A review of the use of magnetic resonance imaging in Parkinson’s disease

Abstract

To date, the most frequently used Parkinson’s disease (PD) biomarkers are the brain imaging measures of dopaminergic dysfunction using positron emission tomography and single photon emission computed tomography. However, major advances have occurred in the development of magnetic resonance imaging (MRI) biomarkers for PD in the past decade. Although conventional structural imaging remains normal in PD, advanced techniques have shown changes in the substantia nigra and the cortex. The most well-developed MRI markers in PD include diffusion imaging and iron load using T2/T2* relaxometry techniques. Other quantitative biomarkers such as susceptibility-weighted imaging for iron load, magnetization transfer and ultra-high-field MRI have shown great potential. More sophisticated techniques such as tractography and resting state functional connectivity give access to anatomical and functional connectivity changes in the brain, respectively. Brain perfusion can be assessed using non-contrast-agent techniques such as arterial spin labelling and spectroscopy gives access to metabolites concentrations. However, to date these techniques are not yet fully validated and standardized quantitative metrics for PD are still lacking. This review presents an overview of new structural, perfusion, metabolic and anatomo-functional connectivity biomarkers, their use in PD and their potential applications to improve the clinical diagnosis of Parkinsonian syndromes and the quality of clinical trials.

Introduction

The diagnosis of Parkinson’s disease (PD) is mainly based on clinical features and does not rely on imaging biomarkers. Commonly used imaging techniques in PD include positron emission tomography (PET) and single photon emission computed tomography (SPECT), but conventional magnetic resonance imaging (MRI) is of little use in clinical practice. However, advances in structural and functional imaging have improved the capacity of MRI to detect changes in PD as well as to differentiate between PD and other Parkinsonian syndromes. MRI has provided several candidate biomarkers that have a potential to inform on the disease process. Biomarkers are quantitative characteristics which are used as indicators of biological or pathological states. In neuroimaging, biomarkers are measures derived from images that reflect the presence of diseases or their severity and that can be used for early diagnosis, prognosis or to monitor responses to therapeutic interventions. Biomarkers are expected to detect early neuropathological features and mechanisms underlying neurodegeneration in PD and to correlate with disease progression in order to allow the monitoring of disease status. Ideally, they should be able to detect preclinical changes. Reliable biomarkers need to be confirmed by independent studies.

Promising candidate biomarkers were able to detect changes at various levels of the central nervous system, including the substantia nigra (SN), the brainstem, the basal ganglia and the cortex. New MR contrasts and image analysis techniques have improved the visualization of the SN and other brainstem nuclei. Cortical lesions can be followed using various techniques that can detect changes in the thickness, volume and density of grey matter in the cortex. Quantitative imaging includes techniques such as relaxometry, susceptibility-weighted, diffusion and magnetization transfer (MT) imaging. Perfusion can be assessed using non-contrast-agent technique MRI and brain metabolism using magnetic resonance spectroscopy. Tractography was used to investigate anatomical connectivity and resting state functional MRI (rsfMRI) provides new information on functional connectivity changes in the brain of PD patients. Task-related activation functional MRI (fMRI) will not be reviewed here. In the SN, although most MRI biomarker candidates were not validated in experimental models and require confirmation in independent studies, it is clear that they have the potential to quantify the pathology and follow the disease progression of PD.

This article reviews new MRI techniques and their use in PD and explains how these techniques may be used to detect changes in the brain of PD patients and their relationships with Parkinsonian symptoms. A summary of these techniques is presented in Table 1 and Figure 1. Each imaging technique and their respective contribution to the study of PD will be presented in the following sections. They include structural imaging, techniques that can assess iron load, other quantitative techniques (e.g. MT, diffusion imaging), perfusion imaging, magnetic resonance spectroscopy and techniques that can assess anatomical and functional connectivity. For the review, the authors searched PubMed from 1990 until early 2013 by using the terms corresponding to each technique and the terms corresponding to the pathological conditions, i.e. Parkinson’s disease and related atypical Parkinsonian conditions.

Table 1.

MRI techniques.

