Macromolecules like proteins cannot be imaged directly with conventional MRI due to their very short T₂ values [15]. However, macromolecules can be interrogated through their interaction with tissue water. For example, the off-resonance RF excitation of the macromolecular protons changes the relaxation times and magnetization in the tissue water proton pool. The magnetization transfer ratio (MTR) can be derived from water proton images with and without off-resonance RF saturation and it was associated with axonal attenuation and myelin content [16 –18]. MTR is reduced in AD in the whole brain [19 –21], GM [20, 22], WM [22], hippocampus [19 , 24], and medial temporal lobes [20]. However, in order to increase the reliability of the measurements, MTR was exclusively assessed in specific regions of interest. This means that the investigators may have excluded regions for analysis in order to optimize power in the other regions. In addition, MTR is affected by the RF excitation amplitude, frequency offset, duty cycle, the acquisition parameters, and tissue properties. Therefore, it is preferable to measure voxel-level MT-related intrinsic tissue properties, e.g., the forward magnetization transfer rate (k_for) map from tissue water to macromolecular protons, that are independent of the details of the data acquisition [25].

In this study, we developed a method to quantify the k_for map on a voxel-by-voxel basis and assessed the sensitivity of the k_for map in AD diagnosis and progression without a priori hypotheses regarding which regions would be affected in both the cross-sectional and longitudinal analyses. The MTI data were acquired to measure tissue T₁ when the arterial spin labeling pulse sequence was applied on upstream blood vessels [26 –28]. Preliminary results of MT effects of the arterial spin labeling sequence were reported [29]. The resulting k_for was used to investigate the extent of microstructural damage of brain GM and WM regions from the entire AD progression spectrum: normal controls (NC), MCI, and AD.

METHODS

Theory

The T₁ value observed in the brain after macromolecular saturation (T_1sat) can be modeled as [30 –32]:

$\frac{1}{T_{1 sat}} = \frac{1}{T_{1 t}} + \frac{f}{λ} + k_{for}$ (1)

where T_1t is the T₁ of brain tissue water protons, f is the perfusion in ml/(100 g·s), λ is the brain-blood partition coefficient (typically 0.9 ml blood/g brain tissue) [33], k_for is the forward magnetization transfer rate for tissue to macromolecular water protons, and is a measure of tissue health. It is worth noting that Equation (1) is true only when the macromolecular pool is fully saturated.

The T₁ value without macromolecular saturation (T_1nosat) is given by [31, 32]:

$\frac{1}{T_{1 nosat}} = \frac{1}{T_{1 t}} + \frac{f}{λ}$ (2)

k_for can be derived from the subtraction of Equations (2) from (1) as:

$k_{for} = \frac{1}{T_{1 sat}} - \frac{1}{T_{1 nosat}}$ (3)

Participants

From 2002–2013, the Cardiovascular Health Study Cognition Study (CHS-CS) [34] measured cerebral blood flow, brain morphology and tissue health in 195 elderly volunteers at baseline. During the study, the volunteers progressed from NC to MCI and AD. Volunteers received follow-up MRIs during the course of the study. The subjects were classified as NC, MCI, or AD based on cognitive status adjudication [34] without verification for deposition of amyloid-β. The CHS-CS diagnostic criteria for MCI in the study included both MCI-amnestic type and MCI-multiple cognitive domain type [34]. Sixty-three age matched elderly volunteers from the CHS-CS study, including 20 NC (mean age±SD, 84.40±4.38 years), 18 MCI (mean age±SD, 84.77±3.01 years), and 25 early AD (mean age±SD, 84.40±3.30 years) were scanned using MRI to measure MT effects at baseline. Of those, 53 participants completed a follow-up MRI scan to measure the same MT effects as the baseline. The longitudinal follow-up scans can be divided into four groups: 15 stable NC (follow-up time±SD: 2.75±1.90 years), 12 NC-to-MCI (follow-up time±SD: 3.08±2.20 years), 12 stable MCI (follow-up time±SD: 2.69±1.81 years), and 14 MCI/AD-to-AD (follow-up time±SD: 1.91±1.37 years) subjects. Modified Mini-Mental State Examination (3MSE) scores [35] were used to assess general cognitive ability for the population. It has been shown that 3MSE scores have increased sensitivity in detecting dementia in comparison to the Mini-Mental State Examination [36, 37].

MRI protocol

All MRIs were performed on a dedicated GE Signa 1.5 T MRI (Version LX) at the University of Pittsburgh MR Research Center after informed consent. An inversion recovery sequence was performed with or without off-resonance macromolecular saturation. The images were acquired using multi-slice echo planar imaging (EPI) (field of view = 24×24 cm, matrix = 64×64, 19 slices, Thickness/Spacing: 5/0 mm, TE/TR: 21/9000 ms, 90°, 977 Hz/pixel) with twelve inversion times (TIs) for the first slice from: 0.025, 0.11, 0.195, 0.28, 0.365, 0.535, 0.705, 0.875, 1.3, 1.725, 2.15, and 2.575 s (sample sets of 12 TI images without and with off-resonance saturation are shown in Fig. 1A and B). Images were acquired using a quadrature transceiver RF coil. The 19 slices were acquired sequentially from superior to inferior during a 1 s multi-slice acquisition. Therefore, TIs varied slice by slice. The off-resonance macromolecular saturation was administered between the inversion pulse and the image acquisition, using a pulse train with an input amplitude of 3.5μT at 92% duty cycle and 85 ms repetition time and a frequency offset of 10.6 kHz below the water frequency. This frequency offset, originally designed to measure the MT effects of the CASL sequence, is therefore larger than the 1–3 kHz frequency offset in typical MTI [38 –41]. T₁ maps were quantified from the images acquired with and without macromolecular saturation, referred to as T_1sat and T_1nosat maps respectively.

Fig. 1

Sample images from a middle axial slice of a normal control subject. (A) 12 TI images of T_1nosat, (B) 12 TI images of T_1sat, (C) T_1nosat image derived from (A), (D) T_1sat image derived from (B), (E) T_1nosat – T_1sat image, and (F) k_for image. The 12 TIs for the slice were 0.499, 0.584, 0.669, 0.754, 0.839, 1.009, 1.179, 1.349, 1.774, 2.199, 2.624, and 3.049s, respectively.

Coronal T₁-weighted three-dimensional anatomical images were acquired with spoiled gradient-echo (SPGR) images (field of view = 24×19 cm, matrix = 256×192, TE/TR = 5/25 ms, flip angle = 40°, slice thickness/gap = 1.5/0 mm, number of slices = 124, bandwidth = 63 Hz/pixel).

