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

Abstract

Patients with amnestic mild cognitive impairment (aMCI) have higher probability to develop Alzheimer’s disease (AD) than elderly controls. The detection of subtle changes in brain structure associated with disease progression and the development of tools to identify patients at high risk for dementia in a short time is crucial. Here, we used probabilistic white matter (WM) tractography to explore microstructural alterations within the main association, limbic, and commissural pathways in aMCI patients who converted to AD after 1 year follow-up (MCI_converters) and those who remained stable (MCI_stable). Both diffusion tensor imaging (DTI) and quantitative magnetization transfer (qMT) parameters have been considered for a comprehensive pathophysiological characterization of the WM damage. Overall, tract-specific parameters derived from qMT and DTI at baseline were able to differentiate aMCI patients who converted to AD from those who remained stable in time. In particular, the qMT exchange rate, RM^B ₀, of the right uncinate fasciculus was significantly decreased in MCI_converters, whereas fractional anisotropy was significantly decreased in the bilateral superior cingulum in MCI_converters compared to MCI_stable. These results confirm the involvement of WM and particularly of association fibers in the progression of AD, highlighting disconnection as a potential mechanism.

Keywords

Diffusion tensor imaging mild cognitive impairment probabilistic tractography quantitative magnetization transfer white matter

INTRODUCTION

Mild cognitive impairment of amnestic type (aMCI) is considered a pre-dementia phase, highly predictive of subsequent conversion to Alzheimer’s disease (AD) [1]. Nonetheless, aMCI is a fairly heterogeneous condition [2, 3], including not only patients at an early AD stage, but also patients who will remain stable, and patients at risk for developing other forms of dementia [4]. Determining the biophysical sensitivity and specificity of quantitative magnetic resonance imaging to AD pathology is essential to develop effective imaging biomarkers of neurodegeneration. While regional assessments of grey matter (GM) atrophy are sensitive to macroscopic changes occurring at relatively advanced phases of the disease [5, 6], this approach is unable to highlight microstructural alterations that accompany and may well precede volume loss. The hypothesis that disconnection may be a contributor to the spread of damage in AD brains is increasingly gaining consensus [7], thus implying that early changes within the white matter (WM) tracts might play a crucial role in the onset and early progression of AD [7 –9], up to the hypothesis that myelin breakdown might be implicated in the pathogenesis of AD [10]. This view is supported by evidence of microstructural changes occurring in the WM already in pre-clinical stages of AD [11]. It still remains to be established, however, whether WM microstructural changes are a primary event, or secondary to GM atrophy [12].

Diffusion tensor imaging (DTI) and magnetization transfer (MT) imaging are among the most popular MRI techniques for assessing WM changes. They are sensitive to micro-structural and metabolic changes, and therefore might provide more accurate prognostic information than atrophy assessment on its own. DTI allows the estimation of a number of indices of microstructural integrity of WM fiber tracts [13, 14]. Among those, fractional anisotropy (FA) reflects the coherence of the orientation of water diffusion, and its variation is influenced by different factors including myelination, axon density, axonal membrane integrity, and axon diameter [15, 16]. Mean diffusivity (MD) is determined by the density of physical obstructions to diffusion such as membranes, and quantifies the resultant distribution of water molecules between different cell compartments [16]. Higher MD values indicate increased diffusion which suggests tissue breakdown and increased brain water content [15]. As diverse pathological events can lead to similar changes in the observed FA and MD, it was proposed that parameters quantifying separately the diffusivity along and across WM tracts—namely axial diffusivity (ADif) and radial diffusivity (RDif)—can be more informative about the integrity of axons versus their surrounding myelin sheaths [17]. Increased RDif has been linked to loss of myelin in animal studies of experimentally induced myelin loss [17 –19], while a decrease in ADif has been reported in both rodents and humans with axonal damage associated with axonal swelling, fragmentation and organelle accumulations [19, 20].

MT imaging is based on the exchange of magnetization between protons in tissue water (liquid pool) and those bound to macromolecules (macromolecular pool). In particular, as some evidence suggests that molecules associated with myelin dominate the MT exchange process in WM, it is conceivable that this technique may provide information about myelin integrity. A simple way to quantify this effect is the MT ratio (MTR) [21], which has been extensively applied to the study of neurological conditions, including dementia. Previous investigations have demonstrated reductions of the MTR in the brain of patients with AD [22, 23] and associations between MTR and measures of global cognitive impairment [24 –26], suggesting that the MTR is sensitive to at least some of the pathological processes characterizing AD. The MTR value, however, is the result of the combination of several more fundamental quantities, and it is known to be affected also by inflammation, T1 changes, and pH [27]. Quantitative MT (qMT) [28] is an extension of MT imaging which attempts to overcome the limitation of MTR by quantifying the physical properties that govern the MT process, including the relaxation rates of the pools, the exchange rate, and the relative size of the macromolecular pool. These parameters deliver information closely related to biological changes of WM and GM, and are therefore of potential interest in several neurological disorders. In particular, preliminary studies suggest that qMT might provide complementary information to that offered by other MRI techniques in the characterization of neurodegenerative diseases, such as AD [23 , 30].

Postmortem studies [31] and animal studies [32 –34] indicate that one of the parameters obtained from qMT, the relative size of the macromolecular pool, or F, is proportional to myelin content in WM [35], while preliminary studies indicates that another qMT parameter, namely the forward exchange rate, RM₀ ^B, might be more sensitive to AD pathology [29, 30]. The interpretation of RM₀ ^B is uncertain, as the exact mechanisms of MT are still unknown. Based on results obtained in the cortex of patients with AD, it was suggested that it might reflect metabolic abnormalities [30], while recent reports indicate that it might be related to mild inflammatory changes [36, 37].

Table 1

Principal demographic and clinical characteristics of the studied subjects

			MCI
	AD	HC	Converters	Stable
Mean (SD) age at scan (years)	70.77 (7.07)	61.48 (6.97) *	61.00 (23.83) *	65.88 (16.10)
Sex (F/M)	29/26	12/19	12/5	7/10
Mean (SD) years of formal education	9.44 (4.19)	13.86 (3.26) *	7.23 (5.04)^#	10.75 (4.83)
Mean (SD) MMSE score	19.11 (3.80)	29.39 (1.03) *	25.42 (2.57) * #	27.23 (1.94) *
IADL	5.03 (1.97)	0.00	6.64 (1.73)	6.33 (1.63)
Mean (SD) CDR score	1.19 (0.67)	0.00	0.50 (0.19)	0.50 (0.02)

*All versus AD; ^#MCI (converters, stable) versus HC; AD, Alzheimer’s disease; HC, healthy controls; MCI, mild cognitive impairment.

