Structural and functional neuroimaging in patients with Parkinson's disease dementia
Gordon W. Duncan, Michael J. Firbank, and John T. o'Brien
Structural and functional neuroimaging studies have provided insights into the heterogeneous pathological, anatomical, and neurochemical correlates of cognitive impairment associated with the progression of Parkinson’s disease (PD) and the subsequent evolution of Parkinson’s disease dementia (PD-D). A key target for clinicians and researchers is to develop robust biomarkers for this common, distressing, and disabling complication of PD. This would permit the early identification of those people at highest risk of PD-D and streamline clinical trials of potentially disease-modifying therapies which are most likely to be effective when initiated early in the disease process, prior to extensive neuronal loss. The absence of such biomarkers has hampered the ability of interventional clinical trials to distinguish between the symptomatic effects and the potential disease-modifying properties of study medications. This chapter will provide an overview of the structural and functional imaging studies of PD and cognitive decline to date.
Structural magnetic resonance imaging
Considerable progress has been made in developing conventional structural magnetic resonance imaging (MRI) into a tool for differentiating Alzheimer’s disease (AD) from other causes of dementia, as a biomarker for observational studies of amnestic mild cognitive impairment (MCI) and AD, and, increasingly, as a surrogate outcome measure in trials of therapeutics (see Fig. 10.1). It is a commonly held view that MRI has the potential to fulfil a similar role in PD. MRI is an attractive option as a biomarker: it is safe, well tolerated by patients, non-invasive, and readily available. Crucially, for observational researchers and the pharmaceutical industry, it is not prohibitively costly and is up to ten times cheaper than many other imaging modalities being used in clinical trials.
Early MRI studies of PD and cognitive impairment commonly adopted a region of interest (ROI) approach; however, such methods are time-consuming, laborious, and involve critical issues regarding the choice of anatomical boundary, with potential for influencing intra- and inter-rater reliability. Although there are now fully automated protocols for segmenting individual brain structures such as the cortical lobar regions, hippocampus, basal ganglia, and thalamus, methods which analyse the entire grey matter volume are increasingly being used. Voxel-based morphometry (VBM) provides an automated, largely operator independent, and unbiased
Fig. 10.1 Coronal Trweighted MRI images of (from left to right) control, PD-D, and AD. Atrophy of the medial temporal structures is most severe in the AD subject (arrows) but mild in the PD subject compared with the control subject of a similar age.
method of assessing differences in the volume of brain tissue between groups of patients. Because the whole of the grey matter may be examined on a voxel-wise basis there is no requirement for an a priori hypothesis, permitting the study of cortical areas which may not have been previously considered relevant and thus ignored by ROI analyses. Corticometry, a measurement of cortical thickness and folding, is a relatively new whole-brain technique and may offer greater sensitivity in detecting early changes to the grey matter than VBM.
Studies using VBM analysis have consistently reported widespread cortical atrophy in patients with established PD-D, with diffuse bilateral loss of grey matter occurring in frontal, parietal, temporal, and occipital areas [1-6]. Studies using cortical thickness analysis have also observed extensive thinning of the frontal, temporal, and parietal cortices in those with PD-D compared with controls  and in the temporal lobes relative to PD patients without dementia . Compared with AD, most [1, 6, 9, 10], though not all [8, 11], studies have found less pronounced hippocampal atrophy in PD-D. Atrophy of other limbic structures, including the entorhinal cortex [11, 12], amygdala [13, 14], and anterior cingulate gyrus [4, 10], has also been reported. Atrophy of the caudate nucleus, putamen, and thalamus has been observed in PD-D using both VBM and ROI analyses [1, 4, 15].
Although one study has reported mild posterior atrophy in patients with early PD and intact cognition (PD-NC) , significant cortical atrophy in such patients has generally not been observed with VBM [6, 16, 17]. Cortical thinning has been reported in early PD-NC in frontal, temporal, and parietal regions ; however, this is not a consistent finding . In studies of PD patients without dementia, with a mean disease duration of over 5 years, VBM studies have demonstrated only mild cortical atrophy relative to control subjects [2, 3, 19, 20]. Mild cortical thinning in frontal, temporal, parietal, and occiptal regions has been observed with corticometry [8, 21, 22]. In PD patients without dementia, hippocampal and medial temporal lobe atrophy is not significant when assessed with ROI or VBM techniques [1, 2, 16, 23], although mild loss of grey matter in the amygdala has been reported .
