Dementia with Lewy bodies (DLB) is the second most common form of neurodegenerative dementia following Alzheimer’s disease (AD), accounting for approximately 15% of cases at autopsy [1]. Characterised by cognitive fluctuations, visual hallucinations, and motor Parkinsonism, DLB shares both clinical and pathological features with other dementia types, including Parkinson’s disease dementia (PDD) and AD. Hence, DLB is a frequently misdiagnosed condition, and previously established consensus criteria are limited by low sensitivity. In light of this uncertainty in diagnosis, and with important implications for subsequent patient management, more reliable imaging markers are needed to help distinguish DLB from other subtypes of dementia. In this review, we provide a literature summary of the principal neuroimaging techniques used to investigate DLB in terms of its differentiation from other dementia types. The diagnostic standard for all mentioned studies is clinical criteria except where otherwise stated.

Besides nuclear imaging methods, the clinical utility of structural neuroimaging with magnetic resonance imaging (MRI) for differential diagnosis of dementias is also well established. Structural MRI has been widely used to compare regional structural changes in patients with DLB against AD, PDD and their healthy controls. In this section, we summarise the principal findings using MRI to distinguish DLB from other subtypes of dementia (Table 1).

Comparison between dementia with Lewy bodies and Alzheimer’s disease

Both cross-sectional and longitudinal studies have shown that DLB is associated with less pronounced global atrophy than AD [28]. The most consistent finding is the relative preservation of the medial temporal lobe (MTL) in DLB compared to AD [28] (Figure 2). Burton and colleagues [29] demonstrated the clinical significance of MTL atrophy for distinguishing DLB from AD in pathologically confirmed cases, although another autopsy study argued that the presence of MTL atrophy may not rule out a diagnosis of DLB, especially amongst patients in the oldest old-age category (>85 years) [30]. In addition, Burton and colleagues [29] reported a strong correlation between hippocampal atrophy and amyloid-beta plaques and neurofibrillary tangles but not Lewy body-associated neuronal inclusions, suggesting that gray matter (GM) loss in DLB could be due to concomitant AD pathology. Hippocampal atrophy in DLB and AD has also been investigated, showing less severe atrophy in DLB than in AD when compared to healthy controls [31]. Similarly, other studies have demonstrated that the entorhinal cortex, CA1, and subiculum areas of the hippocampus may be most affected in AD compared to DLB [28, 32].

In parallel with the MTL findings, patients with DLB generally demonstrate relative preservation of episodic memory capabilities compared to AD while experiencing impairments in attentional and visuospatial tasks. Hence, Ballmaier and colleagues [33] hypothesised that frontal lobe structures could be spared in DLB and the reciprocal connections with relatively preserved temporal regions would be affected to a lesser extent too. Although their study revealed greater frontal GM loss in AD relative to DLB [33], other studies have not found differences between these patient groups [34].

Conversely, the midbrain appears to be more affected in DLB than in AD, with the substantia innominata showing greater atrophy [35, 36]. The substantia innominata contains the nucleus basalis of Meynert, which is highly involved in the cholinergic neurotransmitter system. In addition, Whitwell and colleagues [35] demonstrated that increased dorsal mesopontine GM atrophy distinguished patients with clinically diagnosed DLB and AD, and this finding was further confirmed in an autopsy study involving patients with high likelihood DLB [37]. These findings are thus in agreement with greater cholinergic dysfunction in DLB, and perhaps relate to the presence of midbrain synuclein pathology (see the 'Cholinergic pathways' section).

In light of the ¹²³FP-CIT SPECT functional changes involving the loss of DAT in DLB [4], associated subcortical structures have also been investigated relative to AD. Cousins and colleagues [38] have observed reduced putamen volume in patients with DLB, which may be associated with the striatal synuclein pathology.

Active task and resting state functional MRI (fMRI) are the main neuroscience tools to examine cerebral function related to cognitive tasks or during rest through changes in blood oxygen level-dependent signal. There are still relatively few fMRI studies in DLB, but different patterns of functional connectivity between AD and DLB have been reported (Table 1). A recent resting-state fMRI study showed both increased connectivity between the precuneus and regions in the dorsal attention networks and decreased connectivity with prefrontal and visual cortices in DLB compared to the AD group [46]. Another data-driven independent component analysis study demonstrated increased connectivity in the default mode network in DLB compared to AD [47]. This finding contrasts with reported connectivity dysfunctions between the posterior and anterior portions of the default mode network in AD [48]. Kenny and colleagues [49] found greater connectivity between the putamen and frontal, temporal and parietal regions in DLB patients compared with AD patients, and argued that this might be associated with Parkinsonian features in DLB. Consistent with relative preservation of memory function in DLB compared with AD, hippocampal connectivity was not found to be different in DLB compared to healthy controls. Conversely, the left hippocampal connectivity was greater in AD compared with controls, which might reflect potential compensatory mechanisms.