Here, we combined DTI and qMT to provide a detailed examination of WM abnormalities in aMCI. Our primary aim was to identify tract-specific biomarkers obtained from DTI and qMT able to predict the conversion to AD in patients with aMCI over 1 year. In addition, secondary aims were 1) to assess the microstructural damage of the WM tracts which were recently shown to be involved in AD pathophysiology [38] in patients with MCI; and 2) to evaluate the relationship between qMT and DTI indices within those tracts. Our attention has focused on the cingulum and uncinate fasciculus, whose early involvement in AD pathology and relation to cognitive decline has been described by previous studies [9, 39]. The inferior longitudinal fasciculus has also been considered, given its proximity in the anterior part to the uncinate fasciculus.

MATERIALS AND METHODS

Subjects

A cohort of 43 patients with a diagnosis of probable AD [40], 34 patients with amnestic mild cognitive impairment (aMCI) (as classified following the original Mayo Clinic criteria; [41]), and 21 healthy controls (HC) were enrolled for this study. To classify our MCI patients as a prodromal-AD group, we adopted two main criteria: 1) patients had to report subjective memory impairment at clinical onset, corroborated by an assistant, and confirmed by performances below the normality cut-off scores on at least one of the administered tests for episodic memory; 2) their MRI profile had to be compatible with prodromal-AD (i.e., structural alterations in temporal areas).

Patients were reviewed after 1 year, in order to assess whether they had converted to AD or remained stable. Patients were recruited from the specialist dementia clinic of the Catholic University of Rome (Rome, Italy). All patients had a typical clinical onset of AD, characterized by episodic memory deficits. The principal demographic and clinical characteristics of the participants are summarized in Table 1.

aMCI patients were identified according to current criteria [41]. They had to report a subjective memory impairment corroborated by an assistant and confirmed by performance below the normality cut off scores on at least one of the administered tests for episodic memory (see Table 2), and all MCI patients had not to fulfil the Diagnostic and Statistical Manual of Mental Disorders (DSM-V) criteria for the diagnosis of major neurocognitive disorders [42]. This criterion was confirmed by reporting normal scores at the Mini-Mental State Examination (MMSE) [43, 44] (Italian normality cut-off score > 23.8) [45] and at all tests of the neuropsychological battery assessing cognitive functions other than episodic memory (Table 2). Finally, the memory deficits observed in aMCI patients had to result in no or in a very mild impact on their daily living activities, as confirmed by their total Clinical Dementia Rating (CDR) score [46] which had not to exceed 0.5. None of the aMCI patients was either under anticholinesterase or neuroleptic medication. Patients with probable AD were diagnosed according to the clinical criteria established by the National Institute of Neurological and Communicative Disorders and Stroke Alzheimer’s Disease and Related Disorders Association (NINCDSADRDA) [40].

Table 2

Performance scores obtained by HC, AD, and MCI patients

				MCI
Domain	Test	AD	HC	Converters	Stable
Long Term Memory:
-Verbal	15-Words List:
	Immediate recall (cut-off≥28.5)	21.33 (9.13)	46.26 (8.58)	28.19 (5.84) *^#	32.88 (7.35) * #
	Delayed recall (cut-off≥4.6)	1.79 (1.97)	9.82 (2.04)	3.32 (2.60) # ^∧	5.53 (1.88) * #
	Short Story test:
	Immediate recall (cut-off≥3.1)	1.91 (2.00)	6.48 (1.24)	3.43 (2.80)	4.91 (1.58) *
	Delayed recall (cut-off≥2.9)	0.95 (1.83)	6.21 (1.05)	2.63 (2.40) * #	4.43 (2.03) * #
-Visuo-spatial	Complex Rey’s Figure:
	Immediate recall (cut-off≥6.4)	3.73 (4.29)	15.39 (7.30)	9.29 (7.48) #	12.73 (8.07)
	Delayed recall (cut-off≥6.3)	3.48 (4.58)	14.39 (5.83)	8.20 (7.00) #	12.15 (5.57)
Short-term memory:
-Verbal	Digit span (cut-off≥3.7)	4.97 (1.19)	5.75 (0.98)	4.63 (1.12) #	5.19 (0.94)
-Visuo-spatial	Corsi span (cut-off≥3.5)	3.28 (1.44)	4.99 (0.82)	3.97 (0.72)	4.44 (0.43) *
Executive functions:	Phonological Word Fluency (cut-off≥17.3)	21.96 (10.72)	36.04 (8.28)	29.88 (6.63) *	34.78 (10.38) *
	Digit span backward	2.76 (1.60)	4.75 (0.85)	3.35 (0.70) #	3.94 (0.75) *
	Corsi span backward	2.15 (1.84)	6.69 (8.36)	3.44 (0.96)	4.29 (0.69)
Language:	Naming of objects (cut-off≥22)	20.92 (10.14)	29.15 (1.14)	27.71 (2.08)	28.82 (1.47)
Reasoning:	Raven’s Colored Progressive Matrices (cut-off≥18.9)	21.29 (6.57)	31.06 (5.53)	25.28 (4.08) #	28.99 (4.39) *
Constructional praxis:	Copy of Drawings (cut-off≥7.1)	6.42 (3.72)	11.04 (1.16)	8.94 (1.45)	10.19 (1.36) *
	Copy of Drawings with landmarks (cut-off≥61.8)	50.28 (22.92)	69.20 (0.90)	65.31 (4.54)*	67.66 (2.86) *
	Copy of Complex Rey’s Figure (cut-off≥23.7)	18.26 (12.66)	32.67 (2.57)	25.39 (9.36)	32.46 (4.53) *

*MCI (conv, stab) versus AD; ^#MCI (converters, stable) versus HC; ^∧MCIconverters versus MCIstable. AD, Alzheimer’s disease; HC, healthy controls; MCI, mild cognitive impairment. For each studied group mean (SD) scores obtained in each administered test are reported. For each administered test appropriate adjustments for gender, age and education level were applied according to the Italian normative data. Available cut-off scores of normality (≥95% of the lower tolerance limit of the normal population distribution) are also reported for each test.

HC were enrolled among patients’ relatives and spouses. HC had to report no previous complaint suggestive of cognitive impairment, and they had to score within the range of normality at every administered neuropsychological test (Table 2).

All recruited subjects had to be right-handed, in order to reduce any potential source of variability due to hemispheric dominance. Major systemic, psychiatric, and other neurological illnesses were carefully investigated and excluded in all the studied subjects. This research study was approved by the Ethics Committee of Santa Lucia Foundation, according to the principles expressed in the Declaration of Helsinki. Written informed consent (either from the participants or from their responsible guardians if incapable) were obtained before study initiation.

