Elsevier

NeuroImage

Volume 48, Issue 4, December 2009, Pages 657-667
NeuroImage

Multi-parametric classification of Alzheimer's disease and mild cognitive impairment: The impact of quantitative magnetization transfer MR imaging

https://doi.org/10.1016/j.neuroimage.2009.07.005Get rights and content

Abstract

Multi-parametric and quantitative magnetic resonance imaging (MRI) techniques have come into the focus of interest, both as a research and diagnostic modality for the evaluation of patients suffering from mild cognitive decline and overt dementia. In this study we address the question, if disease related quantitative magnetization transfer effects (qMT) within the intra- and extracellular matrices of the hippocampus may aid in the differentiation between clinically diagnosed patients with Alzheimer disease (AD), patients with mild cognitive impairment (MCI) and healthy controls. We evaluated 22 patients with AD (n = 12) and MCI (n = 10) and 22 healthy elderly (n = 12) and younger (n = 10) controls with multi-parametric MRI. Neuropsychological testing was performed in patients and elderly controls (n = 34). In order to quantify the qMT effects, the absorption spectrum was sampled at relevant off-resonance frequencies. The qMT-parameters were calculated according to a two-pool spin-bath model including the T1- and T2 relaxation parameters of the free pool, determined in separate experiments. Histograms (fixed bin-size) of the normalized qMT-parameter values (z-scores) within the anterior and posterior hippocampus (hippocampal head and body) were subjected to a fuzzy-c-means classification algorithm with downstreamed PCA projection. The within-cluster sums of point-to-centroid distances were used to examine the effects of qMT- and diffusion anisotropy parameters on the discrimination of healthy volunteers, patients with Alzheimer and MCIs. The qMT-parameters T2r (T2 of the restricted pool) and F (fractional pool size) differentiated between the three groups (control, MCI and AD) in the anterior hippocampus. In our cohort, the MT ratio, as proposed in previous reports, did not differentiate between MCI and AD or healthy controls and MCI, but between healthy controls and AD.

Introduction

In Alzheimer's disease (AD), neurodegeneration in the mediotemporal lobe (MTL) often begins many years before the clinical onset of dementia. Neuropathological research has demonstrated a topographically characteristic course of infestation of MTL structures, that strongly correlates with the progression of disease. Thus, early and preclinical diagnosis and staging of AD is mandatory for causal pharmacotherapeutic interventions. Multimodal MRI offers a noninvasive tool for repeatable testing and monitoring of patients with cognitive decline who may benefit from early therapy with N-methyl-d-aspartate (NMDA) antagonists and acetylcholinesterase (AChE) inhibitors.

The formation of plaques is known to be a crucial event in AD pathogenesis and starts several years before the onset of clinical symptoms. At that point, plaque deposition has become extensive in most instances and involves brain regions essential for normal cognition (Cataldo et al., 2000, Morris et al., 1996, Haroutunian et al., 1998). Previous studies (Blessed et al., 1968, Caramelli et al., 1998, Naslund et al., 2000, Bussiere et al., 2002) have shown a correlation between plaque load and severity of dementia within the hippocampus, the associated entorhinal cortex and the perirhinal (transentorhinal) cortex. These regions are especially vulnerable to an early deposition of amyloid plaques (Arnold et al., 1991, Hyman et al., 1984, Hyman et al., 1990, Braak and Braak, 1991). Neurons of the hippocampus are particularly vulnerable to amyloid deposition during the early course of AD (Braak et al., 1993, Gomez-Isla et al., 1996, West et al., 1994). Damage to the hippocampus and to the cortical GM of AD patients has been demonstrated by magnetization transfer imaging, suggesting that the MT technology is sensitive to grey matter abnormalities (Bozzali et al., 2001, Hanyu et al., 2001). In AD, histopathologic findings in the hippocampus resemble a loss of pyramidal cells accompanied by an increase in the number of astrocytes, microglia, and oligodendrocytes, as well as by an accumulation of plaques and neurofibrillary tangles (Mirra et al., 1993). In addition, demyelination and axonal loss may be other additional causative factors that influence MT-parameters (Brochet and Dousset, 1999, van Waesberghe and Barkhof, 1999), since degeneration of intrahippocampal projection neurons frequently occurs in AD (Mizutani & Kasahara, 1997). Cellular changes, synaptic loss, and neuronal degeneration (where MT has been utilized as a surrogate marker) are likely to precede gross regional atrophy (Hyman et al., 1984), thus supporting the hypothesis that the underlying biologic changes in brain tissue may be detected in the absence of obvious volumetric changes in subjects with MCI (Kohler et al., 1998).