Method	Techniques	Information	Changes in PD
Structural	T1-w, T2-w, IR, MT	Morphometry	SN: variable volume changes Cortex: mild reduction in volume and thickness
Neuromelanin	Spin echo T1-w	Presence of melanin-containing cells	Reduced signal intensity
Magnetization transfer	Images with (MT) and without (M0) MT pulse	Degree of myelination, axonal density	Reduced MT ratio (MTR = M0 – MT/M0)
Relaxometry	T2/T2* measurements	Brain iron	Reduced T2/T2*
			Increased R2/R2*
Susceptibility-weighted	Phase images	Brain iron	Increased susceptibility due to iron load
Diffusion	DTI	Diffusion of water in biological tissues	Reduced FA
Tractography	DTI	Fibre tract-specific reconstruction	Reduced probability of connections
Functional	Resting state BOLD fMRI	Functional connectivity within brain networks	Reduced FC in sensorimotor
			Increased FC in associative
MR spectroscopy	1H MRS	NAA, Cho, mIns, GABA, Glx, GSH, lactate	Trend for metabolite reduction
	31P MRS	Energy metabolism: ADP, PCr/ATP	Decreased ATP in midbrain
Perfusion	ASL	rCBF	Reduced in cortex, Variable in basal ganglia

¹H MRS, proton magnetic resonance spectroscopy; ³¹P MRS, phosphorus magnetic resonance spectroscopy; ADP, adenosine diphosphate; ASL, arterial spin labelling; ATP, adenosine triphosphate; BOLD, blood oxygen level-dependent contrast; Cho, choline-containing compounds; DTI, diffusion tensor imaging; FA, fractional anisotropy; FC, functional connectivity; fMRI, functional magnetic resonance imaging; Glx, glutamate/glutamine; GSH, glutathione; IR, inversion recovery; mIns, myo-inositol; MT, magnetization transfer; MTR, magnetization transfer ratio; NAA, N-acetylaspartate; rCBF, regional cerebral blood flow; PCr, phosphocreatine; SN, substantia nigra; T2, T2 relaxation time; T2*, gradient echo T2 relaxation time; T1-w, T1-weighted; T2-w, T2-weighted; R2, T2 relaxation rate; R2*, T2* relaxation rate.

Figure 1.

Overview of MRI methods used to study PD. Cortex: changes were detected using voxel-based techniques, cortical thickness measurements and perfusion imaging. Brain connectivity was investigated using resting-state functional MRI (rs-fMRI) for functional connectivity and tractography for structural connectivity. Substantia nigra (yellow contour): changes were detected using diffusion imaging (reduced fractional anisotropy - FA), relaxometry (increased R2* indicating increased iron load and more recently susceptibility-weighted imaging), magnetization transfer ratio (MTR reduced) and spectroscopy. Basal ganglia: studies showed no or mild changes in FA, R2* or MTR. Locus coeruleus area (white arroxw head): reduced signal intensity was detected using neuromelanin imaging.

Structural imaging

Techniques for visualization and segmentation of brain regions

MRI sequences for visualization of the basal ganglia and the cortex

Conventional three-dimensional T1-weighted (T1-w) sequences present high grey/white matter contrast for cortical and some basal ganglia structures, but poor contrast in many structures of interest in PD (SN, subthalamic nucleus [STN], globus pallidus [GP], red nucleus [RN]). These latter structures contain high iron levels leading to shortened T1 and reduced contrast [Hardy et al. 2005]. Iron load is an advantage in T2-w and T2*-w sequences that use the T2 shortening effects of iron providing enhanced contrast. Susceptibility-weighted imaging (SWI) is promising in PD due to its improved sensitivity to brain mineralization [Haacke et al. 2005]. SWI uses the magnetic susceptibility differences in brain tissues [Haacke et al. 2005; Lotfipour et al. 2012; Schafer et al. 2012; Wang et al. 2012]. SWI images are sensitive to iron concentration and improve visualization of the SN [Haacke et al. 2005; Lotfipour et al. 2012; Schafer et al. 2012; Wang et al. 2012]. Proton-density and short tau inversion recovery (STIR) images were suggested as good alternatives. Other contrasts, such as the MT [Helms et al. 2009; Menke et al. 2010] and inversion recovery sequences [Hutchinson et al. 2003] also improved visualization of the SN. MT pulses increase contrast between the SN and the surrounding white matter, resulting in more accurate delineation and improved visualization of these structures [Helms et al. 2009; Menke et al. 2010]. Lastly, the increase in magnetic field provided by 7T human MRI, provided not only increased spatial resolution but also improved contrast, which allowed better visualization of basal ganglia contours and shapes. The volume and contours of the SN depend on the MRI contrast that was used [Oikawa et al. 2002]. For instance, contours that were shown using the T2-w sequences, which are sensitive to iron load, probably differed from those detected using inversion recovery and neuromelanin imaging.