Neuropsychological assessment

All recruited patients underwent an extensive neuropsychological battery administered by two trained neuropsychologists on the same day of the MRI acquisition and at follow-up. Global cognitive function was assessed by means of the MMSE [43, 44]. Patients underwent also a complete battery of tests specific for each cognitive domain: 1) Verbal episodic long-term memory: Immediate and Delayed recall of a 15-Word List [47], Short Story (Immediate and Delayed) [48]; 2) Visuo-spatial episodic long term memory: Immediate and Delayed recall of Complex Rey’s Figure [48]; 3) short-term memory: Digit span forward and Corsi Block Tapping task forward [49]; 4) executive functions: Phonological Word Fluency [47]. Digit span backward and Corsi Block Tapping task backward [49] Language: Naming objects subtest of the BADA (“Batteria per l’Analisi dei Deficit Afasici”, Italian for “Battery for the analysis of aphasic deficits”); and 5); Problem-solving: Raven’s Colored Progressive Matrices [47]; Praxis: Copy of drawings with and without landmarks [47]; Copy of Complex Rey’s Figure [48]. For all administered tests, we used the Italian normative data for score adjustment (gender, age, and education) and to define cut-off scores of normality, determined as the lower limit of the 95% tolerance interval for a confidence level of 95%. For each test, normative data are reported in the corresponding references. Group comparisons were estimated by seventeen one-way ANOVAs. To avoid the type-I error Bonferroni’s correction was applied (statistical threshold for significance, p = 0.003, α= 0.05/16).

MRI acquisition

All MRI was obtained using a head-only 3T scanner (Siemens Magnetom Allegra, Siemens Medical Solutions, Erlangen, Germany), equipped with a circularly polarized transmit–receive coil. The maximum gradient strength is 40mTm^-1, with a maximum slew rate of 400T/m/s. The following sequences were obtained for each subject during a single scanning session: 1) dual-echo turbo spin echo (TSE) (TR = 6190 ms, TE1 = 12 ms, TE2 = 109 ms, echo train length (ETL) = 5; matrix = 256×192; field of view (FOV) = 230×172.5mm²; 48 contiguous 3 mm thick slices); 2) fluid attenuated inversion recovery (FLAIR) (TR = 8170 ms, TE = 96 ms, TI = 2100 ms; ETL = 13; same FOV, matrix and number of slices as TSE); 3) morphological 3D T1-weighted magnetization prepared rapid acquisition gradient echo (MPRAGE) (TE = 2.74 ms, TR = 2500 ms, inversion time = 900 ms; flip angle = 8; matrix = 256×208×176; FOV = 256×208×176mm³); 4) diffusion weighted twice-refocused SE EPI (TR = 170 ms, TE = 85 ms, maximum b factor = 1000smm², isotropic resolution 2.3mm³; matrix = 9696; 60 slices), collecting seven images with no diffusion weighting (b0) and 61 images with diffusion gradients applied along 61 non-collinear directions [50]; 5) a series of 12 MT-weighted 3D fast low-angle shot (FLASH) sequences (TR = 35 ms, TE = 7.4, flip angle = 7; matrix = 128×96×28; FOV = 230×172.5×140 mm³), with various combinations of on-resonance equivalent flip angle and offset frequency (Δ) of the Gaussian MT pulse (pulse duration = 15 ms). The optimized set of sampling points matched that described in Cercignani et al. [51]; 6) three 3D FLASH sequences with three different flip angles for mapping the observed T1 of the system (TR = 15 ms, TE = 4.8 ms, flip angle = 5, 7, 15, respectively; same matrix and FOV as the MT-weighted FLASH); 7) three 3D FLASH sequences with near-180 flip angles for B1 mapping [52] (TR = 28 ms, TE = 4.8 ms, flip angles = 155, 180, 205, respectively; matrix = 64×64×40, FOV = 220×220×160 mm³); this sequence was collected along the sagittal plane, while all the other sequences were collected in the near-axial plane, with slices positioned to run parallel to a line that joins the most inferoanterior and inferoposterior parts of the corpus callosum (AC–PC line). The total scan time was approximately 60 min.

MR images post-processing and analysis

Diffusion tensor MR imaging and probabilistic tractography

Correction for eddy currents and the effects of involuntary movements on the DTI was by means of affine registration of each diffusion-weighted volume to the first non-diffusion weighted volume using FSL. Then FA, MD, RDif, and ADif were computed from the diffusion tensor (DT) fitted with weighted linear least-square using Camino [53].

The WM tracts of interest were: the inferior longitudinal fasciculus (ILF), the interior (ventral) and superior (dorsal) cingulum bundles (infCi and SupCi), and the uncinated fasciculus (UF). The tracts were reconstructed in native space for all participants, using multi-fiber probabilistic tractography. Tractography was carried out using 10000 iterations of the probabilistic index of connectivity (PICo) algorithm [53] applied to fibre orientation distribution functions estimated with persistent angular structure (PAS) MRI. Seed and way points to reconstruct the tracts of interest were defined on each subject’s colour coded map according to published guidelines. Anatomical images were used to increased confidence in seed-point selection.

Quantitative MT analysis

For all the subjects, FLASH images from sequence 7 were reformatted to the axial plane. Next, in order to compensate for involuntary motion, all images obtained from sequences 5 (MT FLASH), 6 (T1-mapping FLASH) and 7 (reformatted B1-mapping FLASH) were affine-registered using FLIRT [54, 55] to match the 15° flip-angle scan obtained with sequence 6. B1 maps were computed according to Dowell and Tofts, [52] and T1 maps were calculated by fitting the theoretical spoiled gradient echo signal equation [56] as a function of the flip angle (after B1-correction) to the signal measured by sequence 6 [57]. A detailed description of the processing is given elsewhere [51]. For MT modelling, we adopted here Ramani’s signal equation [58] a modification of Henkelman’s model [28] for in vivo applications, where the continuous wave power equivalent (CWPE) approximation is used. As described in detail elsewhere [51, 59] quantitative MT parameters were estimated on a voxel-by-voxel basis by fitting the model equation of the MT-weighted signal to the data from sequence 5 using the Levenberg–Marquardt method, as implemented in Numerical Recipes [60]. The absorption lineshape of the macromolecular pool was modelled with a super-Lorentzian [61]. The longitudinal relaxation rate of the macromolecular pool, RM₀ ^B, was set arbitrarily to 1 s^-1 and kept fixed during the fitting, as in previous quantitative MT experiments [28 , 63]. The fitting yielded maps of F and RM₀ ^B.