The τ-protein is a molecule that normally stabilizes the cytoskeleton of human neurons. This τ-protein is being modified within the nerve cells of our brain during aging — the molecule is loaded with phosphate groups. In patients with mild cognitive impairment (MCI) an enhanced concentration of a special form of such modified τ-231-proteins has been isolated in the cerebrospinal fluid (Watanabe et al., 1993). Previous studies demonstrated, that above a certain threshold concentration MCI patients turn out to show the Alzheimer symptoms with a probability of up to 80% (Parnetti et al., 2001). It is not clear however, if all patients with MCI will proceed to AD — to fortify drug administration during subclinical stages of the disease, reliable biochemical testing and more sophisticated imaging techniques are mandatory.

Nuclear magnetic resonance magnetization transfer was introduced as such a technique to probe the surface-to-volume ratio and morphology of materials with characteristic structure sizes of 1–100 nm (Valiullina et al., 2003) — exactly the characteristic dimensions of the AD plaques in humans to be detected among other effects (Benzinger et al., 2000). NMR facilitates the study at the level of individual amino acid residues during folding/unfolding of proteins. Heteronuclear correlation experiments have been shown to be very sensitive because of the high magnetization transfer between directly bond nuclei (Schulman et al., 1997).

Thus, we hypothesized that, in the presence of extracellular amyloid plaques and intracellular neurofibrillary tangles, demyelination and axonal loss will lead to MT induced signal changes, that are detectable with quantitative magnetization transfer (qMT) MR imaging. The post-processing algorithm and the calculation of the qMT-parameters, applied in this study are based on a sophisticated spin-bath model (Sled and Pike, 2000a, Sled and Pike, 2000b, Sled and Pike, 2001, Sled and Pike, 2004, Kiefer et al., 2004). This approach is in particular different from the approach of using the magnetization transfer ratio, which is often mentioned in the literature — especially the precise mechanism for the reduction of the MTR in the hippocampus of AD patients is not yet clear. Moreover the MTR is a phenomenological measure that has been shown to depend on the amount of magnetization transfer and also on the direct saturation of free water by the rf pulse which has been shown to be a causative for misleading results (Stanisz et al., 2002).

Magnetization transfer imaging has been repeatedly applied as an overall measure to various neurodegenerative and neuroinflammatory disorders, e.g. Alzheimer disease (Ridha et al. 2007), Parkinson disease (Anik et al., 2007, Tambasco et al., 2003), multiple Sclerosis and progressive supranuclear palsy (PSP) (Hanyu et al. 2001). Saturated protons enter the free pool of protons or transfer their magnetization to free water protons. The net effect is a decrease in the MR visible signal in areas of macromolecules effected by magnetization transfer. Thus, it may be inferred, that the MT ratio (MTR) appears to be a non-specific finding with regard to early neurodegeneration. Quantitative MT imaging, in contrast, allows a more comprehensive analysis by investigating more fundamental parameters of qMT. In a first study, Ridha et al. (2007) have applied qMT to a series of 14 patients and 14 healthy controls. They found a reduction of the relaxation times in the free proton pool that aided in their differentiation of healthy controls and AD. In our study, in contrast, qMT is applied not only in AD versus controls, but in a MCI population. Moreover, Ridha has applied volumetry to the whole hippocampus.