Given the high prevalence of visuoperceptual impairments in DLB patients, task-based fMRI studies have investigated the functional integrity of the visual system in DLB. A task-based fMRI study involving visual presentations of colour, face and motion paradigms found greater activation in the superior temporal sulcus in DLB compared to AD during the motor part of the tasks [50]. Another task-based fMRI study did not find any significant differences in functional response between DLB and healthy controls to checkerboard, objects or motion stimuli in V1 and V2/V3, suggesting a relative preservation of function in lower visual areas. Interestingly, ROI analysis revealed decreased V5/MT (middle temporal) activation in response to motion stimuli in the DLB group [51]. Whether these abnormalities at higher levels of the visual system contribute towards the characteristic visuoperceptual impairment and visual hallucinations needs further empirical evidence. In summary, these results seem to indicate that functional abnormalities in DLB affect the visual association areas rather than the primary visual cortex.

Diffusion tensor imaging (DTI) provides in vivo information about white matter microstructural integrity by utilising the anisotropic nature of diffusion in neuronal white matter tracts [52]. White matter diffusion characteristics are commonly assessed by means of mean diffusivity (MD) and fractional anisotropy (FA). MD increases with the degeneration of structural barriers that normally restrict the Brownian motion of water molecules, and reductions in FA occur as diffusion consequently becomes less directionally oriented.

DTI studies in AD have consistently found elevated MD in the hippocampus and decreased FA in the main limbic pathways. Much less is known about DTI changes in DLB. Previous DTI studies on DLB used ROI-based or voxel-based techniques with considerable variability in their findings. Some studies have reported diffusion abnormalities of the corpus callosum and the frontal, parietal, occipital, and, to a lesser extent, temporal white matter when compared to controls, while other studies have found very little change in DTI parameters compared to controls and AD patients [53, 54]. The modest involvement of the temporal lobe is consistent with the relative preservation of global neuropsychological measures and memory domain in DLB compared to AD. Two DTI studies in DLB patients have also identified white matter alterations in the longitudinal fasciculus [54, 55]. Considering the significant role of longitudinal fasciculus in ventral visual pathway, these abnormalities could be associated with visuospatial impairment and visual hallucinations in DLB patients. In addition, elevated MD in the amygdala was also found in DLB, which was associated with Unified Parkinson’s Disease Rating Scale scores [54]. However, AD was not accompanied by loss of GM in this brain region, implying a different pathologic mechanism, such as vacuolisation. However, no changes in diffusivity measures were identified between DLB and AD [54]. On the contrary, a recent DTI study demonstrated distinct patterns of white matter alterations between DLB and AD, with a more focal posterior predominance of FA change in DLB (parieto-occipital) as opposed to a more diffuse pattern of change in AD. DLB was also associated with reduced FA in the pons and the left thalamus compared to AD [56].

1H magnetic resonance spectroscopy (MRS) measures the peak signals from several different metabolites within a single examination period, and it has increasingly been used in differential diagnosis of dementia through the identification of respective spectroscopic profiles of various dementia subtypes (Table 1). In AD, the metabolite N-acetylaspartate is consistently found to be reduced in temporal lobe tissue (approximately 15%) [57], and is associated with disease severity [58]. In addition, increases in myo-inositol (15%) are also commonly reported [57]. In comparison, DLB is characterised by relatively normal N-acetylaspartate/creatine and myo-inositol levels, suggesting neuronal integrity and a lack of gliosis, respectively [59]. Further research with larger sample sizes will determine the clinical utility of these findings in distinguishing DLB from other types of dementia.