Extraction of tract-specific indices

For every subject, the spatial transformation matching the DTI scan to the MT-weighted scans was then computed by affine registration and followed by nonlinear registration (see FLIRT and FNIRT, http://www.fmrib.ox.ac.uk/fsl/fsl/list.html). The non-linear step was added to compensate for EPI geometric distortions. The transformations were then applied to the maps defining the WM tracts which were used to compute the tract mean F, RM₀ ^B. The tract average FA, MD, ADif, and RDif were also computed.

Statistical analysis

Statistical analyses were performed using SPSS 15.0 for Windows (SPSS Inc, USA). Age and gender were compared between groups using a one-way ANOVA and a χ ²-test, respectively. Between-group comparisons of tract volumes, mean qMT (F and RM₀ ^B) and DTI (FA, MD, ADif, and RDif) parameters were performed using a one-way ANOVA, Tukey tests were used to assess group differences (post-hoc analyses). The significance level for the ANOVAs was set to p = 0.008, after Bonferroni’s correction (α= 0.05/6). The presence of any association between DTI and qMT parameters was assessed using the Pearson correlation. The significance level was set after Bonferroni’s correction (p = 0.015; α= 0.05/4).

RESULTS

After 1 year, 17 aMCI remained stable (MCI_stable) and 17 converted to probable AD (MCI_converters). Table 1 summarizes the demographical and clinical characteristics of all the patients and HC, together with the statistical differences. A significant difference in age was detected between HC and AD, and between MCI_converters and AD (p < 0.05 in both cases). Moreover, on average, HC had received more years of education when compared to AD (p < 0.01). Both age and years of education were introduced in the statistical analysis as variables of no interest. As expected, the baseline MMSE scores were significantly lower in AD when compared to MCI_converters, MCI_stable and HC_, and significantly lower in MCI_converters when compared to HC. No significant differences were found between MCI_converters and MCI_stable.

Neuropsychological and behavioral assessments

Neuropsychological data for the group of AD and MCI patients and HC volunteers are summarized in Table 2. As expected, the performance of AD patients was significantly poorer than HC in all cognitive domains. Both MCI groups performed worse than HC in immediate and delayed recall of the 15-word list, and delayed recall of short story test. MCI_converters but not MCI_stable performed worse than HC in immediate and delayed recall of Complex Rey’s figure, verbal digit span forward and backward, modified card sorting test and Raven’s colored progressive matrices. Conversely, when compared to AD, both MCI_converters but not MCI_stable reported better performance in immediate recall of 15 words, delayed recall of short story, world fluency test, copy of drawings with landmarks. MCI_stable but not MCI_converters performed better than AD in delayed recall of 15 words, Corsi span, digit span backward, Raven’s colored progressive matrices and copy of drawings. When compared to each other, MCI_converters performed worse than MCI_stable at the delayed recall of 15 word list and modified card sorting test (see Table 2).

Fig.1

Average group FA (A) and MD (B) values in the tracts of interest. Data are reported only for tracts that showed at least one between group significant difference. Differences between AD and HC are widespread. By contrast, only FA of supCI differentiated MCIconverters from MCIstable already at baseline. AD, Alzheimer’s disease; HC, healthy controls; MCIconv, MCI that converted to AD at follow-up; MCIstab, MCI that remained stable (did not convert to AD) at follow-up; Fa, franctional anisotropy; MD, mean diffusivity; sup, superior; inf, inferior; UF, uncinated fasciculus; ILF, inferior longitudinal fasciculus.

Fig.2

Grup-averaged RDif (A) and ADif (B) values in the tracts of interest. Data are reported only for tracts that showed at least one between group significant difference. None of these indices was able to differentiate MCconv from MCstab. AD, Alzheimer’s disease; HC, healthy controls; MCIconv, MCI that converted to AD at follow-up; MCIstab, MCI that remained stable (did not convert to AD) at follow-up; RDif, radial diffusivity; ADif, axial diffusivity; sup, superior; inf, inferior; UF, uncinated fasciculus; ILF, inferior longitudinal fasciculus.

DTI results

We first investigated the presence of between-group differences in the volume of WM tracts, in order to exclude that difference in DTI and qMT indices were affected by partial volume related to the overall atrophy rather than to mictrostructural damage. A one-way ANOVA reported significant differences in the volume of infCI (for both left and right p < 0.005) and ILF (for both left and right p = 0.05) in AD and MCI_converters reported reduced volume in left infCI when compared to HC, whereas only MCI_converters reported reduced volume in right infCi when compared to HC (all p < 0.005). MCI_stable reported reduced volume in left ILF when compared to HC (p < 0.05). Volumes of infCi and ILF have been introduced as covariates of no interest in the subsequent analysis.

Significant differences were found in all DTI measures when comparing the three groups (Figs. 1 and 2). Overall, AD patients presented changes in all DTI parameters (but FA) in all tracts when compared to HC. Several differences were found also when comparing the two subgroups of MCI (MCI_converters and MCI_stable) to HC, as shown in Fig. 1 (FA and MD) and Fig. 2 (RDif and ADif). The only significant differences between MCI_converters and MCI_stable were found with FA of the supCi.

Table 3

Correlation between DTI (FA, MD, RDif, ADif) and qMT (F, RM₀ ^B) indices

			F	RM₀ ^B				F	RM₀ ^B
supCi Left	FA	r	0.161	0.131	supCi Right	FA	r	0.252^*	0.072
		p	0.119	0.206			p	0.013	0.482
	MD	r	–0.264**	–0.488^**		MD	r	–0.235^*	–0.413^**
		p	0.010	0.000			p	0.021	0.000
	ADif	r	–0.100	–0.143		ADif	r	–0.125	–0.198
		p	0.339	0.172			p	0.232	0.057
	RDif	r	–0.173	–0.276**		RDif	r	–0.214*	–0.259^*
		p	0.098	0.008			p	0.040	0.012
infCi Left	FA	r	–0.031	0.187	infCi Right	FA	rr	0.040	0.012
		p	0.761	0.066			p	0.697	0.910
	MD	r	–0.188	–0.600^**		MD	r	–0.053	–0.389**
		p	0.065	0.000			p	0.613	0.000
	ADif	r	–0.085	–0.480^**		ADif	r	0.004	–0.300^**
		p	0.414	0.000			p	0.968	0.004
	RDif	r	–0.114	–0.517^**		RDif	r	–0.038	–0.275^**
		p	0.274	0.000			p	0.720	0.008
UF Left	FA	r	0.373**	0.172	UF Right	FA	r	0.191	0.229^*
		p	0.000	0.097			p	0.066	0.027
	MD	r	–0.247*	–0.067		MD	r	–0.231^*	–0.465^**
		p	0.016	0.520			p	0.026	0.000
	ADif	r	–0.127	–0.228^*		ADif	r	–0.025	–0.235^*
		p	0.224	0.028			p	0.817	0.025
	RDif	r	–0.283^**	–0.319**		RDif	r	–0.119	–0.467^**
		p	0.006	0.002			p	0.261	0.000
ILF Left	FA	r	0.130	0.274^**	ILF Right	FA	r	0.204^*	0.271^**
		p	0.208	0.007			p	0.047	0.008
	MD	r	–0.023	–0.349**		MD	r	–0.014	–0.165
		p	0.821	0.000			p	0.889	0.107
	ADif	r	0.025	–0.162		ADif	r	0.052	–0.051
		p	0.813	0.116			p	0.614	0.626
	RDif	r	–0.023	–0.294^**		RDif	r	–0.020	–0.149
		p	0.823	0.004			p	0.844	0.150

