Functional Brain Connectivity Patterns Associated with Visual Hallucinations in Dementia with Lewy Bodies

INTRODUCTION

The presence of recurrent, complex visual hallucinations (VH) is among the core clinical features for a clinical diagnosis of dementia with Lewy bodies (DLB) [1]. Different models have been proposed to understand the mechanisms underlying VH. Collerton et al. [2] suggested that recurrent, complex VH result from the combination of impaired top-down (attention) and bottom-up (visual perception) processes. Similarly, another recent model highlights the importance of dysfunctional attentional mechanisms in the development of VH [3]. Accordingly, neuropsychological studies have shown deficits in visual perception [4–7] and visual attention [8–10] in DLB and Parkinson’s disease (PD) patients with VH.

Even though it has been proposed that VH generate from a disrupted organization of functional brain networks [3, 11], there is still insufficient evidence fully corroborating this hypothesis. Resting-state functional MRI studies of VH have been carried out mainly in Parkinson’s disease (PD), and less in DLB [12]. Franciotti et al. [13] focused on the role of the default mode network (DMN) in explaining the occurrence of VH in PD. They tested the hypothesis proposed by previous theoretical articles based on the intrusion of self-referential images into misperceptions due to a dysfunctional DMN [14–16]. Independent component analysis (ICA) revealed increased connectivity in VH patients compared with those without between the superior frontal sulcus bilaterally and parietal regions, and also between contralateral parietal regions. Consistently, in another study, Yao et al. [17] reported increased activity in hallucinating PD patients within the DMN, specifically in fronto-parietal regions. In both studies, patients with and without VH had reduced functional connectivity when compared with control participants [13, 17], suggesting that increased functional connectivity found in patients with VH, compared with those without, may reflect dysfunctional compensatory mechanisms fostering the development of these symptoms. Another recent seed-based resting-state fMRI study of minor hallucinations in PD showed increased connectivity between the posterior cingulate, a central hub of the DMN, and temporal and parietal areas [18]. In another study, Yao et al. [19] investigated the connectivity of the hippocampus, showing increased functional connectivity with fronto-parietal areas, and decreased functional connectivity with occipito-temporal regions.

Generally, the most consistent finding is represented by increased functional connectivity of DMN-related regions in hallucinating patients with PD compared to those without VH, especially with frontal and parietal areas. However, due to the limited number of investigations available, the overall picture of what dysfunction in brain functional connectivity underpins or fosters VH in Lewy body disease is not clear, especially in patients with dementia. The aim of the present pilot study was to explore brain functional connectivity in DLB patients with VH in comparison with non-hallucinating patients. We undertook seed-based analyses to investigate the functional connectivity of regions of interest (ROIs) consistent with the hypothesis of an involvement of fronto-parietal regions, underlying attention deficits, and occipital/occipito-temporal areas, leading to impaired visual perception [9]. Then, we performed ICA to identify intrinsic resting-state functional brain networks, hypothesizing an involvement of the DMN.

MATERIALS AND METHODS

MRI acquisition, pre-processing, and analyses

Three dimensional T1-weighted structural scans were acquired with a 1.5 Tesla Philips Achieva MRI scanner. The sequence used was a Turbo Field Echo and acquisition was sagittal. The following acquisition parameters were applied: repetition time = 7.4 ms, echo delay time =3.4 ms, field of view: 220 mm for DLB and 230 for PD, flip angle: 8°, 160 slices, voxel dimension 1.04×1.04×0.66 mm, gap 0.6 mm. T1-weighted images were segmented to obtain total gray matter, white matter, and cerebrospinal fluid volumes, determined using the MATLAB ‘get_totals’ script (http://www0.cs.ucl.ac.uk/staff/g.ridgway/vbm/get_totals.m) from each image in native space. Total intracranial volume (TIV) was calculated for each patient by summing gray matter, white matter, and cerebrospinal fluid total volumes.