The anatomy of the cortex is usually better depicted using conventional three-dimensional T1-weighted sequences which present high grey/white matter contrast.

Segmentation and volume calculation

Small structures such as the SN are usually outlined and segmented manually. Some cortical regions (e.g. the hippocampus, the striatum and the thalamus) can be segmented manually, but automated techniques are preferred as exploratory tools to detect grey and white matters changes in the cortex [Destrieux et al. 2010; Patenaude et al. 2011]. These techniques are designed for group studies, are automated, whole brain and rater independent. They include voxel-based, deformation-based and tensor-based morphometry, cortical thickness and sulci measurements. Voxel-based morphometry (VBM) is now a standard technique, sensitive to differences in grey and white matter [Ashburner and Friston, 2000]. VBM provides metrics such as the concentration, density and volume of grey matter. A major limitation of VBM is the nonspecificity with respect to the underlying tissue changes. Deformation-based and tensor-based morphometry provide information about global or local differences in shape, respectively [Apostolova and Thompson, 2007; Good et al. 2001]. Other software allows measurements of topographic differences in thickness, surface area and curvature of the cortex [Dale et al. 1999; Fischl et al. 1999; Makris et al. 2005].

Volume and shape changes in Parkinson disease

Substantia nigra

The volume of the SN is usually measured using manual segmentation. Results in PD are discordant with normal [Oikawa et al. 2002; Peran et al. 2010], reduced [Hutchinson et al. 2003; Menke et al. 2009; Minati et al. 2007] and even increased volumes [Cho et al. 2011; Kwon et al. 2012] (Table 1). The origin of this variability is probably due to differences in methodology between studies. First, the variability of volume measurements is large, given the small size of the structure. Second, different contrasts were used resulting in changes in SN contours. Third, some methods relied on tractography and anatomical connectivity changes which are completely different in nature as compared with other structural methods. Lastly, increased SN volume was obtained using 7T imaging and may be explained by the increased volume of regions with iron load. Inversion recovery methods were able to detect a predominant involvement of the lateral segments of the SN [Hutchinson et al. 2003; Minati et al. 2007] in line with the preferential degeneration of dopaminergic neurons in the caudal and lateral SN in histological studies [Redgrave et al. 2010]. Overall, although volume changes were detected in the SN of PD patients, changes varied between studies and the reproducibility was low. Ultra-high-field MRI may provide improved results due to increased spatial resolution.

Other basal ganglia and brainstem nuclei

Global measurements of brainstem volumes [Longoni et al. 2011; Oba et al. 2005] and rate of atrophy [Paviour et al. 2006] were usually normal in PD patients. Similarly, most studies did not find any global volumetric changes in the basal ganglia [Bonneville et al. 2005; Oba et al. 2005; Paviour et al. 2006]. Neuromelanin-sensitive imaging showed reduced signal intensity in the locus coeruleus in PD patients compared with controls, probably reflecting loss of neuromelanin-containing neurons [Sasaki et al. 2006] (Table 1). A reduction in signal intensity in the locus coeruleus/subcoeruleus area correlated with the percentage of atonia during rapid eye movement (REM) sleep suggesting that damage to this region is implicated in the pathophysiology of REM sleep behaviour disorder [Garcia-Lorenzo et al. 2013].