FA, fractional anisotropy; MD, mean diffusivity; RDif, radial diffusivity; ADif, axial diffusivity; RM₀ ^B, forward exchange rate; F, relative size of the macromolecular pool; supCi, superior cingulum; infCi, inferior cinguilum; UF, uncinated fasciculus; ILF, inferior longitudinal fasciculus. Significant p-values are printed in Bold.

qMT results

Among the three groups, significant differences were found only with RM₀ ^B. When compared to HC, AD patients reported lower RM₀ ^B in left ILF, bilateral infCI and left supCI (all p < 0.005). Similarily, MCI_converters (but not MCI_stable) had lower levels of RM₀ ^B in left ILF, bilateral infCi and right UF when compared to HC. A difference between MCI_stable and AD patients emerged in right infCi. When directly comparing MCI_converters to MCI_stable, the former group reported lower RM₀ ^B in right UF (Fig. 3). No significant results emerged when considering F values.

Estimated cut-off DTI and qMT values for prediction to AD

Fig.3

Group-averaged Graph RM₀ ^B values estimated in the tracts of interest. RM₀ ^B in UF differentiated MCIconverters from MCIstable already at baseline. AD, Alzheimer’s disease; HC, healthy controls; MCIconv, MCI that converted to AD at follow-up; MCIstab, MCI that remained stable (did not convert to AD) at follow-up; RM₀ ^B, forward exchange rate; sup, superior; inf, inferior; UF, uncinated fasciculus.

To estimate the optimal DTI and qMT cut-off values for predicting MCI to AD conversion, we chose the point, based on the ROC analysis, which had the greatest combined specificity and sensitivity. As shown in Fig. 4, the AUC statistic (95% CI) for the optimal RM₀ ^B value in the right UF tract was 0.743(0.572–0.913), whereas the cut-off values for the conversion was 0.683. In the case of SupCi left (AUC = 0.735; 0.564–0.906) and SupCi right (AUC = 0.741; 0.571–0.911), the optimal cut-off values for FA were 0.355 and 0.345, respectively (Fig. 4).

Fig.4

Receiver-operating characteristic (ROC) curve and area under the curve (AUC) showing the sensitivity and specificity of RM₀ ^B values in uncinated fasciculus (UF) and FA values in bilateral superior cingulum bundle in predicting conversion from MCI to AD.

Correlation between DTI and qMT metrics

Table 3 summarizes the correlations between DTI and qMT matrices in the WM tracts of interest. MD and RM₀ ^B values were negatively correlated in almost all tracts (except right ILF and UF left). MD correlated negatively also with F values in supCI left and UF left. Similarly, in all tracts (but left ILF), we observed a negative correlation between RDif and RM₀ ^B parameter. A negative correlation was evident between RDif and F in left UF. As regards ADif, a negative correlation was evident with RM₀ ^B values of the bilateral inferior cingulum, whereas no correlation was evident with F values. Finally, a positive correlation was observed between FA and RM₀ ^B values in bilateral ILF and F values in left UF. It is interesting to observe that most of these correlations, albeit significant, were modest (R < 0.5).

DISCUSSION

This study sought to identify alterations in WM tracts, measurable in the brain of MCI patients up to a year before they converted to AD. In order to capture the full extent of WM changes in aMCI and AD, we considered both DTI and qMT indices in the major tracts involved in AD: inferior longitudinal fasciculus, inferior and superior cingulum and uncinate fasciculus.

A number of major results emerged. First, we identified 2 potential biomarkes of conversion to AD: FA in supCi and RM₀ ^B in UF, which can differentiate MCI_converters from MCI_stable already at baseline.

Both UF and the cingulum are known to be involved in AD pathology [39 , 65]. The UF is one of principal pathways of connection between temporal and frontal lobes, and therefore damage to this tract might explain some aspects of memory impairment, such as those related with encoding and retrieval of new material [66, 67]. Previous studies have also postulated a role for the UF as predictor of the conversion from aMCI to AD [68], and this tract has been linked to memory and emotional recognition alterations in aMCI [69]. Interestingly, a previous study from our group showed that damage to the UF, by means of reduced FA, was measurable in patients with AD, but not in those with MCI [39].

The cingulum bundle is the major median associative WM fasciculus reciprocally connecting various frontal, temporal, and parietal regions. It is part of both, the executive control and default mode networks, which are responsible for executive function, memory and other cognitive functions [70]. The damage in the cingulum has been found to correlate with measures of cognitive impairment [71], and we have previously shown that it correlates with atrophy of the nearby GM structures [9]. It is therefore unsurprising that measures of microstructural damage to this tract can be sensitive to the conversion to AD.

In line with these findings, we show that, at baseline, MCI_converters already show some microstructural differences compared to HC, which resemble those exhibited by AD patients. This result matches what has already been shown for GM atrophy [5], i.e., that MCI who are about to convert to AD will show a degree of tissue damage comparable to AD patients prior to conversion, while such damage is undetectable in those MCI patients who remain stable. Interestingly, these WM abnormalities can be detected by the DTI indices MD, RDif and ADif, and by the qMT parameters RM₀ ^B, but not by FA or F. This probably indicates a pathological substrate of inflammatory nature, with no frank demyelination or tissue loss.

MCI_stable only show signs of mild microstructural damage when compared to HC, evidenced by an increase in ADif in infCi and UF. The interpretation of these findings is highly speculative, as we cannot be sure whether this group of patient will ever develop a form of dementia. However, if we were to hypothesize that they represent a group of very early AD patients, the involvement of these tracts fits with the frequently reported (and above discussed) evidence of microstructural damage to the cingulum and the UF in AD patients [9, 39] and MCI [68, 72]. Anyhow, the lack of a ‘control’ tract in our analyses (i.e., a tract of no-interest) does not allow us to conclude about the ‘focal’ involvement of these tracts in early AD pathology. Indeed, the possibility of a more generic disruption of the WM, that goes beyond the areas primarily damaged by AD degeneration, has been also recently suggested [73], supporting the view of AD as a ‘disconnection syndrome’ [7].