Echo planar T2^*-weighted MRI images were acquired on a 1.5 Tesla Philips Achieva MRI scanner using the following parameters: repetition time =2.02 s, echo delay time = 50 ms, acquisition time =2.1 s, flip angle: 90°, voxel dimension: 1.80×1.80×6.00 mm, field of view: 220 mm. Images were acquired in a single run including 250 volumes, with interleaved slice acquisition (21 axial slices per volume).

Seed-based analyses were undertaken to explore the functional connectivity of regions of interest (ROIs) based on the hypothesis of an involvement of attention and perception-related mechanisms in the development of VH. The following ROIs were selected: left and right pulvinar, lateral geniculate body, caudate nucleus, putamen, frontal eye field (BA 8), superior parietal lobule (BA 7), primary visual cortex (BA 17), secondary visual cortex (BA 18), visual associative cortex (BA 19), inferior temporal area (BA 20), occipitotemporal area (BA 37). All seeds are displayed in a supplementary material file (Supplementary Figure 1). These areas were chosen since they are involved in attention and visuoperceptive abilities. In particular, the frontal eye field, the superior parietal lobule, the pulvinar and the striatum (putamen and caudate nucleus) were chosen for their role in goal-directed visual attention [33–36]. The formation of objects visual representation, on the other hand, relies on ventral visual pathways, including primary and associative visual cortices, occipito-temporal cortices, and inferior temporal areas [37, 38]. The lateral geniculate body was chosen as a control nucleus of the thalamus, the hippocampus as a control region involved in cognitive processing, and the primary motor cortex as a control region not involved in high cognitive functioning. Pre-processing and statistical analyses were performed using the CONN toolbox [39]. Sequences were slice-timed, realigned, coregistered to their respective anatomical image, normalized to the MNI space, and smoothed with a 6 mm³ full-width at half-maximum Gaussian kernel. An outlier-detection step was also carried out to tag volumes showing displacement above the 97th percentile. A number of denoising methods were thus applied: outlier censoring via the ‘scrubbing’ approach, regressing out the first five principal components of WM and CSF signal (aCompCor [40]), and regressing out a total of 24 motion parameters. Acquisitions were also band-pass filtered (0.01 –0.1 Hz). Individual maps of functional connectivity were calculated for each seed.

OPEN IN VIEWER

Fig. 1

The six brain large-scale brain functional networks investigated in this study, specifically: a) DMN, b) salience, c) fronto-parietal, d) occipital, e) sensory-motor, and f) cerebellar networks.

Large-scale functional networks of brain regions were identified by means of ICA. Pre-processing and statistical analyses were undertaken using SPM 12 running on MATLAB R2014a (v8.3). The pre-processing routine described above was relaunched prior to ICA, and further diagnostics were carried out. Linear and rotational parameters of head motion were estimated for each subject, plotted and inspected to identify potentially excessive head motion (exceeding 1.5 mm and 3° for linear and rotational movement, respectively) [41]. No acquisition was found in breach of these limits and no participant was thus excluded at this stage. Framewise and absolute displacement were also calculated in order to characterize volume-to-volume motion and check whether there were between-group differences in these aspects of motion parameters. Framewise displacement refers to the movement of the head from one volume to the next [42]. It was calculated as the sum of the absolute values of the difference of realignment estimates for each volume compared to the previous one. Absolute displacement refers to the movement of the head from the original position, which was calculated for each volume for both rotation and translation. Translation was calculated as the sum of the absolute values of the X, Y, and Z estimates, while rotation was the sum of the absolute values in pitch, yaw, and roll for each volume [42]. Finally, the REST toolbox (http://www.restfmri.net) was used to apply a band-pass temporal filter, and, compared to seed-based analyses, low and high-pass filters were set to define an analogous yet slightly larger bandwidth (0.008 –0.1 Hz). A group ICA was performed with the GIFT toolbox (v1.3i; mialab.mrn.org/software/gift), allowing the identification of brain networks. This technique also allows for the separation of signal and noise-based components, including those resulting from in-scanner motion [43]. The Infomax algorithm was used and the number of components set at 20 following published recommendations [44]. Following visual inspection of all estimated components, and agreement between two independent raters, six functional brain networks of interest were identified, specifically the salience, occipital, sensory-motor, cerebellar, DMN and fronto-parietal networks. All networks are displayed in Fig. 1. Three-dimensional maps of z scores for each network were extracted for each participant, and used for inferential models, aiming at investigating between-group differences, controlling for TIV and age. An uncorrected threshold of p < 0.001 was set to display significant results. Only peaks surviving a cluster-level threshold of p < 0.05 Family-Wise Error corrected for multiple comparisons are reported in the results section.