Cortex

The understanding of brain mechanisms underlying cognitive impairment is of particular interest in PD. Mild cognitive impairment is common in nondemented PD patients and is a risk factor of progression to dementia, which develops in about 30% of PD patients [Litvan et al. 2011; Svenningsson et al. 2012]. Impaired cognition in PD has a heterogeneous profile including attention, executive, visuospatial and memory deficits [Svenningsson et al. 2012]. Different lesions may explain the different cognitive deficits [Brooks and Pavese, 2010]. Executive dysfunction may be related to dopaminergic fronto-subcortical deficits whereas dementia may be associated with widespread atrophy in posterior parietal and temporal cortical areas, associated with Lewy body deposition or Alzheimer plaques, and with nondopaminergic neurotransmitter dysfunctions such as cholinergic deficiency [Brooks and Pavese, 2010; Ibarretxe-Bilbao et al. 2009, 2011a; Svenningsson et al. 2012]. Structural MRI is a powerful technique to detect cortical changes in PD and evidence for cortical damage will be presented below.

Nondemented cognitively intact PD

Cortical morphology in cognitively intact PD patients may be normal [Cerasa et al. 2011; Feldmann et al. 2008; Messina et al. 2011; Paviour et al. 2006; Rizzo et al. 2008; Seppi and Poewe, 2010; Tessitore et al. 2012a] or altered in frontal regions [Biundo et al. 2011; Burton et al. 2004; Karagulle Kendi et al. 2008] or in more widespread cortical regions including parietal, temporal and occipital areas [Jubault et al. 2011; Lyoo et al. 2010; Meppelink et al. 2010; Tinaz et al. 2011]. Variations may be related to changes in methodology or disease stage but changes were most often mild. Cortical atrophy in fronto-temporal regions accelerates with disease progression [Ibarretxe-Bilbao et al. 2012].

Mild cognitive impairment and dementia

In contrast, impaired cognition and dementia in PD were associated with more severe patterns of cortical atrophy [Beyer et al. 2007; Burton et al. 2004] as reviewed by Ibarretxe-Bilbao and colleagues [Ibarretxe-Bilbao et al. 2009, 2011a]. Regions predictive of cognitive decline included the hippocampus and parieto-temporal regions [Weintraub et al. 2012]. The Alzheimer’s disease pattern of brain atrophy may thus be a preclinical biomarker of impaired cognition in PD patients.

Cortical correlates of other nonmotor dysfunctions

Apart from impaired cognition and dementia, other nonmotor symptoms in PD include mood disorders, psychotic symptoms, hyposmia and sleep disorders [Brooks and Pavese, 2010]. Depression in PD patients was associated with structural changes in the limbic system including medial, orbitofrontal and temporal regions [Feldmann et al. 2008; Kostic et al. 2010] and the mediodorsal thalamus [Cardoso et al. 2009]. Visual hallucinations were associated with atrophy in limbic, paralimbic and neocortical areas in some studies [Ibarretxe-Bilbao et al. 2010, 2011b], but not others [Meppelink et al. 2010]. Olfactory dysfunction was related to grey matter atrophy in olfactory regions including the piriform cortex and amygdala [Wattendorf et al. 2009]. Nonmotor symptoms were thus associated with structural changes in specific related brain networks.

Morphometric and signal changes in other Parkinsonian syndromes

Volumetric MRI was efficient for differentiating Parkinsonian disorders such as progressive supranuclear palsy (PSP) and the Parkinson variant of multiple system atrophy (MSA-P) from PD. There changes were reviewed extensively [Seppi and Poewe, 2010]. In contrast to PD, where structural changes are subtle or absent, PSP patients present with extensive atrophy in the brainstem, basal ganglia and cortex. Midbrain atrophy in PSP has a characteristic shape described as the ‘penguin’ or ‘hummingbird’ sign [Kato et al. 2003; Righini et al. 2004] and can be quantified using various indexes [Oba et al. 2005; Paviour et al. 2006; Quattrone et al. 2008; Righini et al. 2004]. Using VBM, grey and white matter reductions have been constantly reported in PSP in the midbrain and more variably in the basal ganglia, the frontal cortex, the insula and the thalamus [Boxer et al. 2006; Brenneis et al. 2004; Cordato et al. 2005; Josephs et al. 2008; Lehericy et al. 2010; Messina et al. 2011; Padovani et al. 2006]. In MSA-P, atrophy and signal changes were observed in the putamen (hypointensity in the dorsolateral part and hyperintense rim around the putamen), the pons (cross sign), the middle cerebellar peduncle (T2 hyperintensity) and the cerebellum [Seppi and Poewe, 2010]. PSP patients had greater midbrain and superior cerebellar atrophy whereas MSA-P patients had greater pontine and middle cerebellar atrophy [Oba et al. 2005; Paviour et al. 2006; Quattrone et al. 2008; Righini et al. 2004]. In MSA-P patients atrophy was also evidenced in the cortex using VBM [Brenneis et al. 2003; Minnerop et al. 2007]. Quantitative volumetric measures of pontine and midbrain areas [Oba et al. 2005; Quattrone et al. 2008] were able to discriminate accurately PSP and MSA-P from PD patients or healthy controls.