At present, neuroimaging findings in dementia are generally derived from group-level analyses. While these have enriched our understanding about the neurobiological differences between DLB and other dementia types, the diagnostic value of most neuroimaging methods is still limited due to the lack of sensitivity and specificity when applied at the individual subject level. As such, there has been increased emphasis on maximizing the utility of more advanced data analysis methods to bridge the gap between the basic research and clinical practice. In this regard, novel machine learning techniques have been developed to allow individual classification of patients. Machine learning techniques extract features from neuroimaging data and construct models of different dementia types and healthy population. These models have been used to differentiate AD from controls, mild cognitive impairment, and frontotemporal lobar degeneration [60], while the potential of this approach in classification among dementia types has been highlighted by a recent autopsy-based machine-learning study by Vemuri and colleagues [61] involving pathologically confirmed dementia patients (AD, DLB, and frontotemporal lobar degeneration). Besides the advent of machine learning techniques, multimodal neuroimaging studies have become increasingly popular as researchers are recognising the benefits from integrating more than one imaging modality (for example, jointly analysing brain volume with perfusion or white matter changes). Recent multimodality strategies have shown superior predictive power than using any single modality domain in the diagnosis of AD [62–64], and distinguishing AD from frontotemporal lobar degeneration [65] and mild cognitive impairment [66]. Similar approaches have recently been applied also to the differentiation of DLB from other conditions. Goto and colleagues [67] integrated MRI striatal volumetric data with occipital perfusion SPECT to distinguish patients with mild DLB from patients with mild AD with high sensitivity and specificity. Kantarci and colleagues [68] achieved increased accuracy (98%) for distinguishing DLB from AD by combining information from occipital FDG uptake, global Pittsburgh compound B retention, and hippocampal volume. Given the multifactorial nature for pathological involvement in patients with DLB, differentiation between DLB and other dementia types will be aided by combining imaging modalities that are sensitive to different components of the disease process.

Longitudinal analyses of neuroimaging data will also provide information about the trajectory of the disease and its underlying neurobiological changes. By allowing the assessment of brain changes over time in individual subjects using multiple serial MRI scans, longitudinal designs have the advantage of reducing within-subject variation. Different rates of cerebral atrophy in dementia types could also inform differential diagnosis. In fact, DLB patients showed a much slower rate of global atrophy (0.4%) compared to AD patients (1.1%) [69].

This review summarises the current imaging literature of DLB in the context of its differentiation from other causes of dementia, discusses the increasingly important role of imaging biomarkers in differential diagnosis, and outlines promising areas for future research. Collectively, the imaging findings have yielded important insights into the underlying pathophysiology of this condition while showing potential promise in improving clinical differentiation of DLB from other types of dementia.

To date, the marked reduction of dopaminergic activity in the basal ganglia is the most characteristic imaging finding of DLB, and has been considered as a significant step in aiding the clinical diagnosis of DLB. Cerebral perfusion studies have also revealed a distinctive pattern of deficits in parietal and occipital regions. At the structural level, another robust observation concerns the relative preservation of MTL in DLB that is consistent with preserved memory functions. Other techniques such as DTI, MRS, and fMRI coupled with novel analytical approaches have also revealed information for differential diagnosis of DLB.

While the field has made substantial progress in delineating the imaging characteristics associated with dementia subtypes, the ability to detect structural patterns that enable accurate prediction of diagnosis for specific individuals ultimately determines the clinical value of MRI and the measurements obtained from it. The reliable application of these methods in routine radiological practice may be facilitated by non-expert-dependent, automated methods of analysis.

In terms of practical considerations, SPECT and PET are both generally well tolerated investigations with few contra-indications, though both involve radiation exposure, limiting the number that any one subject should have. In most countries, SPECT is more widely available than PET. MRI is now widely accessible but more unsuitable for those with claustrophobia, pacemakers or metallic implants in or around the head. fMRI requires specialist stimulus presentation, which limits its applicability to research studies. In addition, some computerized paradigms in task-based fMRI might not be suitable for elderly dementia patients. Other modalities, particularly DTI and MRS, require extensive post-processing analysis time, also potentially limiting their routine application for clinical use.

It is worth mentioning the limitations of the current research. At present, the overwhelming majority of neuroimaging studies in DLB are cross-sectional, relatively small in size, and in participants in established stages of the disease. Therefore, larger prospective longitudinal studies are warranted to confirm the utility of many imaging techniques and monitor disease progression in early disease stages as well as at risk individuals and patients with mild cognitive impairment. Furthermore, studies involving multimodal neuroimaging data and larger cohorts are likely to make novel contributions in evaluating the utility of combined biomarkers in DLB.

This article is part of a series on Lewy Body Dementia, edited by Ian McKeith and James Galvin. Other articles in this series can be found at http://alzres.com/series/LewyBodyDementia.