Image analysis

Calculation of T_1sat and T_1nosat

To correct for the subject motion, SPM12 “realign” was used to align all 12 images acquired with and without off-resonance saturation. In principle, T_1sat at each single voxel can be calculated by fitting the inversion recovery signals as a function of the TIs. $s (T I) = P (1 - 2 e^{- T I / T_{1 sat}})$ (4) where TI is the inversion time, s(TI) is the signal at inversion time TI, P is the proton density, and T_1sat is the T₁ value measured after the application of off-resonance RF pulse. Equation (4) represents a two-parameter fit that may assume perfect inversion that is unrealistic when the B₁ field is inhomogeneous. However, combining the Equation (4) with the signal-weighted inversion time (see Equation (5)) one can calculate T_1sat numerically, resulting in better results than a nominal three-parameter fit.

The signal-weighted inversion time is calculated as below. $WI (T_{1 sat}) = \frac{\sum_{i = 1}^{12} T I_{i} s (T I_{i})}{\sum_{i = 1}^{12} s (T I_{i})}$ (5)

This simple first moment noniterative algorithm proved to be very robust. The signal-weighted inversion time, WI, was numerically calculated using Equation (4) for a range of T_1sat. WI was found to be a function consisting of two monotonic functions of T_1sat with a singular point T₁max, as shown in Fig. 2A. These two monotonic curves can be reorganized to form a single monotonic curve (Fig. 2B), which could be numerically inverted to calculate T_1sat from WI. T_1nosat was calculated using the same method as T_1sat but with the signal images acquired without macromolecular saturation. Sample images of T_1nosat and T_1sat are shown inFig. 1C and D.

Fig. 2

Numerical calculation of T₁ (T_1sat and T_1nosat) using signal weighted inversion (WI). (A) Theoretical graph of WI as a function of T₁ (Equation (5), with T_1sat replaced by T₁ for generality), WI reaches its maximum value at T_1max, (B) WI forms a monotonically increasing curve of T₁ if the curve at [0 T_1max] ms is shifted to the right of [T_1max 4000] ms. WI is an inversible function except for the values within [WI(4000) WI(0)]. For the value of WI that falls into the range, T1 is assigned to 0 ms or 4000ms depending on whether WI is closer to WI(0) or WI(4000).

Calculation of k_for

The realigned 12 image sets with and without macromolecular saturation were averaged to calculate mean images, respectively. For each subject, the mean realigned image with macromolecular saturation was co-registered to the mean realigned image without saturation and the co-registration parameters were used to transform the T_1sat map to align with the T_1nosat map. k_for was calculated from the aligned T_1sat and T_1nosat maps using Equation (3). To eliminate extreme values in these k_for maps, the signal less than –0.1 and greater than 0.6 was assigned to –0.1 and 0.6, respectively. Although k_for should not be negative in general, those negative values were left in k_for maps to avoid violating noise distribution. Sample images of T_1nosat - T_1sat and k_for are shown in Fig. 1E and F. The large frequency offset in the pulse sequence cannot fully saturate the macromolecular pool, and therefore the measured k_for will be underestimated.

Normalization of T_1sat, T_1nosat, and k_for maps

Anatomical SPGR images were segmented into GM, WM, and cerebrospinal fluid (CSF) probability maps using SPM12’s “segment” tool. The GM probability map was co-registered to the mean realigned image without saturation so the transformed GM probability map were aligned with T_1sat, T_1nosat, and k_for maps. The transformed GM probability map was registered to the GM template in the standard MNI brain, and the warping parameters were used to transform the T_1sat, T_1nosat, and k_for maps to the standard MNI brain space.

Statistical analysis

Demographic and neuropsychological data were analyzed with MATLAB, version R2019a. All tests were 2-tailed and the significance was set for p < 0.05. Normality was assessed with the Shapiro-Wilk test. Demographic and neuropsychological characteristics for the NCs, MCIs, and ADs at baseline and for stable-NCs, NC-to-MCIs, stable-MCIs, and MCI/AD-to-ADs at follow-up were compared using either two-sample t-tests (for normal distribution) or Mann-Whitney nonparametric U-tests (for non-normal distribution) for continuous variables and χ² test for categorical variables.

Generation of global regions of interest (ROIs) (whole brain, GM, and WM)

The global ROIs (whole brain, GM, and WM) were obtained by thresholding the corresponding SPM a priori probability map in the standard space with probability greater than 0.3, 0.8, and 0.95, respectively. The selected threshold levels allow us to minimize partial volume effects and have a sufficient number of voxels for reliable averaging. Regional k_for values were calculated as the median over each brain global ROI.

Voxel-level analysis on k_for maps at baseline

k_for maps were compared voxel-by-voxel among the three baseline groups using SPM’s general linear model (GLM), controlling for age and gender. Because of partial coverage, only the voxels in which k_for maps had valid values (intersection area from both T_1sat and T_1nosat maps) from all the participants were analyzed. The analyzed area in the SPM analysis is shown in Supplementary Figure 1A. A voxel-level p-value of 0.01 was used to threshold the statistical maps. A cluster-level p-value of 0.05 was used to guard against false positives from multiple comparisons. The MCI and AD-affected regions were used for further region-based analyses.

Voxel-level analyses on longitudinal changes of k_for

The longitudinal changes of k_for were compared among the stable-NCs, NC-to-MCIs, stable-MCIs, and MCI/AD-to-ADs using a multiple linear regression. The longitudinal change of k_for maps for each subject was calculated between the baseline and follow-up scans and compared voxel-by-voxel among the four longitudinal groups using SPM’s GLM, controlling for age, gender, and years between baseline and follow-up scans. The analyzed area in the SPM analysis is shown in Supplementary Figure 1B. A voxel-level p-value of 0.01 was used to threshold the statistical maps. A cluster-level p-value of 0.05 was used to guard against false positives from multiple comparisons. A more liberal voxel-level threshold of p < 0.025 was used to view the trends of longitudinal changes, although a corrected cluster-level threshold of p < 0.05 was still used for comparisons.

Region-level analyses on longitudinal changes of k_for

Regional k_for comparisons were performed among the four longitudinal groups for the whole brain, GM, WM, and MCI/AD-affected regions detected in the voxel-level cross-sectional analysis, controlling for age, gender and follow-up time between baseline and follow-up scans. Post hoc regional analyses were also performed to determine whether the regions detected in the voxel-level longitudinal analysis had better sensitivity in group differences.