Iron load

Iron load may be estimated using measurements of T2 and T2* relaxation times and more recently, using SWI (Table 1). T2 relaxation times characterize how fast water magnetization returns to equilibrium after perturbation by a radiofrequency pulse during an MRI sequence. R2 relaxation rates correspond to 1/T2. T2 relaxation times are characteristic of tissue composition and depend on its molecular structure. The apparent transverse relaxation time T2* (T2 star) is sensitive to the macroscopic and microscopic inhomogeneities of the magnetic field. Macroscopic inhomogeneities are linked to hardware imperfections or to the presence of tissue interfaces characterized by a significant gradient of susceptibility. Microscopic inhomogeneities are linked to the tissue microstructure. Many experiments have shown that T2/R2 and T2*/R2* relaxation rates are noninvasive estimates of iron content either in primate or postmortem studies in humans, with better results obtained using R2* [Hardy et al. 2005; Langkammer et al. 2010; Ordidge et al. 1994]. Improved relaxation measurement techniques have recently been proposed based on adiabatic pulse sequences (T1ρ and T2ρ) [Menke et al. 2010; Michaeli et al. 2005, 2007]. Changes in iron concentration also result in changes in the phase of the tissue relative to its surroundings, a property which is used in SWI to quantify iron load [Haacke et al. 2005; Lotfipour et al. 2012; Schafer et al. 2012; Wang et al. 2012] (Table 1).

In PD iron load is increased [Dexter et al. 1989]. T2 relaxometry has provided largely consistent and reproducible results with decreased T2/T2*/adiabatic T2ρ relaxation times and increased R2/R2* relaxation rates [Baudrexel et al. 2010; Dexter et al. 1989; Du et al. 2011; Graham et al. 2000; Martin et al. 2008; Michaeli et al. 2007; Ordidge et al. 1994; Peran et al. 2010; Wang et al. 2012; Zhang et al. 2010]. However, some studies observed a change in T2 only in a particular subgroup [Ordidge et al. 1994] or no change at all. A predominance of R2* changes in the lateral SNc [Martin et al. 2008], in line with the location of predominant cell loss reported in histological studies [Damier et al. 1999]. Changes in R2* in the SN correlated with the motor score of the UPDRS [Martin et al. 2008] or with the levodopa equivalent daily dose scores [Peran et al. 2010].

In PSP [Boelmans et al. 2012] and MSA-P, changes in T2 relaxation times affected more widespread regions including the striatum and the GP as shown using phase-contrast susceptibility imaging [Wang et al. 2012] or relaxometry [Arabia et al. 2010; Boelmans et al. 2012].

Other quantitative biomarkers

In addition to relaxation times and rates, other quantitative markers include magnetization transfer ratio (MTR) and diffusion metrics such as fractional anisotropy (FA) and mean diffusivity (MD).

Magnetization transfer

MT imaging is a technique which relies on the transfer of energy between highly bound protons, e.g. linked to macromolecules, and mobile protons, i.e. in free water [Wolff and Balaban, 1989] (Table 1). Highly bound protons linked to macromolecules are present within structures such as myelin. The amount of MT is thus considered to correlate with the degree of myelination [Rademacher et al. 1999] and axonal density [van Waesberghe et al. 1999]. MT is therefore used in diseases such as multiple sclerosis [Dousset et al. 1992]. MT provides a quantitative parameter, the MTR [Dousset et al. 1992]. The MTR was reduced in the SN as well as in other basal ganglia (GP, the putamen, caudate nucleus) in PD patients [Eckert et al. 2004; Tambasco et al. 2011].