In our study, we have chosen only the anterior HC, since degeneration of intrahippocampal and entorhinal projection neurons frequently occurs in the CA 1 subregion of the hippocampal head. We investigated, to which degree the physical processes of diffusion, relaxometry and magnetization transfer may act as a classifier between AD, MCI, younger and elderly controls. As the essential reference the results of the neuropsychological examination and the routine MRI dementia protocol (Scheltens et al., 1995) were used.

The protons in biological systems can be described as existing in two pools: the free and the bound protons. The most sophisticated model of Balaban and Ceckler (1992) characterizes the free pool as consisting of mobile bulk protons and so-called ‘hydration water’, protons connected to the surface of macromolecules via dipole-dipole interactions through space (cross relaxation). The free pool has a narrow spectral line (10–100 Hz), whereas the bound pool is characterized by a broad spectral line of 10–50 kHz, but each centered at the same Larmor frequency. This fact can be used to pre-saturate the bound protons by using frequency-selective rf pulses irradiated at a frequency offset Δf with respect to the central 1H Larmor frequency while affecting the ‘free pool’ due to magnetization transfer. In this work, exclusively Gaussian modulated pre-saturation pulses were used. The determination of fractional pool sizes and T2-times of restricted protons necessitates an elaborate model to describe the dynamics of the magnetization transfer mechanism. In this work a quantitative imaging technique is used that yields all of the observable properties of the binary spin-bath model for MT. Adapted from a model of the steady-state behavior of the magnetization during a pulsed MT-weighted imaging sequence, as well as suitable methods in MRI relaxometry (1), this approach yields parametric images of the fractional size of the restricted pool, the T2 of the restricted pool, and the relaxation times in the free pool. It has been shown (Sled and Pike, 2000a, Sled and Pike, 2000b, Sled and Pike, 2001, Sled and Pike, 2004), that the restricted pool can be adequately described by the Bloch equations (Bloch et al., 1946). A simplified graphical representation of a binary spin-bath model is shown in Fig. 1a: the longitudinal magnetization M0 of the different pools of free (indices F, f) and restricted protons ((indices R, r) is presented by a basin filled with water. The MT exchange and relaxation phenomena are presented by channels connecting the basins with each other (kf (F→R), kr (R→F)) and with the lattice (R1,f, R1,r). If the restricted pool is saturated by an rf pulse, a new equilibrium between the pools F and R establishes and the longitudinal magnetization of the free pool is reduced.

Section snippets

MRI acquisition and sequence parameters

All experiments were performed on a 1.5 Tesla MR scanner (Siemens Magnetom Sonata Vision, Siemens Erlangen, Germany) at the Institute of Diagnostic and Interventional Neuroradiology, Inselspital, University of Bern.

A diagnostic routine MRI was performed to exclude other treatable causes of cognitive decline (i.e. vascular dementia, normal pressure hydrocephalus or brain tumor).

Structural imaging was obtained using a T1-weighted, sagittal oriented 3D-MPRAGE sequence (TR/TE/TI 2000/3.42/1100 ms,

Results

For the study, the DTI and qMT-parameters and their neuropsychological performance have been correlated to AD or MCI in 22 patients and 12 controls. The subgroups of the patients with AD and MCI did not differ significantly in terms of mean age (p = 0.14), whereas the healthy elderly controls were younger (p < 0.01) than the MCI and AD patients (paired t-test). The demographic data of the patients and controls are listed in Table 1a. An ANOVA one-way analysis of variance was performed for testing

Discussion

In this study, we have addressed the question, if qMT and DT-imaging can be used to differentiate between clinically classified subgroups of patients with established AD from MCI and healthy controls. Here, we have shown that the qMT-parameters T2r, F, MTR, kf, kr in the anterior hippocampus, i.e. the hippocampal head, are the crucial factors to differentiate between these different cohorts. Moreover the fact that most of the MR parameter values of the MCI group, presented in Table 4, rather

Conclusion

The qMT-parameters T2r and F significantly differentiated healthy controls, MCI and AD in the HH subregion. This is in contrast to previous studies using the MT ratio (MTR) parameter as a discriminator between AD and healthy controls, but did not succeed to differentiate between MCI and AD or healthy controls and MCI patients. In order to avoid mixing effects of MT and relaxometry we propose a model based qMT-parameter analysis to characterize disease related changes of the hippocampal matrices

Acknowledgment

This project was granted by the Kamillo-Eisner foundation — Grant no. EK 145/05.