RESULTS

Demographic and cognitive information

Table 1 summarizes the subjects’ demographic and cognitive data for the cross-sectional and longitudinal analyses. For the cross-sectional analysis, no significant difference was found for age and gender across the NC, MCI, and AD groups. The AD group showed significantly lower 3MSE scores than the NC and MCI groups. For the longitudinal analysis, no significant difference was found for age, gender, follow-up time, and changes in 3MSE scores across the stable-NC, NC-to-MCI, stable-MCI, and MCI/AD-to-AD groups. The MCI/AD-to-AD group showed significantly lower 3MSE scores at baseline compared to the stable-NC, NC-to-MCI, and stable-MCI groups.

Table 1

Demographic and cognitive scores at baseline and follow-up. The differences between two time points were calculated by subtracting the baseline values from follow-up values

Baseline (n = 63)
	NC (n = 20)	MCI (n = 18)	AD (n = 25)	p
Age (y)	84.40±4.38	84.40±3.30	84.77±3.01	0.1986
Gender (F, %)	10 (50%)	12 (67%)	20 (80%)	0.1054
3MSE scores^†	94.72±3.87	91.89±8.54	82.70±11.92	0.0001
	(n’ = 17)	(n’ = 15)	(n’ = 23)
Follow-up (n = 53)
	Stable NC (n = 15)	NC to MCI (n = 12)	Stable MCI (n = 12)	MCI/AD to AD (n = 14)	p
Baseline age (y)	83.13±3.94	83.45±3.06	85.99±4.05	82.91±3.30	0.7435
Gender (F, %)	9 (60%)	6 (50%)	8 (67%)	9 (64%)	0.8425
Follow-up Time (y)	2.75±1.90	3.08±2.20	2.69±1.81	1.91±1.37	0.4494
Baseline 3MSE scores	95.26±3.67	95.41±4.39	95.62±2.46	86.73±6.96	0.0055
3MSE change ^†	–0.27±3.77	–0.94±2.26	–0.05±3.17	–3.14±4.65	0.2125
	(n’ = 13)	(n’ = 11)	(n’ = 9)	(n’ = 12)

^†3MSE scores were not completed for all the subjects at the baseline and follow-up studies. The actual number of subjects with the 3MSE scores measured is indicated inside the brackets.

Cross-sectional comparisons of T_1nosat, T_1sat, and k_for in the global ROIs

T_1nosat values were not significantly different in the whole brain (561±119 ms), GM (957±246 ms), and WM (688±133 ms) among the NC, MCI, and AD groups. T_1sat values were significantly higher in WM for ADs versus NCs (p = 0.0061; 672±158 ms versus 556±92 ms) and for ADs versus MCIs (p = 0.022; 672±158 ms versus 560±142 ms), but not different for whole brain (529±97 ms) and GM (978±247 ms) among the NC, MCI, and AD groups. k_for values were significantly lower in WM for ADs versus NCs (p = 0.0049; 0.17±0.16/s versus 0.35±0.23/s) and for ADs versus MCIs (p = 0.00087; 0.17±0.16/s versus 0.36±0.17/s) and in the whole brain for ADs versus MCIs (p = 0.0041; 0.16±0.11/s versus 0.27±0.11/s), but not different for GM (0.065±0.036/s) among the NC, MCI, and AD groups.

Cross-sectional voxel-wise group comparisons

Example k_for maps from each group of NC, MCI, and AD subjects are shown in Fig. 3. Compared to the NC group, the AD group had significantly decreased k_for in three clusters: the left corona radiata WM region (cluster 1, in green circle, p = 0.044), right corona radiata extending to superior longitudinal fasciculus WM regions and right frontal lobe extending to precentral GM regions (cluster 2, in purple circle, p = 0.001) (Fig. 4A), and the right superior longitudinal fasciculus WM region and right postcentral, paracentral lobule and precuneus GM regions (cluster 3, p = 0.031) (Fig. 4B).

Fig. 3

Representative k_for images from (A) NCs, (B) MCIs, and (C) ADs.

Fig. 4

Compared to the NC group, the AD group had significantly lower k_for values in three clusters: (A) the left corona radiata white matter region (cluster 1 in green circle), right corona radiata extending to superior longitudinal fasciculus white matter regions and right frontal lobe extending to precentral gray matter regions (cluster 2 in purple circle), (B) the right superior longitudinal fasciculus white matter region and at right postcentral, paracentral lobule and precuneus gray matter regions. Compared to the MCI group, the AD group had significantly lower k_for values in (C) two clusters: the left corona radiata and internal capsule white matter regions and left anterior cingulum and superior medial frontal gray matter regions (cluster 4 in green circle), and the right corona radiata and external capsule white matter regions and right superior frontal gray matter regions (cluster 5 in purple circle). The color bar stands for the range of t values.

Compared to the MCI group, the AD group had significantly decreased k_for in two clusters: the left corona radiata, internal capsule, external capsule and corpus callosum WM region and left cingulum, putamen, precentral, insula, and superior medial frontal GM regions (cluster 4, in green circle, p < 0.001), and the right corona radiata, external capsule, superior longitudinal fasciculus and internal capsule WM regions and putamen, insula, middle frontal, and right superior frontal GM regions (cluster 5, in purple circle, p < 0.001) (Fig. 4C). Statistics of the clusters are listed in Table 2.

Table 2

Summary of cluster-level statistics for clusters with significant k_for differences among the NC, MCI, and AD groups

Cluster	No. voxels	Peak-t	Peak-t MNI coordinates	GM/WM	Anatomical locations	% Cluster	% Region
1 NC > AD	407	3.78	–28.8 14.0 26.0	WM	Corona radiata [L]	60.69%	13.81%
2 NC > AD	983	3.77	34.0 –8.0 50.0	WM	Corona radiata [R]	28.89%	12.75%
					Superior longitudinal fasciculus [R]	4.07%	4.85%
				GM	Middle frontal [R]	14.04%	2.70%
					Precentral [R]	11.60%	3.37%
					Operculum inferior frontal [R]	3.56%	2.50%
					Inferior triangularis frontal [R]	2.54%	1.16%
3 NC > AD	485	3.77	6.0 –36.0 58.0	WM	Superior longitudinal fasciculus [R]	4.95%	2.91%
				GM	Postcentral [R]	46.19%	5.86%
					Paracentral Lobule[R]	24.74%	14.35%
					Precuneus [R]	7.63%	1.13%
4 MCI > AD	2401	4.97	–14.0 40.0 2.0	WM	Corona radiata [L]	31.57%	42.37%
					Internal capsule [L]	6.16%	12.54%
					External capsule [L]	4.00%	13.87%
					Corpus callosum	3.62%	3.04%
				GM	Anterior cingulum [L]	8.25%	5.93%
					Superior medial frontal [L]	5.58%	4.48%
					Putamen [L]	5.16%	12.29%
					Precentral [L]	2.79%	1.90%
					Insula [L]	2.67%	3.44%
					Superior frontal [L]	2.58%	1.72%
					Middle frontal [L]	2.04%	1.01%
					Supplementary motor area [L]	1.58%	1.77%
5 MCI > AD	2502	4.51	30.0 22.0 10.0	WM	Corona radiata [R]	29.78%	41.95%
					External capsule [R]	7.43%	25.51%
					Superior longitudinal fasciculus [R]	5.64%	17.09%
					Internal capsule [R]	1.84%	3.76%
				GM	Superior frontal [R]	8.83%	5.45%
					Putamen [R]	5.48%	12.88%
					Middle frontal [R]	5.12%	2.51%
					Insula [R]	4.96%	7.01%
					Inferior triangularis frontal [R]	1.44%	1.67%