Water diffusion

Diffusion MRI is technique that is sensitive to water diffusion in biological tissues [Le Bihan, 2003]. Diffusion MRI provides several indexes that characterize the overall displacement of molecules (apparent diffusion coefficient [ADC] and MD), the orientation of diffusion (FA) and the characteristics of diffusion along the main direction of diffusion (axial or longitudinal diffusivity) and perpendicular to it (radial or transverse diffusivity) [Le Bihan, 2003]. At the cellular level, anisotropy may result from the presence of obstacles to diffusion due to oriented structures such as the membrane, myelin, longitudinal filaments and cytoskeleton [Le Bihan, 2003]. Changes in axial diffusivity may reflect axonal damage whereas changes in radial diffusivity may reflect myelin damage. Although efforts are pursued to develop methods that better characterize brain changes [Alexander, 2008; Assaf et al. 2008; Jensen et al. 2005; Zhang et al. 2012], to date diffusion parameters suffer from a lack of specificity and poor knowledge of the underlying neuronal substrates.

In the SN of PD patients, changes in the orientation of diffusion, i.e. anisotropy, appeared more consistent than changes in diffusivity (Table 1). Most studies reported reduced anisotropy in early PD patients [Chan et al. 2007; Du et al. 2011; Peran et al. 2010; Vaillancourt et al. 2009; Yoshikawa et al. 2004; Zhan et al. 2012], however others did not [Menke et al. 2009, 2010]. Changes in diffusion kurtosis, a method that quantifies non-Gaussian water diffusion [Jensen et al. 2005] were also reported in PD patients [Wang et al. 2011]. In contrast, diffusivity was mostly unchanged in PD patients [Seppi and Poewe, 2010; Seppi and Schocke, 2005]. Reduction in FA in the SN inversely correlated with clinical severity in some studies [Chan et al. 2007; Zhan et al. 2012] but not in other studies [Du et al. 2011].

In the striatum, diffusion measures were most often normal [Cochrane and Ebmeier, 2013; Nicoletti et al. 2006; Paviour et al. 2007] although few studies reported increased diffusivity in the striatum and thalamus in PD [Peran et al. 2010; Zhan et al. 2012].

Diffusion imaging is also useful to discriminate PD from other Parkinsonian syndromes [Cochrane and Ebmeier, 2013]. In PSP, diffusivity was increased in the basal ganglia [Rizzo et al. 2008; Seppi et al. 2003], the midbrain [Blain et al. 2006; Seppi and Poewe, 2010; Seppi and Schocke, 2005], the superior cerebellar peduncle [Nicoletti et al. 2008], the precentral white matter [Ohshita et al. 2000] and the total white matter [Padovani et al. 2006]. In MSA-P, diffusivity was increased and anisotropy was decreased in the putamen, pons and middle cerebellar peduncle [Blain et al. 2006; Nicoletti et al. 2006; Paviour et al. 2007; Schocke et al. 2004] and MTR was reduced in the putamen [Eckert et al. 2004].

Differential diagnosis of Parkinsonian syndromes using quantitative biomarkers

Several studies have suggested that the combination of R2* and FA markers may better differentiate people with PD from healthy aged subjects with greater than 95% global accuracy [Du et al. 2011; Peran et al. 2010].

In PSP, increased diffusivity and reduction in T2 relaxation times were widespread involving the basal ganglia [Boelmans et al. 2012; Rizzo et al. 2008; Seppi et al. 2003], the midbrain [Blain et al. 2006; Seppi and Poewe, 2010; Seppi and Schocke, 2005], the superior cerebellar peduncle [Nicoletti et al. 2008] and the white matter [Ohshita et al. 2000; Padovani et al. 2006].