PD with mild cognitive impairment (PD-MCI) may represent a transitional state to dementia in patients with PD. The publication in 2012 by the Movement Disorder Society (MDS) of guidance on the diagnosis and classification of PD-MCI aims to harmonize the definition of MCI and standardize methods of testing and classification . Using these criteria, Melzer et al.  reported mild loss of grey matter in the temporal, parietal, and frontal cortices and in the hippocampi of PD-MCI patients. In a further study, Mak et al.  found loss of grey matter volume in the left insular, left superior frontal, and left middle temporal areas in patients with PD-MCI compared with PD-NC. The predictive value of the VBM approach for determining which patients with PD-MCI will progress to PD-D has been addressed . At the initial scan, patients with PD-MCI who developed PD-D exhibited lower grey matter density in the left frontal area, left insular cortex, and bilateral caudate nucleus compared with patients who did not subsequently develop PD-D. Interestingly, the volume of the substantia innominata was significantly lower in patients who progressed to PD-D. This structure, which is located within the basal forebrain, contains the nucleus basalis of Meynert, one of the major suppliers of cholinergic innervation to the cerebral cortex.
Although not using MDS criteria, Pagonabarraga et al.  used corticometry to compare the thickness of grey matter across different cognitive stages of PD. Relative to PD-NC, subjects with PD-MCI displayed grey matter thinning involving the left anterior temporal pole, anterior cingulate cortex, entorhinal cortex, lingual gyrus, and precuneus. On the right, grey matter thinning was seen to involve the cuneus, lateral occipital cortex, inferior temporal cortex, and the fusiform gyrus. Compared with PD-MCI subjects, patients with PD-D displayed even greater atrophy in similar regions.
High-dimensional pattern classification algorithms, such as the regional analysis of volumes examined in normalized space (RAVENS) have been used to predict disease in individual subjects in studies of AD and MCI [28, 29]. An individual’s score is generated based upon measures of atrophy of regions such as the hippocampus, posterior cingulate, and peri-hippocampal white matter. Positive scores are indicative of a greater likelihood of dementia and a negative score is more in keeping with healthy control brains. This has been used identify patients who will subsequently progress to PD-D [20, 30]. After 2 years, those PD subjects without dementia with higher baseline SPARE-PDD (spatial pattern of abnormality for recognition of Parkinson’s disease with dementia- level cognitive deficits) scores had greater subsequent worsening of global cognitive performance.
Performance in neuropsychological tests has been correlated with atrophy of specific neuroanatomical structures. Atrophy of the prefrontal cortex has been correlated with increased reaction times and hippocampal atrophy has been associated with impairments of verbal memory . Semantic fluency has been correlated with grey matter volume in frontal and temporal areas ; visuospatial and visuoperceptual function with occipital grey matter ; and impaired decision making with reduced volume of the orbito-frontal cortex . Atrophy of the substantia innominata has been observed across all stages of PD [34, 35], and has been correlated with worsening attention, executive function, and verbal fluency .
Visual hallucinations are a prominent feature of PD-D [37, 38]. Patterns of grey matter loss similar to those observed in PD-D are reported in PD subjects without dementia who suffer from visual hallucinations. Compared with those without visual hallucinations, patients who hallucinate display diffuse grey matter loss in the superior parietal lobes (bilaterally), the right medial frontal gyrus, right lingual gyrus, left inferior parietal lobe, and left occipital lobe . After a mean of 29 months’ follow-up, the patients with hallucinations had higher rates of conversion to dementia and extensive grey matter loss than those without hallucinations .
Few studies have examined longitudinal morphological brain changes in PD. Using the brain boundary shift integral technique, Burton et al.  observed significantly higher rates of brain atrophy in those with PD-D (12.2 ml/year; 1.12%) compared with PD patients without dementia (3.4 ml/year; 0.31%) and healthy controls (3.8 ml/year; 0.34%). In PD without dementia rates of atrophy similar to controls have been reported in some, but not all, studies [41-43]. Ventricular dilation may reflect concurrent cerebral atrophy and has been used as a measure of atrophy in studies of AD. Progressive ventricular enlargement of the lateral ventricles in PD-D has been observed , and enlargement of the left inferior lateral ventricle and third ventricle is reported in patients with PD-MCI [45, 46].
In routine clinical practice the principal role of MRI is to assist in the differential diagnosis of PD or PD-D from another akinetic-rigid disorder or another dementia syndrome. As discussed, atrophy of structures within the medial temporal lobe is not severe early in the disease course of PD or PD-D when compared with AD; however, with disease progression there is often mild generalized global atrophy. Patients with progressive supranuclear palsy may display prominent midbrain atrophy, while those with multiple system atrophy often show either pontine or cerebellar atrophy. More common, however, is the demonstration of cerebrovascular disease in the form of stroke, lacunar infarcts, microbleeds, or white matter hyperintensities. White matter hyperintensities are a frequent neuroimaging finding in older people and have been associated with both cognitive impairments and mild parkinsonian signs [47-49].