The Johns Hopkins University (JHU) WM atlas and automated anatomical labeling (AAL) GM atlas were used to label anatomical locations. % Cluster indicates the percentage of each cluster that falls within the defined region, % Region indicates the percentage of each defined region that falls within the cluster. The listed anatomical regions are either “% Cluster”>1% or “% Region”>1%.

Longitudinal voxel-wise group comparisons

Compared to the stable-NC group, the stable-MCI group exhibited larger k_for decreases in the left corona radiata extending to the left superior longitudinal fasciculus WM regions (cluster 6, in green circle, p = 0.003) and the right corona radiata and right superior frontal GM regions (cluster 7, in purple circle, p = 0.017) (Fig. 5A, B). No significant changes were observed for the other longitudinal groups for voxel-level p-value of 0.01. With less stringent voxel-level p-value of 0.025, compared to the stable-NC group, a symmetrical k_for decrease pattern of left and right sides was observed in the stable-MCI group (Supplementary Figure 2); the MCI/AD-to-AD group exhibited a larger trend of k_for decreases in the corpus callosum, right superior longitudinal fasciculus, right corona radiata, and cingulum WM regions and the left precuneus, posterior cingulate, and left inferior parietal GM regions (cluster 8, p = 0.012) (Fig. 5C). Statistics of the clusters are listed in Table 3.

Fig. 5

Compared to the stable-NC group, the stable-MCI group exhibited larger k_for decreases in (A) the left corona radiata extending to left superior longitudinal fasciculus white matter regions (cluster 6, in green circle) and (B) the right corona radiata and right superior frontal gray matter regions (cluster 7, in purple circle); the MCI/AD-to-AD group exhibited larger k_for decreases in (C) the corpus callosum, right superior longitudinal fasciculus, right corona radiate and cingulum white matter regions and at left precuneus, both sides of the posterior cingulum and left inferior parietal gray matter regions (cluster 8). The color bars stand for the range of t values.

Table 3

Summary of cluster-level statistics for clusters with significantly larger k_for decreases among the stable-NC, NC-to-MCI, stable-MCI and MCI/AD-to-AD groups. The k_for decreases were calculated by subtracting follow-up k_for values from baseline values

Cluster	No. voxels	Peak-t	Peak-t MNI coordinates	GM/WM	Anatomical locations	% Cluster	% Region
6 stable-MCI > stable-NC	652	3.47	–38.0 –32.0 42.0	WM	Corona radiata [L]	12.58%	5.99%
					Superior longitudinal fasciculus [L]	11.04%	8.83%
				GM	Inferior parietal [L]	16.41%	4.37%
					Postcentral [L]	15.80%	2.65%
7 stable-MCI > stable-NC	435	3.32	26.0 20.0 28.0	WM	Corona radiata [R]	33.79%	8.28%
				GM	Superior frontal [R]	11.49%	1.23%
8 MCI/AD-to-AD > stable-NC	1010	3.73	–20.0 –48.0 24.0	WM	Corpus callosum	24.06%	7.43%
					Superior longitudinal fasciculus [R]	6.34%	7.76%
					Corona radiata [R]	6.24%	7.02%
					Cingulum	4.26%	6.81%
				GM	Precuneus [L]	11.19%	3.20%
					Posterior cingulum [L]	10.79%	4.53%
					Inferior parietal [L]	5.94%	2.45%
					Posterior cingulum [R]	1.58%	4.78%

Longitudinal regional group comparisons

Longitudinal k_for decreases were observed significantly different from zero (p < 0.001) on the global ROIs and the clusters 1–5 derived from the cross-sectional study. However, compared to the changes in the stable NC group, longitudinal k_for changes in the other groups were found only marginally different on the global ROIs and the cross-sectional clusters. Specifically, compared to the stable NC group, the stable MCI group exhibited marginally larger longitudinal k_for decreases in the whole brain (p = 0.068), cluster 2 (p = 0.080), and cluster 5 (p = 0.072). The MCI/AD-to-AD group had marginally larger longitudinal k_for decreases in the WM (p = 0.062), cluster 1 (p = 0.058), cluster 2 (p = 0.083), and cluster 5 (p = 0.085).

The regional k_for values in the clusters derived from the longitudinal voxel-wise analyses corroborated the voxel-level results and exhibited more significant changes. We used larger and symmetrical clusters 6–7 (voxel-level p value of 0.025) for the regional analyses. In cluster 6, significantly larger longitudinal k_for decreases were observed in the NC-to-MCI, stable MCI, and MCI/AD-to-AD groups (Fig. 6, p = 0.05, p = 0.0006, and p = 0.0115 respectively) when compared to the stable NC group. In clusters 7 and 8, compared to the stable NC group, significantly larger longitudinal k_for decreases were observed only in the stable MCI group (Fig. 6, p = 0.0010 and p = 0.0078 for the clusters 7 and 8, respectively) and the MCI/AD-to-AD groups (Fig. 6, p = 0.0129 and p = 0.0008 for the clusters 7 and 8, respectively), while marginally larger longitudinal k_for decreases were observed in the NC-to-MCI group (Fig. 6, p = 0.0796 and p = 0.0719 for the clusters 7 and 8, respectively). Years between baseline and follow-up scans were significantly associated with the longitudinal changes in k_for (p < 0.05).

Fig. 6

Z scores of regional k_for changes in the NC-to-MCI, Stable-MCI, and MCI/AD-to-AD groups compared to the stable-NC group. ∧, ^*, and ^** represent the significance group differences with 0.05≤p < 0.1, 0.01≤p < 0.05, and p < 0.01, respectively. Clusters 6–8 were identified from the voxel-based longitudinal group comparisons of k_for changes with voxel-level p < 0.025.