In MSA-P, increased diffusivity and reduced anisotropy was found in the putamen, pons and middle cerebellar peduncle [Blain et al. 2006; Nicoletti et al. 2006; Paviour et al. 2007; Schocke et al. 2004] and greater iron deposition in the putamen, using phase-contrast susceptibility imaging [Wang et al. 2012] or relaxometry [Arabia et al. 2010; Boelmans et al. 2012].

Tractography and anatomical connectivity

Tractography is a technique based on the anisotropy of water diffusion that allows reconstructing fibre tracts in the brain. Tracks are reconstructed by assuming that the main direction of diffusion in a voxel indicates the local orientation of white matter fibres [Dell’acqua and Catani, 2012; Mori et al. 1999]. Using tractography, diffusion measures (MD, FA) and connectivity measures (e.g. the number of tracks or connection probability between brain regions) can be calculated within the specific fibre tracks. Tractography has also been successfully used to parcel out the SN [Menke et al. 2009] and the basal ganglia into specific territories [Draganski et al. 2008; Lehericy et al. 2004].

In PD, reduced connectivity was observed between the SN and ipsilateral putamen and thalamus [Menke et al. 2009; Sharman et al. 2012] as well as in the sensorimotor circuit of the basal ganglia [Sharman et al. 2012] (Table 1). Automated diffusion-based parcellation of SN subregions showed that the SNr and SNc in PD patients showed a general atrophy [Menke et al. 2010]. Tractography is therefore used to investigate changes in anatomical connectivity in PD patients [Menke et al. 2009, 2010; Sharman et al. 2012].

Perfusion

Perfusion can be assessed noninvasively with MRI without contrast agent administration using arterial spin labelled (ASL) techniques [Detre et al. 1992]. ASL techniques have been recently introduced as a noninvasive alternative to PET or SPECT imaging for perfusion measurements in PD [Fernandez-Seara et al. 2012; Kamagata et al. 2011; Ma et al. 2010; Melzer et al. 2011]. In ASL, the tracer is endogenous arterial blood water that is labelled electromagnetically using radiofrequency (RF) irradiation [Detre et al. 2012]. ASL perfusion imaging provides quantitative measurements of cerebral blood flow [Detre et al. 2012].

Compared with healthy controls, perfusion in PD was decreased in the cortex [Fernandez-Seara et al. 2012; Kamagata et al. 2011; Melzer et al. 2011] and either preserved [Melzer et al. 2011] or decreased [Fernandez-Seara et al. 2012] in the basal ganglia and preserved in the sensorimotor areas [Melzer et al. 2011] (Table 1). Perfusion patterns were correlated with metabolic changes using fluorodeoxyglucose PET [Ma et al. 2010].

Metabolic biomarkers

Proton magnetic resonance spectroscopy (¹H MRS) provides information about the levels of metabolites in the brain (Table 1). Metabolites that are evaluated include N-acetylaspartate (NAA; considered as a marker of neuronal number and health), choline-containing compounds (Cho; markers of demyelination and cell proliferation), myo-inositol (mIns; considered as a marker for osmotic stress or astrogliosis), lactate (frequently associated brain pathologies), glutathione (GSH) as well as neurotransmitters such as glutamate/glutamine or GABA (using specific editing techniques). Metabolite levels may be reported as their ratios over sum of creatine and phosphocreatine (Cr; which used a concentration reference) or as absolute concentrations. MRS is highly sensitive to magnetic field homogeneity and lipid artefact signals when the regions examined are too close to the skull. MRS of the SN is therefore very challenging in humans due to the location, small size and high iron content of the structure. Regions studied present partial voluming with the rest of the midbrain.

Using ¹H-MRS, two groups reported changes in metabolites at 3T [Groger et al. 2011; Hattingen et al. 2009]. This was not confirmed at higher field as MRS at 4T [Oz et al. 2006] or 7T [Emir et al. 2012] did not show significant changes in the SN of PD patients. The authors only report a trend for Glu, NAA and GSH to decrease, and for Cho to increase [Oz et al. 2006].

Phosphorus magnetic resonance spectroscopy (³¹P MRS) allows direct monitoring of the energy metabolism of the brain [Ross and Bluml, 2001] (Table 1). PD patients showed abnormal phosphate metabolisms [Barbiroli et al. 1999; Hattingen et al. 2009; Hu et al. 2000] that suggested mitochondrial dysfunction of mesostriatal neurons [Barbiroli et al. 1999; Hattingen et al. 2009; Hu et al. 2000; Rango et al. 2006].