Montreal Neurological Institute (MNI) coordinates were converted into Talairach coordinates using GingerALE, v2.3.6 (http://www.brainmap.org/ale/). Then, the Talairach Client, v2.4.3 (http://www.talairach.org/client.html) was used to determine brain region labels for each significant cluster. In case of uncertain identification of coordinates labels, some peaks and sub-peaks were further checked manually on the 1988 Talairach atlas [45].

RESULTS

Demographic, clinical, and neuropsychological features are reported in Tables 1 and 2. Briefly, patients with and without VH did not differ for any demographic, clinical or neuropsychological characteristics, except for the NPI total score (p = 0.01) and the presence of rapid eye movement sleep behavior disorder, which was marginally significant (p = 0.05). There were no significant differences between patients with and without VH in framewise displacement (p = 0.46), absolute displacement translation (p = 0.46) and rotation (p = 0.29).

Seed based analyses yielded regions of significantly decreased and increased functional connectivity in DLB VH patients as opposed to patients without hallucinations (Table 3, Figs. 2 and 3). In particular, the seeds located in the left geniculate nucleus, left superior parietal lobule, and right putamen showed decreased functional connectivity with the medial frontal gyrus bilaterally. The left primary visual cortex presented regions of decreased connectivity with the cerebellum, while the right primary visual cortex had significantly lower connectivity with temporo-parietal regions. Increased functional connectivity was found between the right lateral geniculate nucleus and temporo-parietal regions, between the right secondary visual cortex and the postcentral and precentral gyri, between the right superior parietal lobule and the cerebellum, and between the right putamen and the cingulate gyrus.

OPEN IN VIEWER

Fig. 2

Regions of decreased functional connectivity in DLB patients with VH compared to those without obtained in the analyses of the following seeds: a) right putamen; b) left superior parietal lobule (BA 7); c) left lateral geniculate nucleus; d) right primary visual cortex (BA 17); e) right primary visual cortex (BA 17); f) left primary visual cortex (BA 17). The color bar indicates the z scores with the cluster-level threshold of p < 0.05 Family-Wise Error corrected for multiple comparisons with total intracranial volume and age as covariates of no interest.

OPEN IN VIEWER

Fig. 3

Regions of increased functional connectivity in DLB patients with VH compared to those without obtained in the analyses of the following seeds: a) right superior parietal lobule (BA 7); b) right secondary visual cortex (BA 18); c) right lateral geniculate nucleus; d) right putamen. The color bar indicates the z scores with the cluster-level threshold of p < 0.05 Family-Wise Error corrected for multiple comparisons with total intracranial volume and age as covariates of no interest.

From the 20 components estimated by the ICA, six large-scale functional brain networks were identified, namely the salience, occipital, sensory-motor, cerebellar, DMN and fronto-parietal networks (Fig. 1). Only the statistical analysis on the DMN yielded regions of significant decreased and increased functional connectivity (Table 3 and Fig. 4). The DMN was identified as showing synchronous BOLD signal in the following regions: posterior cingulate, inferior parietal lobule, lateral temporal and medial prefrontal cortices [46]. Patients with VH showed a cluster of decreased functional connectivity with the DMN, which was located in the left insula. Increased functional connectivity was found with parietal regions, namely the left inferior parietal lobule and supramarginal gyrus (Table 3 and Fig. 4).

OPEN IN VIEWER

Fig. 4

Regions of decreased and increased functional connectivity within the DMN in DLB patients with VH compared with those without. The color bar indicates the z scores with a cluster-level threshold of p < 0.05 Family-Wise Error corrected for multiple comparisons with total intracranial volume and age as covariates of no interest.