DISCUSSION

We found lower k_for in the frontal GM, parietal GM, frontal corona radiata WM tracts, frontal and parietal superior longitudinal fasciculus WM tracts in ADs relative to both NCs and MCIs. From the longitudinal study, we also observed progressive decreases in k_for in the frontal GM, parietal GM, frontal and parietal corona radiata WM tracts, and parietal superior longitudinal fasciculus WM tracts in stable MCIs; and in the parietal GM, parietal corona radiata WM tracts, and parietal superior longitudinal fasciculus WM tracts in MCI/AD-to-ADs.

Lower k_for values were previously reported in the hippocampus, temporal lobe, posterior cingulate cortex, and parietal GM regions [42] and preselected WM fiber tracts [43] of AD patients. Lower k_for values in the WM tract were reported to differentiate MCI converters from stable MCIs [43]. Other quantitative MT parameters, such as the T₂ of the restricted pool, fractional pool size, and the ratio of relaxation times of the free proton pool to restricted pool, were associated with AD [44, 45]. However, these quantitative MT parameters were mostly analyzed on predetermined regions of interest for increased signal-to-noise ratio. Our study explored voxel-wise k_for abnormalities in AD and AD progression and allowed accurate detection for abnormal k_for spatial locations related to AD without a priorihypotheses.

The detected abnormal k_for GM regions included the superior frontal GM, anterior cingulate, inferior parietal, posterior cingulate, and precuneus regions. These regions are consistent with regions with reduced cerebral blood flow [26, 27] and reduced glucose metabolism [46, 47] that are typically reported in AD. Due to its overlap with regions with blood flow and metabolic deficits, this finding supports that k_for is related to metabolic processes [48] and mitochondrial dysfunction may play a role in the pathophysiology of AD for GM damage [49]. N-acetylaspartate (NAA), a neuron-specific metabolite synthesized by the mitochondria, has been reported with significant reduction in the whole brain tissue of MCI patients [50]. The reduction of NAA in the MCI further supports energy impairment and mitochondrial dysfunction may be already evident before individuals are clinically demented [50, 51]. Energy impairment from mitochondrial dysfunction causes neuronal and axonal damage, and therefore reduces magnetization transfer effects. The abnormal k_for WM regions included the frontal and parietal corona radiata and superior longitudinal fasciculus. Reduced k_for values in WM tracts of AD patients have been correlated to smaller fractional anisotropy values using DTI [43], which can be influenced by demyelination. Mitochondrial dysfunction was found to contribute to WM demyelination in multiple sclerosis [52, 53]. Therefore, mitochondrial dysfunction may play a similar role in the pathophysiology of AD for WM damage. Although highly speculative, our study supports a uniform role of mitochondrial dysfunction in the pathophysiologyof AD.

We also observed k_for decreases in the whole brain and global WM in ADs relative to MCIs or NCs and longitudinal decreases of k_for in the whole brain and global WM in ADs and MCIs. Existing studies using MTR mostly focused on histogram analyses in large regions [19 , 55], such as the whole brain, GM, WM, and some hypothesized regions, such as hippocampal regions and deep GM regions. Our study supports the whole brain and global WM microstructure deficits in AD. An MTR study analyzed the frontal lobe and found reduced histogram properties in the region [21]. Therefore, our findings extend current microstructural deficits of AD to the frontal WM and parietal WM, and also specifies the deficits in twoWM tracts.

Only one longitudinal MTR study was found, and it reported that AD patients had significant longitudinal decreases of MTR in the whole brain and all deep GM regions that were investigated over 12 months [55]. We extended these findings by demonstrating that the longitudinal decreases in k_for are significantly larger in stable MCIs and MCI/AD-to-ADs than in stable NCs, indicating that AD pathology accelerates k_for declines in frontal and parietal GM regions and WM tracts.

The frontal and parietal WM deficits in k_for are also in line with the locations for WM microvascular damage observed with mean diffusivity and fractional anisotropy using DTI [56 –58]. Altered mean diffusivity and free water signal fraction in the frontal WM and temporal WM regions have been significantly correlated with CSF biomarkers in AD [58]. We did not observe deficits in the k_for of the temporal region. However, the temporal region was located in the inferior portion of the imaging volume and only the overlapped regions covered across the subjects were analyzed. It is worth noting that anterior (frontal) corona radiata and posterior (parietal) corona radiata were listed as clusters with significant association with k_for declines. A DTI meta-analysis showed significant decreases of both WM volume and fractional anisotropy in corona radiata in ADs [10]. The alterations from both corona radiata and superior longitudinal fasciculus WM tracts were also supported by longitudinal DTI analysis in AD[11, 59].

Years between baseline and follow-up scans were significantly associated with the longitudinal changes in k_for. Our results support the involvement of frontal and parietal GM regions and WM tracts (corona radiata and superior longitudinal fasciculus) in the progression of AD, highlighting MTRs of those regions as promising biomarkers and disconnection of fibers as a potential mechanism.

This study has limitations, the most important of which is a consequence of the technical aspects of scanning and measuring magnetization transfer. Specifically, the data becomes less reliable in the more ventral aspects of the brain. in addition, we set the stringent criterion that each voxel must have representation from every subject in order to be included in the analyses. Therefore, the more ventral regions of the brain, most importantly including the inferior and middle temporal gyri, and the mesial temporal lobe including the hippocampus, were not included in the analyses. Consequently, there may be important, subtle changes in MTR in these regions, especially in the transition from normal cognition to MCI. For future studies we will need more focused protocols in order to examine these brain regions, but critically, we have derived a measure of magnetization transfer rate from a data set that was not designed for thispurpose.

Second, the images acquired with and without macromolecular saturation were acquired at 1.5 T before the widespread availability of 3 T. We expect that MTI studies at 3 T can benefit from longer T1 relaxation times in addition to higher steady state magnetization. However, the method of continuous off-resonance saturation used in this study would probably not be practical at 3 T due its high specific absorption rate (SAR). Nevertheless, we were able to observe the k_for differences between NCs, MCIs, and ADs at 1.5 T.

Third, we had small sample sizes, especially for the longitudinal analysis. This limits the statistical power to detect longitudinal changes for the groups with large variations. The fourth longitudinal group was a combined group from the subjects who turned into AD from MCI stage and those who stayed in the AD stage, resulting from the small numbers of participants in these two categories. If repeated in a larger population, there is a potential for a better understanding of the full trajectory of disease progression using k_for.