The main novelty of this paper is in the combination of qMT and DTI parameters, and in the multi-parametric evaluation of the major WM tracts. The potential role of qMT in the evaluation of patients with AD was highlighted by three previous papers [23 , 30]. Here we extended the analysis to the population of MCI individuals. Our data confirm the sensitivity of the parameter RM₀ ^B in detecting pathophysiological changes in AD [30]. In contrast to our initial hypothesis, F, which has been linked to myelin damage, was not sensitive to alterations in WM tracts, while the DTI parameter RDif, believed to also reflect myelin damage, showed some abnormalities in MCI_converters. The lack of correlation between the two parameters, although unexpected, has been already described in other studies [32, 74]. The reason for such a discrepancy might be explained by the underlying physical principles between the two imaging techniques [33, 74], where DTI indices reflect mainly the directionality of WM fibers whereas qMT parameters reflects tissue composition rather than spatial organization. An overall increase in diffusivity (which can be associated with tissue loss and inflammation) would also result in an increase of RDif, with no associated change in F.

In support of this speculation, values of RDif mostly correlated with values of RM₀ ^B in almost all tracts (but ILF right), suggesting that the two parameters might reflect similar, or highly linked, structural contributions to WM integrity in this sample. An intriguing speculation on the clinical relevance of the parameter RM₀ ^B in AD has been put forward in a previous study from our group [30], suggesting that this parameter might reflect some metabolic information [75] and supporting the hypothesis that mitochondrial dysfunction might play a relevant role in the pathophysiology of AD [76]. Whereas this still remains a purely speculative explanation, it is worth noting that mitochondrial dysfunction contributes to WM demyelination in multiple sclerosis [77, 78], suggesting a similar mechanism in the WM damage in AD and providing a possible explanation of the correlation between RDif and RM₀ ^B indices in our study. Of note, changes in RM₀ ^B of the insular cortex were recently shown in experimentally induced systemic inflammation, and were interpreted as localized increase in metabolically-active macromolecules [36]. This might also explain why in our study different indices of WM microstructure are more sensitive to the damage in different tracts. While part of it might be due to the different fiber architecture in the various tracts [79], other might be explained by the specific pathological process occurring in specific tracts. For example, UF carries cholinergic fibers from the basal nucleus of Meynert innervating frontal and temporal regions [80]. The integrity of cholinergic system is highly vulnerable to metabolic state, suggesting again that RM₀ ^B (the most sensitive marker in spotting WM damage in UF when comparing MCI_converters and MCI_stable) might be eloquent of underlying metabolic processes. In fact, the hypothesis of a cholinergic deficit in AD and its pathophysiological consequences have a long history. It is known that an impairment of cholinergic transmission one of the features of neurodegeneration which occurs early in AD, and the loss of cholinergic neurons and projections is well described in the AD pathology [81]. Albeit these hypotheses are intriguing, further interpretations are beyond the scope of this study and future research is needed to investigate the clinical relevance of qMT parameters extracted from WM tracts.

In conclusion, our study reveals that qMT and DTI indices extracted from tractography are efficient in spotting the WM microstructural damage and are valid biomarkers of the conversion from MCI to AD. Combined with neuropsychological evaluation, the complementary information from these markers could be applied in future for the development of multivariate tools for the early identification of patients who are likely to benefit from therapeutic intervention.

Footnotes

ACKNOWLEDGMENTS

This project has been supported by grants of the Italian Ministry of Health and of the Italian Ministry of University and Research.

Authors’ disclosures available online ().

References

Mitchell

, Shiri-Feshki

(2009) Rate of progression of mild cognitive impairment to dementia- meta-analysis of 41 robust inception cohort studies. Acta Psychiatr Scand 119, 252–265.

Perri

, Serra

, Carlesimo

, Caltagirone

(2007) Amnestic mild cognitive impairment: Difference of memory profile in subjects who converted or did not convert to Alzheimer’s disease. Neuropsychology 21, 549–558.

Mufson

, Binder

, Counts

, DeKosky

, deTolledo-Morrell

, Ginsberg

, Scheff

, (2012) Mild cognitive impairment: Pathology and mechanisms. Acta Neuropathol 123, 13–30.

Petersen

, Doody

, Kurz

, Mohs

, Morris

, Rabins

, Ritchie

, Rossor

, Thal

, Winblad

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

Bozzali

, Filippi

, Magnani

, Cercignani

, Franceschi

, Schiatti

, Castiglioni

, Mossini

, Falautano

, Scotti

, Comi

, Falini

(2006) The contribution of voxel-based morphometry in staging patients with mild cognitive impairment. Neurology 67, 453–460.

Devanand

, Liu

, Tabert

, Pradhaban

, Cuasay

, Bell

, de Leon

, Doty

, Stern

, Pelton

(2008) Combining early markers strongly predicts conversion from mild cognitive impairment to Alzheimer’s disease. Biol Psychiatry 64, 871–879.

Bozzali

, Padovani

, Caltagirone

, Borroni

(2011) Regional grey matter loss and brain disconnection across Alzheimer disease evolution. Curr Med Chem 18, 2452–2458.

Bozzali

, Cherubini

(2007) Diffusion tensor MRI to investigate dementias: A brief review. Magn Reson Imaging 25, 969–977.

Bozzali

, Giulietti

, Basile

, Serra

, Spanó

, Perri

, Giubilei

, Marra

, Caltagirone

, Cercignani

(2012) Damage to the cingulum contributes to Alzheimer’s disease pathophysiology by deafferentation mechanism. Hum Brain Mapp 33, 1295–1308.

10.

Bartzokis (2011) Alzheimer’s disease as homeostatic responses to age-related myelin breakdown. Neurobiol Aging 32, 1341–1371.

11.

Gold

, Johnson

, Powell

, Smith

(2012) White matter integrity and vulnerability to Alzheimer’s disease: Preliminary findings and future directions. Biochim Biophys Acta 1822, 416–422.

12.

Pierpaoli

, Barnett

, Pajevic

, Chen

, Penix

, Virta

, Basser

(2001) Water diffusion changes in Wallerian degeneration and their dependence on WM architecture. Neuroimage 13, 1174–1185.

13.

Le Bihan

, Mangin

, Poupon

, Clark

, Pappata

, Molko

, Chabriat

(2001) Diffusion tensor imaging: Concepts and applications. J Magn Reson Imaging 13, 534–546.

14.

Sundgren

, Dong

, Gómez-Hassan

, Mukherji

, Maly

, Welsh

(2004)Neuroradiology 46, 339–350 Diffusion tensor imaging of the bra, Review of clinical alications.

15.