References (76)

  • KohlerS. et al.

    Memory impairments associated with hippocampal versus parahippocampal-gyrus atrophy: an MR volumetry study in Alzheimer's disease

    Neuropsychologia

    (1998)
  • LeskiJ.

    Towards a robust fuzzy clustering

    Fuzzy Sets Syst.

    (2003)
  • MacFallJ.R. et al.

    Magn. Reson. Imag.

    (1987)
  • MizutaniT. et al.

    Hippocampal atrophy secondary to entorhinal cortical degeneration in Alzheimer-type dementia

    Neurosci. Lett.

    (1997)
  • SledJ.G. et al.

    Quantitative interpretation of magnetization transfer in spoiled gradient echo MRI sequences

    J. Magn. Reson.

    (2000)
  • WatanabeA. et al.

    J. Biol. Chem.

    (1993)
  • WestM.J. et al.

    Differences in the pattern of hippocampal neuronal loss in normal ageing and Alzheimer's disease

    Lancet

    (1994)
  • AlexanderD.C. et al.

    Spatial transformations of diffusion tensor magnetic resonance images

    IEEE Trans. Med. Imag.

    (2001)
  • AndersonT.W.

    Asymptotic theory for principal component analysis

    Ann. Math. Stat.

    (1963)
  • ArnoldS.E. et al.

    The topographical and neuroanatomical distribution of neurofibrillary tangles and neuritic plaques in the cerebral cortex of patients with Alzheimer's disease

    Cereb. Cortex

    (1991)
  • BalabanR.S. et al.

    Magnetization transfer contrast in magnetic resonance imaging

    Magn. Reson. Q.

    (1992)
  • BardY.

    Nonlinear Parameter Estimation

    (1974)
  • BelkinM. et al.

    Semi-supervised learning on Riemannian manifolds

    Mach. Learn.

    (2004)
  • BenzingerT.L.S. et al.

    Two-dimensional structure of β-amyloid(10–35) fibrils

    Biochemistry

    (2000)
  • BezdekJ.

    A convergence theorem for the fuzzy data clustering algorithms

    IEEE Trans. Pattern Anal. Mach. Intell. PAMI-2

    (1980)
  • BezdekJ. et al.

    Fuzzy Models and Algorithms for Pattern Recognition and Image Processing

    (1999)
  • BlessedG. et al.

    The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects

    Br. J. Psychiatry

    (1968)
  • BlochF. et al.

    Phys. Rev.

    (1946)
  • BozzaliM. et al.

    Quantification of tissue damage in AD using diffusion tensor and magnetization transfer MRI

    Neurology

    (2001)
  • BraakH. et al.

    Neuropathological stageing of Alzheimer-related changes

    Acta Neuropathol. (Berl)

    (1991)
  • BraakH. et al.

    Staging of Alzheimer-related cortical destruction

    Eur. Neurol.

    (1993)
  • BrochetB. et al.

    Pathological correlates of magnetization transfer imaging abnormalities in animal models and humans with multiple sclerosis

    Neurology

    (1999)
  • CaramelliP. et al.

    Structural correlates of cognitive deficits in a selected group of patients with Alzheimer's disease

    Neuropsychiatry Neuropsychol. Behav. Neurol.

    (1998)
  • ChandlerM.J. et al.

    A total score of the CERAD neuropsychological battery

    Neurology

    (2005)
  • ChangH. et al.

    Super-resolution through neighbor embedding

  • Deichmann, R., Haase, A., 1991. SMRM, “Book of Abstracts”,...
  • DespaF. et al.

    Dielectric modulation of biological water

    Phys. Rev. Lett. PRL

    (2004)
  • DraperN.R. et al.

    Applied Regression Analysis

    (1981)
  • Cited by (0)

    View full text