Fourth, no differences were found among the AD progression groups (NC-to-MCI, stable-MCI, and MCI/AD-to-AD groups) thus changes in k_for during the middle and late stages of AD remains incomplete. The lack of sensitivity may be partially attributed to the reduced sensitivity from the large RF frequency offset used in our protocol. Future studies should evaluate the k_for maps using the optimal sequence design (lower RF frequency offset and short RF duration). In addition, the participants were classified into NC, MCI, or AD based on neuropsychological assessments. The lack of neurobiological definition of AD in the study may have mixed subjects with AD pathology (e.g., those with positive amyloid-β) with those with severe cognitive deficits, leading to reduced sensitivity in k_for longitudinal changes. Unfortunately, neither amyloid-β nor tau PET imaging was available at the beginning of the CHS-CS.

Fifth, the range of k_for is less than 0.6/s, which is on the lower side of the k_for ranges from the literature, e.g., 1.35±0.19 /s for a rat study [32] and 0.51∼1.43 /s for a human study [30]. The low k_for values result from the large RF frequency offset only partially saturating the macromolecules [60]. Hence, the k_for values we presented are relative, not absolute. In addition, the k_for values are biased by spatially-dependent B₁ inhomogeneities, inversion times, and off-resonance saturation including excitation crosstalk. 3D acquisitions would eliminate the bias from the slice-dependent inversion times and excitation crosstalk but could exacerbate spatially-dependent B₁ inhomogeneities. Nevertheless, we assumed the systematic bias was consistent between individuals in the group analysis. We were able to observe the k_for changes in both cross-sectional and longitudinal studies despite potentially reduced sensitivity associated with these biases.

Sixth, WM hyperintensities were not accounted for in the current analysis. WM hyperintensities are common in older adults with AD subjects having increased WM hyperintensity volume [54, 61]. AD subjects demonstrated decreased MTR in normal-appearing WM and even lower MTR in WM hyperintense regions [54]. Not accounting for WM hyperintensities in this study represents a potential limitation in understanding how different types of WM (normal-appearing WM and hyperintense WM) contributed to reduced k_for in AD.

Conclusion

Our study reveals that voxel-wise forward magnetization transfer rate maps are efficient in detecting GM and WM microstructural damage in AD diagnosis and AD progression. Future studies are needed to investigate the clinical relevance of k_for in posterior cingulate, precuneus, inferior parietal, anterior cingulate, and superior frontal GM regions and corona radiata and superior longitudinal fasciculus WM tracts.

Footnotes

ACKNOWLEDGMENTS

This research was supported by contracts HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, 75N92021D00006, and grants U01HL080295 and U01HL130114 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke (NINDS). Additional support was provided by R01AG023629 from the National Institute on Aging (NIA). A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org. The research was also supported by the State University of New York at Binghamton, the Nevada Cancer Institute, the University of Pittsburgh, and Washington University in St. Louis. Weiying Dai was supported by the National Institute on Aging (NIA) under award number R01AG066430 and the National Science Foundation (NSF) under award number CMMI-2123061. Tony D. Zhou, affiliated with Vestal High School, Vestal, NY, USA, conducted his research under the guidance of Dr. H. Michael Gach.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health and National Science Foundation.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: https://dx.doi.org/10.3233/JAD-220335.

References

Petersen

, Doody

, Kurz

, Mohs

, Morris

, Rabins

, Ritchie

, Rossor

, Thal

, Winblad

(2001) Current concepts in mild cognitive impairment. Arch Neurol 58, 1985–1992.

Ledig

, Schuh

, Guerrero

, Heckemann

, Rueckert

(2018) Structural brain imaging in Alzheimer’s disease and mild cognitive impairment: Biomarker analysis and shared morphometry database. Sci Rep 8, 11258.

Vemuri

, Jack

Jr. (2010) Role of structural MRI in Alzheimer’s disease. Alzheimers Res Ther 2, 23.

Chandra

, Dervenoulas

, Politis

, Alzheimer’s Disease Neuroimaging Initiative (2019) Magnetic resonance imaging in Alzheimer’s disease and mild cognitive impairment. J Neurol 266, 1293–1302.

Alsop

, Dai

, Grossman

, Detre

(2010) Arterial spin labeling blood flow MRI: Its role in the early characterization of Alzheimer’s disease. J Alzheimers Dis 20, 871–880.

Wang

(2014) Characterizing early Alzheimer’s disease and disease progression using hippocampal volume and arterial spin labeling perfusion MRI. , S. J Alzheimers Dis 42 Suppl 4, 495–502.

Brueggen

, Dyrba

, Cardenas-Blanco

, Schneider

, Fliessbach

, Buerger

, Janowitz

, Peters

, Menne

, Priller

, Spruth

, Wiltfang

, Vukovich

, Laske

, Buchmann

, Wagner

, Roske

, Spottke

, Rudolph

, Metzger

, Kilimann

, Dobisch

, Duzel

, Jessen

, Teipel

, DELCODE Study Group (2019) Structural integrity in subjective cognitive decline, mild cognitive impairment and Alzheimer’s disease based on multicenter diffusion tensor imaging. J Neurol 266, 2465–2474.

Molinuevo

, Ripolles

, Simo

, Llado

, Olives

, Balasa

, Antonell

, Rodriguez-Fornells

, Rami

(2014) White matter changes in preclinical Alzheimer’s disease: A magnetic resonance imaging-diffusion tensor imaging study on cognitively normal older people with positive amyloid beta protein 42 levels. Neurobiol Aging 35, 2671–2680.

Mayo

, Garcia-Barrera

, Mazerolle

, Ritchie

, Fisk

, Gawryluk

, Alzheimer’s Disease Neuroimaging Initiative (2018) Relationship between DTI metrics and cognitive function in Alzheimer’s disease. Front Aging Neurosci 10, 436.

10.

Yin

, Tan

, Liu

, Wang

, Jiang

, Radua

, Zhang

, Gao

, Canu

, Migliaccio

, Filippi

, Gorno-Tempini

, Yu

(2015) Multimodal voxel-based meta-analysis of white matter abnormalities in Alzheimer’s disease. J Alzheimers Dis 47, 495–507.

11.

Mayo

, Mazerolle

, Ritchie

, Fisk

, Gawryluk

, Alzheimer’s Disease Neuroimaging Initiative (2017) Longitudinal changes inmicrostructural white matter metrics in Alzheimer’s disease. Neuroimage Clin 13, 330–338.

12.

van Zijl

, Yadav

(2011) Chemical exchange saturation transfer (CEST): What is in a name and what isn’t? Magn Reson Med 65, 927–948.

13.

Haris

, Singh

, Cai

, Davatzikos

, Trojanowski

, Melhem

, Clark

, Borthakur

(2011) T1rho (T_1r) MR imaging inAlzheimer’s disease and Parkinson’s disease with and withoutdementia. J Neurol 258, 380–385.

14.