DISCUSSION

The present pilot study provides evidence of both increased and decreased functional connectivity in DLB patients with VH. The lateral geniculate nucleus, the superior parietal lobule and the putamen all showed decreased connectivity with the medial frontal gyrus bilaterally. The top-down control of spatial attention has been shown to rely on a large-scale network, including three primary cortical hubs located in the parietal, frontal and cingulate cortices [34]. These regions are directly interconnected with each other, and indirectly through subcortical hubs in the striatum and the thalamus [34]. Reduced gray matter volume has been found in DLB patients with VH [9, 47], and altered frontal and parietal metabolism/perfusion has been reported [48]. Furthermore, high density of Lewy body pathology has been shown in frontal and anterior cingulate areas in patients with VH, and alterations of the cholinergic receptors within the anterior cingulate have been related to VH in DLB [49, 50]. Cholinergic dysfunction has been proposed to foster the development of VH [51], and its treatment has been shown to ameliorate this symptom, as well as attention deficits [52]. Alterations in the functional connectivity of medial frontal areas might be related to the more severe attention deficits found in DLB and PD patients with VH [8–10], which might constitute a vulnerability to the symptomatology. Collerton et al. [2] proposed an involvement of visual processing impairment, in addition to attention deficits. Occipital and occipito-temporal hypometabolism/perfusion has been repeatedly found in DLB and PD with VH, suggesting a contribution of ventral visual pathways [48]. In the present study, however, visual areas showed a pattern of both decreased and increased functional connectivity. Future studies are needed to understand the significance of this bidirectional pattern of functional connectivity, and the role of such alterations in the genesis and persistence of VH in this disease.

ICA revealed differences between hallucinating and non-hallucinating patients only in the DMN. Specifically, higher functional connectivity was detected in hallucinating patients with the right inferior parietal lobule, while it was lower with the left insula. The inferior parietal lobule is a core region of the DMN [46]. Shine et al. [3] proposed that VH may be the result of disrupted engagement of attention networks, specifically overactivity of the DMN and the ventral attention network (VAN), and inappropriate engagement of the dorsal attention network (DAN). Therefore, overactivity of the DMN might reflect excessive self-referential internal processing, that may form the basis for the emergence of false images [3]. Within this framework, VH might generate from dysfunctional top-down control mechanisms of medial frontal areas over visual areas that might foster the generation of false images arising from an increased activity of the DMN. This speculative interpretation, however, demands further investigation to shed light on the mechanisms behind the pattern of functional connectivity associated with VH in DLB. The insula, on the other hand, is a central hub of the VAN, which has been proposed to coordinate the activity of the DMN and of the goal-directed DAN network [3]. The lower functional connectivity between the DMN and the insula detected in the present study appears to be in contrast with previous findings showing higher connectivity of the VAN associated with visual misperceptions in patients with hallucinations [12]. In this context, the role of the VAN in the development of VH and the direction of its impaired functional connectivity in relation to other attention networks require further investigation.

The main limitation of the present study is represented by the small sample size. Although the findings survived a Family-Wise Error corrected cluster-level p < 0.05, larger samples are needed to corroborate the hypotheses put forward by the present investigation. Thus, the results should be interpreted with caution, and can only be considered as a reference pilot study for future investigations. Another limitation is related to the use of the NPI to assess the presence, severity and frequency of VH. Although the NPI has been used by most studies in the literature, it is not sufficiently specific to detect VH-related phenomenological features, which might be helpful in clarifying the mechanisms involved [12].

VH are severe and disabling symptoms frequently observed in DLB. They have a deleterious impact on the overall quality of life of patients and lead to significantly higher caregiver distress [15, 53]. However, there is currently no effective targeted treatment for these symptoms, problem that is worsened further by the lack of evidence-based interventions [53, 54]. In this context, a better understanding of the functional brain features underlying this symptomatology may aid the detection of symptom-specific biomarkers, which may be helpful in the development of new treatment targets. In the present study, functional connectivity analyses suggest dysfunctional top-down and bottom-up processes and DMN-related alterations in DLB patients with VH. This impairment might foster the generation of false visual images that are misinterpreted, ultimately resulting in VH.