Pierpaoli

and Basser

(1996) Toward a quantitative assessment of diffusion anisotropy. Magn Reson Med 36, 893–906.

16.

Beaulieu

(2002) The basis of anisotropic water diffusion in the nervous system–a technical review. NMR Biomed 15, 435–455.

17.

Song

, Sun

, Ramsbottom

, Chang

, Russell

, Cross

(2002) Dysmyelination revealed through MRI as increased radial (but unchanged axial) diffusion of water. Neuroimage 17, 1429–1436.

18.

Song

, Yoshino

, Le

, Lin

, Sun

, Cross

(2005) Armstrong R.C. Demyelination increases radial diffusivity in corpus callosum of mouse brain. Neuroimage 26, 132–140.

19.

Sun

, Liang

, Trinkaus

, Cross

, Armstrong

, Song

(2006) Noninvasive detection of cuprizone induced axonal damage and demyelination in the mouse corpus callosum. Magn Reson Med 55, 302–308.

20.

Concha

, Gross

, Wheatley

, Beaulieu

(2006) Diffusion tensor imaging of time-dependent axonal and myelin degradation after corpus callosotomy in epilepsy patients. Neuroimage 32, 1090–1099.

21.

Dousset

, Brochet

, Vital

, Gross

, Benazzouz

, Boullerne

, Bidabe

, Gin

, Caille

(1995) Lysolecithin-induced demyelination in primates: Preliminary in vivo study with MR and magnetization transfer. AJNR Am J Neuroradiol 16, 225–231.

22.

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.

23.

Ridha

, Tozer

, Symms

, Stockton

, Lewis

, Siddique

, MacManus

, Rossor

, Fox

, Tofts

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

24.

Fornari

, Maeder

, Meuli

, Ghika

, Knyazeva

(2010) Demyelination of superficial white matter in early Alzheimer’s disease: A magnetization transfer imaging study. Neurobiol Aging 33, 428.

25.

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.

26.

Lee

, Kim

, Park

, Ko

, Song

, Cho

(2004) Age-related changes in conventional and magnetization transfer MR imaging in elderly people: Comparison with neurocognitive performance. Korean J Radiol 5, 96–101.

27.

Henkelman

, Stanisz

, Graham

(2001) Magnetization transfer in MRI: A review. NMR Biomed 14, 57–64.

28.

Henkelman

, Huang

, Xiang

, Stanisz

, Swanson

, Bronskill

(1993) Quantitative interpretation of magnetization transfer. Magn Reson Med 29, 759–766.

29.

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 MRimaging. Neuroimage 48, 657–667.

30.

Giulietti

, Bozzali

, Figura

, Spanó

, 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.

31.

Schmierer

, Tozer

, Scaravilli

, Altmann

, Barker

, Tofts

, Miller

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

32.

, Sun

, Liang

, Song

, Gochberg

(2009) Quantitative magnetization transfer measured pool-size ratio reflects optic nerve myelin content in ex vivo mice. Magn Reson Med 61, 364–371.

33.

, Sun

, Liang

, Song

, Gochberg

(2009) The MT pool size ratio and the DTI radial diffusivity may reflect the myelination in shiverer and control mice. NMR Biomed 22, 480–487.

34.

Turati

, Moscatelli

, Mastropietro

, Dowell

, Zucca

, Erbetta

, Cordiglieri

, Brenna

, Bianchi

, Mantegazza

, Cercignani

, Baggi

, Minati

(2015) In vivo quantitative magnetization transfer imaging correlates with histology during de- and remyelination in cuprizone-treated mice. NMR Biomed 28, 327–337.

35.

Janve

, Zu

, Yao

, Li

, Zhang

, Wilson

, Ou

, Does

, Subramaniam

, Gochberg

(2013) The radial diffusivity and magnetization transfer pool size ratio are sensitive markers for demyelination in a rat model of type III multiple sclerosis (MS) lesions. Neuroimage 74, 298–305.

36.

Harrison

, Cooper

, Dowell

, Keramida

, Voon

, Critchley

, Cercignani

(2014) Quantitative magnetization transfer imaging as a biomarker for effects of systemic inflammation on the brain. Biol Psychiatry 78, 49–57.

37.

Agosta

, Pievani

, Sala

, Geroldi

, Galluzzi

, Frisoni

, Filippi

(2011) White matter damage in Alzheimer disease and its relationship to gray matter atrophy. Radiology 258, 853–863.

38.

Serra

, Cercignani

, Basile

, Spanó

, Perri

, Fadda

, Marra

, Giubilei

, Caltagirone

, Bozzali

(2012) White matter damage along the uncinate fasciculus contributes to cognitive decline in AD and DLB. Curr Alzheimer Res 9, 326–333.

39.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

Jr , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

40.

Petersen

, Smith

, Waring

, Ivnik

, Tangalos

, Kokmen

(1999) Mild cognitive impairment: Clinical characterization and outcome. Arch Neurol 56, 303–308.

41.

American Psychiatric Association (2000) DSM-IV-TR. Diagnostic and Statistical Manual of Mental Disorder, Fourth Edition, Text Revision, Milano.

42.

Folstein

, Folstein

, McHugh

(1975) “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

43.

Measso

, Cavartezan

, Zappal‘a

, Lebowitz

, Crook

, Pirozzolo

, Amaducci

, Massari

, Grigoletto

(1993) The Mini Mental State Examination: Normative study of a random sample of Italian population. Dev Neuropsychol 9, 77–85.

44.

Magni

, Binetti

, Padovani

, Cappa

, Bianchetti

, Trabucchi

(1996) The Mini-Mental State Examination in Alzheimer’s disease and multi-infarct dementia. Int Psychogeriatr 8, 127–134.

45.

Hughes

, Berg

, Danziger

, Coben

, Martin

(1982) A new clinical scale for the staging of dementia. Br J Psychiatry 40, 566–572.

46.

Carlesimo

, Caltagirone

, Gainotti

(1996) The mental Deterioration Battery: Normative data, diagnostic reliability and qualitative analyses of cognitive impairment. The group of the Standardization of the Mental Deterioration Battery. Eur Neurol 36, 378–384.

47.

Carlesimo

, Buccione

, Fadda

, Graceffa

, Mauri

, Lo russo

, Bevilacqua

, Caltagirone

(2002) Standardizzazione di due test di memoria per uso clinico: Breve Racconto e Figura di Rey. Nuova Riv Neurol 12, 1–13.

48.

Monaco

, Costa

, Caltagirone

, Carlesimo

(2013) Forward and backward span for verbal and visuo-spatial data: Standardization and normative data from an Italian adult population. Neurol Sci 34, 749–754.

49.