Wolff

, Balaban

(1994) Magnetization transfer imaging: Practical aspects and clinical applications. Radiology 192, 593–599.

15.

Edzes

, Samulski

(1977) Cross relaxation and spin diffusion in the proton NMR or hydrated collagen. Nature 265, 521–523.

16.

Schmierer

, Scaravilli

, Altmann

, Barker

, Miller

(2004) Magnetization transfer ratio and myelin in postmortem multiple sclerosis brain. Ann Neurol 56, 407–415.

17.

Mottershead

, Schmierer

, Clemence

, Thornton

, Scaravilli

, Barker

, Tofts

, Newcombe

, Cuzner

, Ordidge

, McDonald

, Miller

(2003) High field MRI correlates of myelin content and axonal density in multiple sclerosis–a post-mortem study of the spinal cord. J Neurol 250, 1293–1301.

18.

Schmierer

, Tozer

, Scaravilli

, Altmann

, Barker

, Tofts

, Miller

(2007) Quantitative magnetization transfer imaging in postmortem multiple sclerosis brain. J Magn Reson Imaging 26, 41–51.

19.

Ridha

, Symms

, Tozer

, Stockton

, Frost

, Siddique

, Lewis

, MacManus

, Boulby

, Barker

, Rossor

, Fox

, Tofts

(2007) Magnetization transfer ratio in Alzheimer disease: Comparison with volumetric measurements. AJNR Am J Neuroradiol 28, 965–970.

20.

Bozzali

, Franceschi

, Falini

, Pontesilli

, Cercignani

, Magnani

, Scotti

, Comi

, Filippi

(2001) Quantification of tissue damage in AD using diffusion tensor and magnetization transfer MRI. Neurology 57, 1135–1137.

21.

van der Flier

, van den Heuvel

, Weverling-Rijnsburger

, Bollen

, Westendorp

, van Buchem

, Middelkoop

(2002) Magnetization transfer imaging in normal aging, mild cognitive impairment, and Alzheimer’s disease. Ann Neurol 52, 62–67.

22.

Kabani

, Sled

, Chertkow

(2002) Magnetization transfer ratio in mild cognitive impairment and dementia of Alzheimer’s type. Neuroimage 15, 604–610.

23.

Hanyu

, Asano

, Iwamoto

, Takasaki

, Shindo

, Abe

(2000) Magnetization transfer measurements of the hippocampus in patients with Alzheimer’s disease, vascular dementia, and other types of dementia. AJNR Am J Neuroradiol 21, 1235–1242.

24.

Hanyu

, Asano

, Sakurai

, Takasaki

, Shindo

, Abe

(2001) Magnetization transfer measurements of the hippocampus in the early diagnosis of Alzheimer’s disease. J Neurol Sci 188, 79–84.

25.

Sled

(2018) Modelling and interpretation of magnetization transfer imaging in the brain. Neuroimage 182, 128–135.

26.

Duan

, Zhou

, Balachandrasekaran

, Bhumkar

, Boraste

, Becker

, Kuller

, Lopez

, Gach

, Dai

(2021) Cerebral blood flow predicts conversion of mild cognitive impairment into Alzheimer’s disease and cognitive decline: An arterial spin labeling follow-up study. J Alzheimers Dis 82, 293–305.

27.

Duan

, Sehrawat

, Balachandrasekaran

, Bhumkar

, Boraste

, Becker

, Kuller

, Lopez

, Gach

, Dai

(2020) Cerebral blood flow is associated with diagnostic class and cognitive decline in Alzheimer’s disease. J Alzheimers Dis 76, 1103–1120.

28.

Dai

, Lopez

, Carmichael

, Becker

, Kuller

, Gach

(2009) Mild cognitive impairment and Alzheimer disease: Patterns of altered cerebral blood flow at MR imaging. Radiology 250, 856–866.

29.

Duan

, Sehrawat

, Becker

, Lopez

, Gach

, Dai

(2019) White matter damage of Alzheimer’s disease using magnetization transfer imaging. Proc Int Soc Magn Reson Med 27, 625.

30.

Ropele

, Strasser-Fuchs

, Augustin

, Stollberger

, Enzinger

, Hartung

, Fazekas

(2000) A comparison of magnetization transfer ratio, magnetization transfer rate, and the native relaxation time of water protons related to relapsing-remitting multiple sclerosis. AJNR Am J Neuroradiol 21, 1885–1891.

31.

Silva

, Zhang

, Williams

, Koretsky

(1997) Estimation of water extraction fractions in rat brain using magnetic resonance measurement of perfusion with arterial spin labeling. Magn Reson Med 37, 58–68.

32.

Zhang

, Silva

, Williams

, Koretsky

(1995) NMR measurement of perfusion using arterial spin labeling without saturation of macromolecular spins. Magn Reson Med 33, 370–376.

33.

Herscovitch

, Raichle

(1985) What is the correct value for the brain-blood partition coefficient for water? J Cereb Blood Flow Metab 5, 65–69.

34.

Lopez

, Jagust

, DeKosky

, Becker

, Fitzpatrick

, Dulberg

, Breitner

, Lyketsos

, Jones

, Kawas

, Carlson

, Kuller

(2003) Prevalence and classification of mild cognitive impairment in the Cardiovascular Health Study Cognition Study: Part 1. Arch Neurol 60, 1385–1389.

35.

Teng

, Chui

(1987) The Modified Mini-Mental State (3MS) examination. J Clin Psychiatry 48, 314–318.

36.

McDowell

, Kristjansson

, Hill

, Hebert

(1997) Community screening for dementia: The Mini Mental State Exam (MMSE) and Modified Mini-Mental State Exam (3MS) compared. J Clin Epidemiol 50, 377–383.

37.

Nadler

, Relkin

, Cohen

, Hodder

, Reingold

, Plum

(1995) Mental status testing in the elderly nursing home population. J Geriatr Psychiatry Neurol 8, 177–183.

38.

Rovaris

, Agosta

, Sormani

, Inglese

, Martinelli

, Comi

, Filippi

(2003) Conventional and magnetization transfer MRI predictors of clinical multiple sclerosis evolution: A medium-term follow-up study. Brain 126, 2323–2332.

39.

Brex

, Leary

, Plant

, Thompson

, Miller

(2001) Magnetization transfer imaging in patients with clinically isolated syndromes suggestive of multiple sclerosis. AJNR Am J Neuroradiol 22, 947–951.

40.

Foong

, Symms

, Barker

, Maier

, Woermann

, Miller

, Ron

(2001) Neuropathological abnormalities in schizophrenia: Evidence from magnetization transfer imaging. Brain 124, 882–892.

41.