Functional connectivity

Initially designed to detect task-related signal changes in the brain, blood oxygen level-dependent (BOLD) fMRI is also increasingly used to study functional connectivity between distant brain regions. Spatially remote regions forming distributed cortical and subcortical networks can present temporal correlations of the fluctuations of their fMRI signal at rest [Biswal et al. 1995; Friston et al. 1993]. These networks are extracted from rsfMRI data using various mathematical approaches [Bullmore and Sporns, 2009; Ramnani et al. 2004] (Table 1).

Functional interactions in resting brain networks are abnormal in PD [Helmich et al. 2009; Hacker et al. 2012; Sharman et al. 2012; Skidmore et al. 2011b; Wu et al. 2011]. rsfMRI studies have shown that PD is associated with changes in cerebral connectivity between the basal ganglia and the cortex or cerebellum [Helmich et al. 2009] and between the STN and cortical motor and premotor areas [Baudrexel et al. 2011]. Changes predominate in the sensorimotor circuit with decreased functional coupling whereas functional connectivity was increased in the associative territory [Sharman et al. 2012; Hacker et al. 2012; Helmich et al. 2009]. Abnormal functional connectivity was also evidenced in the default-mode network in cognitively unimpaired PD patients that correlated with cognitive performances in memory and visuospatial tests [Tessitore et al. 2012b]. In drug-naïve PD patients, the supplementary motor area showed reduced signal fluctuations in PD whereas levodopa enhanced functional connectivity in this region [Esposito et al. 2013]. Changes were observed in a specific frequency band [Esposito et al. 2013]. Changes in resting state BOLD fluctuations were successful in predicting the presence of PD [Long et al. 2012; Skidmore et al. 2011b].

Changes in functional connectivity were differently associated with symptoms in PD. Tremor has been related to increased functional connectivity between the STN and the primary sensorimotor area [Baudrexel et al. 2011] and to abnormal functional interactions between the basal ganglia and the cerebello-thalamic circuit [Helmich et al. 2011]. Apathy was related to abnormal functional connectivity in the orbitofrontal, supplementary motor and middle frontal areas [Skidmore et al. 2011a]. In contrast, depression was associated with functional connectivity changes in the subgenual cingulate cortex [Skidmore et al. 2011a]. Increased amplitude of low-frequency BOLD signal oscillations in the premotor cortex predicted motor performances [Kwak et al. 2012].

In PSP, rsfMRI showed connectivity disruptions between the dorsal midbrain tegmentum and the cerebellum, diencephalum, basal ganglia and cortex that were associated with more severe functional impairment [Gardner et al. 2013]. Another study found disconnection between the thalamus and the striatum, supplementary motor area and cerebellum [Whitwell et al. 2011].

Overall, findings with resting state fMRI suggest that dopamine depletion in PD leads to a remapping of cerebral connectivity that affects predominantly the sensorimotor circuit and sensorimotor integration, that is influenced by levodopa and differently associated with motor and nonmotor symptoms. rsfMRI methods also showed reduced brainstem and thalamic connectivity in PSP that correlated with clinical deficits.

Conclusion

Over the last decade and a half, research in PD imaging has focused on developing quantitative techniques in the aim of providing measurements that can be used a markers of the disease process. In the SN, several techniques including diffusion imaging, T2* relaxometry and MT have shown great promises for detecting pathological changes. Investigators have developed computerized neuroanatomical techniques to study cortical changes that have been related to specific motor and nonmotor symptoms in the disease. MRI gives also access to the anatomical and functional connectivity changes in the brain of PD patients. MRI techniques could be used to monitor disease progression and to detect brain changes in preclinical patients or in patients at risk of developing PD, such as gait REM sleep behaviour disorder patients. However, to date these techniques suffer from the lack of standardization, particularly methods for extracting quantitative information from images, and of validation in large cohorts of subjects in longitudinal studies. It is expected that future efforts in imaging research will provide significant improvement in this respect.