Jones

, Horsfield

, Simmons

(1999) Optimal strategies for measuring diffusion in anisotropic systems by magnetic resonance imaging. Magn Reson Med 42, 515–525.

50.

Cercignani

, Basile

, Spanò

, Comanducci

, Fasano

, Caltagirone

, Nocentini

, Bozzali

(2009) Investigation of quantitative magnetisation transfer parameters of lesions and normal appearing white matter in multiple sclerosis. NMR Biomed 22, 646–653.

51.

Dowell

, Tofts

(2007) Fast, accurate, and precise mapping of the RF field in vivo using the 180 degrees signal null. Magn Reson Med 58, 622–630.

52.

UCL Camino Diffusion MRI Toolkit.http://camino.cs.ucl.ac.uk/ .

53.

Jenkinson

, Smith

(2001) A global optimisation method for robust affine registration of brain images. Med Image Anal 5, 143–156.

54.

Jenkinson

, Bannister

, Brady

, Smith

(2002) Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage 17, 825–841.

55.

Ernst

, Anderson

(1966) Application of Fourier transform spectroscopy to magnetic resonance. Rev Sci Instrum 37, 93–102.

56.

Dowell

, Cooper

, Tibble

, Voon

, Critchley

, Cercignani

, Harrison

(2016) Acute changes in striatal microstructure predict the development of interferon-alpha induced fatigue. Biol Psychiatry 79, 320–328.

57.

Venkatesan

, Lin

, Haacke

(1998) Accurate determination of spin-density and T1 in the presence of RF-field inhomogeneities and flip-angle miscalibration. Magn Reson Med 40, 592–602.

58.

Cercignani

, Alexander

(2006) Optimal acquisition schemes for in vivo quantitative magnetization transfer MRI. Magn Reson Med 56, 803–810.

59.

Cercignani

, Symms

, Schmierer

, Boulby

, Tozer

, Ron

, Tofts

, Barker

(2005) Three-dimensional quantitative magnetisation transfer imaging of the human brain. Neuroimage 27, 436–441.

60.

Press

, Teukolsky

, Vettering

, Flannery

(1992) Numerical recipes in C. Cambridge University Press, Cambridge.

61.

Morrison

, Stanisz

, Henkelman

(1995) Modeling magnetization transfer for biological-like systems using a semi-solid pool with a super-Lorentzian lineshape and dipolar reservoir. J Magn Reson B 108, 103–113.

62.

Tozer

, Ramani

, Barker

, Davies

, Miller

, Tofts

(2003) Quantitative magnetization transfer mapping of bound protons in multiple sclerosis. Magn Reson Med 50, 83–91.

63.

Narayanan

, Francis

, Sled

, Santos

, Antel

, Levesque

, Brass

, Lapierre

, Sappey-Marinier

, Pike

, Arnold

(2006) Axonal injury in the cerebral normal-appearing white matter of patients with multiple sclerosis is related to concurrent demyelination in lesions but not to concurrent demyelination in normal-appearing white matter. Neuroimage 29, 637–642.

64.

Bosch

, Arenaza-Urquijo

, Rami

, Sala-Llonch

, Junqué

, Solé-Padullés

, Peña-Gómez

, Bargalló

, Molinuevo

, Bartrés-Faz

(2012) Multiple DTI index analysis in normal aging, amnestic MCI and AD. Relationship with neuropsychological performance. Neurobiol Aging 33, 61–74.

65.

Stebbins

, Murphy

(2009) Diffusion tensor imaging in Alzheimer’s disease and mild cognitive impairment. Behav Neurol 21, 39–49.

66.

Horel

(1978) The neuroanatomy of amnesia. A critique of the hippocampal memory hypothesis. Brain 101, 403–445.

67.

Levine

, Black

, Cabeza

, Sinden

, McIntosh

, Toth

, Tulving

, Stuss

(1998) Episodic memory and the self in a case of isolated retrograde amnesia. Brain 121, 1951–1973.

68.

Hiyoshi-Taniguchi

, Oishi

, Namiki

, Miyata

, Murai

, Cichocki

, Fukuyama

(2015) The uncinate fasciculus as a predictor of conversion from aMCI to Alzheimer disease. J Neuroimaging 25, 748–753.

69.

Fujie

, Namiki

, Nishi

, Yamada

, Miyata

, Sakata

, Sawamoto

, Fukuyama

, Hayashi

, Murai

(2008) The role of the uncinate fasciculus in memory and emotional recognition in amnestic mild cognitive impairment. Dement Geriatr Cogn Disord 26, 432–439.

70.

van den Heuvel

, Mandl

, Luigjes

, Hulshoff Pol

(2008) Microstructural organization of the cingulum tract and the level of default mode functional connectivity. J Neurosci 28, 10844–10851.

71.

Xie

, Xiao

, Wang

, Wu

, Gong

, Jiang

(2005) Evaluation of bilateral cingulum with tractography in patients with Alzheimer’s disease. Neuroreport 16, 1275–1278.

72.

Ito

, Sasaki

, Takahashi

, Uwano

, Yamashita

, Higuchi

, Goodwin

, Harada

, Kudo

, Terayama

(2015) Detection of early changes in the parahippocampal and posterior cingulum bundles during mild cognitive impairment by using high-resolution multi-parametric diffusion tensor imaging. Psychiatry Res 231, 346–352.

73.

Palesi

, Castellazzi

, Casiraghi

, Sinforiani

, Vitali

, Gandini Wheeler-Kingshott

, D’Angelo

(2016) Exploring patterns of alteration in Alzheimer’s disease brain networks: A combined structural and functional connectomics analysis. Front Neurosci 10, 380.

74.

Underhill

, Yuan

, Yarnykh

(2009) Direct quantitative comparison between cross-relaxation imaging and diffusion tensor imaging of the human brain at 3.0 T. Neuroimage 47, 1568–1578.

75.

Louie

, Gochberg

, Does

, Damon

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

76.

Moreira

, Carvalho

, Zhu

, Smith

, Perry

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

77.

Witte

, Mahad

, Lassmann

, van Horssen

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

78.

Carvalho

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

79.

De Santis

, Drakesmith

, Bells

, Assaf

, Jones

(2014) Why diffusion tensor MRI does well only some of the time: Variance and covariance of white matter tissue microstructure attributes in the living human brain. Neuroimage 89, 35–44.

80.

Selden

, Gitelman

, Salamon-Murayama

, Parrish

, Mesulam

(1998) Trajectories of cholinergic pathways within cerebral hemispheres of the human brain. Brain 121, 2249–2257.

81.

Francis

, Palmer

, Snape

, Wilcock

(1999) The cholinergic hypothesis of Alzheimer’s disease: A review of progress. J Neurol Neurosurg Psychiatry 66, 137–147.