Yarnykh

(2002) Pulsed Z-spectroscopic imaging of cross-relaxation parameters in tissues for human MRI: Theory and clinical applications. Magn Reson Med 47, 929–939.

42.

Giulietti

, Bozzali

, Figura

, Spano

, Perri

, Marra

, Lacidogna

, Giubilei

, Caltagirone

, Cercignani

(2012) Quantitative magnetization transfer provides information complementary to grey matter atrophy in Alzheimer’s disease brains. Neuroimage 59, 1114–1122.

43.

Makovac

, Serra

, Di Domenico

, Marra

, Caltagirone

, Cercignani

, Bozzali

(2018) Quantitative magnetization transfer of white matter tracts correlates with diffusion tensor imaging indices in predicting the conversion from mild cognitive impairment to Alzheimer’s disease. J Alzheimers Dis 63, 561–575.

44.

Kiefer

, Brockhaus

, Cattapan-Ludewig

, Ballinari

, Burren

, Schroth

, Wiest

(2009) Multi-parametric classification of Alzheimer’s disease and mild cognitive impairment: The impact of quantitative magnetization transfer MR imaging. Neuroimage 48, 657–667.

45.

Ridha

, Tozer

, Symms

, Stockton

, Lewis

, Siddique

, MacManus

, Rossor

, Fox

, Tofts

(2007) Quantitative magnetization transfer imaging in Alzheimer disease. Radiology 244, 832–837.

46.

Marcus

, Mena

, Subramaniam

(2014) Brain PET in the diagnosis of Alzheimer’s disease. Clin Nucl Med 39, e413–422.quiz e423-416.

47.

Kato

, Inui

, Nakamura

, Ito

(2016) Brain fluorodeoxyglucose (FDG) PET in dementia. Ageing Res Rev 30, 73–84.

48.

Louie

, Gochberg

, Does

, Damon

(2009) Transverse relaxation and magnetization transfer in skeletal muscle: Effect of pH. Magn Reson Med 61, 560–569.

49.

Moreira

, Carvalho

, Zhu

, Smith

, Perry

(2010) Mitochondrial dysfunction is a trigger of Alzheimer’s disease pathophysiology. Biochim Biophys Acta 1802, 2–10.

50.

Falini

, Bozzali

, Magnani

, Pero

, Gambini

, Benedetti

, Mossini

, Franceschi

, Comi

, Scotti

, Filippi

(2005) A whole brain MR spectroscopy study from patients with Alzheimer’s disease and mild cognitive impairment. Neuroimage 26, 1159–1163.

51.

Signoretti

, Marmarou

, Tavazzi

, Lazzarino

, Beaumont

, Vagnozzi

(2001) N-Acetylaspartate reduction as a measure of injury severity and mitochondrial dysfunction following diffuse traumatic brain injury. J Neurotrauma 18, 977–991.

52.

Witte

, Mahad

, Lassmann

, van Horssen

(2014) Mitochondrial dysfunction contributes to neurodegeneration in multiple sclerosis. Trends Mol Med 20, 179–187.

53.

Carvalho

(2013) Mitochondrial dysfunction in demyelinating diseases. Semin Pediatr Neurol 20, 194–201.

54.

Colonna

, Koini

, Pirpamer

, Damulina

, Hofer

, Schwingenschuh

, Enzinger

, Schmidt

, Ropele

(2021) Microstructural tissue changes in Alzheimer disease brains: Insights from magnetization transfer imaging. AJNR Am J Neuroradiol 42, 688–693.

55.

Ropele

, Schmidt

, Enzinger

, Windisch

, Martinez

, Fazekas

(2012) Longitudinal magnetization transfer imaging in mild to severe Alzheimer disease. AJNR Am J Neuroradiol 33, 570–575.

56.

Bozzali

, Falini

, Franceschi

, Cercignani

, Zuffi

, Scotti

, Comi

, Filippi

(2002) White matter damage in Alzheimer’s disease assessed using diffusion tensor magnetic resonance imaging. }. J Neurol Neurosurg Psychiatry 72, 742–746.

57.

Choi

, Lim

, Monteiro

, Reisberg

(2005) Diffusion tensor imaging of frontal white matter microstructure in early Alzheimer’s disease: A preliminary study. J Geriatr Psychiatry Neurol 18, 12–19.

58.

Hoy

, Ly

, Carlsson

, Okonkwo

, Zetterberg

, Blennow

, Sager

, Asthana

, Johnson

, Alexander

, Bendlin

(2017) Microstructural white matter alterations in preclinical Alzheimer’s disease detected using free water elimination diffusion tensor imaging. PLoS One 12, e0173982.

59.

Genc

, Steward

, Malpas

, Velakoulis

, O’Brien

, Desmond

(2016) Short-term white matter alterations in Alzheimer’s disease characterized by diffusion tensor imaging. J Magn Reson Imaging 43, 627–634.

60.

Finelli

, Reed

(1998) Flip angle dependence of experimentally determined T1sat and apparent magnetization transfer rate constants. J Magn Reson Imaging 8, 548–553.

61.

Wang

, Chen

, Cai

, Hu

, Xu

, Wang

, Cao

, Tan

, Yu

, Alzheimer’s Disease Neuroimaging Initiative (2020) Associations of white matter hyperintensities with cognitive decline: A longitudinal study. J Alzheimers Dis 73, 759–768.

Pattern of Altered Magnetization Transfer Rate in Alzheimer’s Disease

Abstract

Background:

Objective:

Methods:

Results:

Conclusion:

Keywords

INTRODUCTION

METHODS

Theory

Participants

MRI protocol

Image analysis

Calculation of T1sat and T1nosat

Calculation of k for

Normalization of T1sat, T1nosat, and k for maps

Statistical analysis

Generation of global regions of interest (ROIs) (whole brain, GM, and WM)

Voxel-level analysis on k for maps at baseline

Voxel-level analyses on longitudinal changes of k for

Region-level analyses on longitudinal changes of k for

RESULTS

Demographic and cognitive information

Cross-sectional comparisons of T1nosat, T1sat, and k for in the global ROIs

Cross-sectional voxel-wise group comparisons

Longitudinal voxel-wise group comparisons

Longitudinal regional group comparisons

DISCUSSION

Conclusion

Footnotes

ACKNOWLEDGMENTS

References

Calculation of T_1sat and T_1nosat

Calculation of k_for

Normalization of T_1sat, T_1nosat, and k_for maps

Voxel-level analysis on k_for maps at baseline

Voxel-level analyses on longitudinal changes of k_for

Region-level analyses on longitudinal changes of k_for

Cross-sectional comparisons of T_1nosat, T_1sat, and k_for in the global ROIs