Analysis of resting-state data

To estimate independent healthy RSNs, 44 healthy control participants from the ICICLE and VEEG-Stim studies were selected. All participants were scanned on the same scanner as the participants from the main analysis. Eighteen of the additional HC participants were scanned with a slightly different scanner protocol with a change in the TR to 2072 ms and a change in the voxel size of the resting-state scans to 3 x 3 x 4 mm3. The resting-state data were

preprocessed in the same way as described in Section 4.2.3. Two subjects were excluded because they exceeded the motion exclusion criteria resulting in 42 independent HC

79

participants that were included in the generation of the RSN templates.

The temporally concatenated data from all independent HC participants were subjected to a group-ICA using FSL’s Multivariate Exploratory Linear Optimised Decomposition into Independent Components (MELODIC). To obtain more reliable components, a meta ICA approach was adopted (Biswal et al., 2010; Poppe et al., 2013). To this end, MELODIC was repeated 25 times on randomised subsets of 30 out of the 42 HC participants. Subsequently, a meta ICA run was performed on the concatenated components from all individual ICA runs. A model order of 70 independent components was chosen for the individual as well as the meta ICA as this has been shown to be optimal for assessing disease-related group differences (Abou Elseoud et al., 2011; Dipasquale et al., 2015). To identify reliable components, the spatial correlation of each meta component across the individual ICA runs was calculated and components with a correlation <0.6 across runs were excluded (Cerliani et al., 2015).

Furthermore, the meta ICA procedure was repeated using all HC participants from the main analysis and compared to the components from the independent group to ensure that the selected RSNs were present in both cohorts. All meta ICA components from the independent cohort that survived these reliability checks were visually inspected with respect to their spatial maps (Kelly et al., 2010) and 27 were identified as being of biological interest

according to the previous literature (Agosta et al., 2012; Beckmann et al., 2005; Damoiseaux et al., 2008) (Figure 4.1 and Table 4.1).

Subsequently, within-network connectivity was assessed by running FSL-dual regression with all 27 identified RSNs concatenated in a single 4D image. First, for each participant, the RSN spatial maps were regressed (as spatial regressors in a multiple regression) into the

participant’s 4D dataset, resulting in a subject-specific timeseries, one for each RSN. Second, these timeseries were regressed (as temporal regressors, again in a multiple regression) into the same 4D dataset, resulting in a set of subject-specific spatial maps, one for each RSN. These spatial maps represent the participant’s functional connectivity map for the respective RSN. Group differences in these functional connectivity maps between AD and HC, between DLB and HC, and between DLB and AD were assessed using FSL’s randomise function with 10,000 permutations and family-wise error (FWE) correction for multiple comparisons using threshold-free cluster enhancement (TFCE). Covariates of no interest were included to control for age, sex, and study membership. Additionally, in order to reduce the impact of cortical atrophy differences between the participant groups, grey matter probability maps were also included as voxel-wise regressors in the linear model (Damoiseaux et al., 2012). To

investigate between-network connectivity, the FSLNets package was applied to the subject- specific time series from dual regression (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLNets). Full

80

and partial correlations were calculated between all pairs of RSNs and the resulting

correlation coefficients were converted to z-scores for further analysis. Partial correlations are computed as correlations between two RSNs while controlling for the effect of all other RSNs and are thought to reflect more direct connections (Smith et al., 2011). FSL-randomise with 10,000 permutations was then applied to assess group differences in between-network connectivity including covariates for age, sex, and study membership. Results were FWE- corrected for multiple comparisons.

Table 4.1. List of all resting-state networks (RSNs) included in the analysis. Anatomical

labels refer to bilateral areas if not stated otherwise. Locations of RSNs were estimated from the Harvard-Oxford Cortical and Subcortical Structural Atlases and the Cerebellar Atlas in FSL.

RSN name Brain regions

Lateral sensorimotor network LSMN Pre- and postcentral gyrus

Medial sensorimotor network MSMN Pre- and postcentral gyrus, supplementary motor area

Supplementary motor area network

SMAN Supplementary motor area, precentral gyrus

Left motor network LMN Left post- and precentral gyrus

Right motor network RMN Right post- and precentral gyrus

Basal ganglia network BGN Putamen, caudate

Thalamic network THN Thalamus

Cerebellar network 1 CBN1 Cerebellum crus I, crus II

Cerebellar network 2 CBN2 Cerebellum V, VI

Medial visual network MVN Intracalcarine cortex, supracalcarine cortex, lingual gyrus

Lateral visual network LVN Superior lateral occipital cortex, precuneus

Occipital pole network OPN Occipital pole

Lingual gyrus network LGN Lingual gyrus, intracalcarine cortex

Superior visual network SVN Superior lateral occipital cortex, occipital pole

Temporal network TN Planum temporale, Heschl’s gyrus

Temporal pole network TPN Temporal pole

Insular network 1 ISN1 Insular cortex, frontal operculum cortex

Insular network 2 ISN2 Insular cortex, planum polare

Anterior cingulate network ACN Anterior cingulate cortex

Default mode network 1 DMN1 Precuneus, posterior cingulate cortex

Default mode network 2 DMN2 Precuneus

Default mode network 3 DMN3 Precuneus, superior lateral occipital cortex Supramarginal gyrus network SPGN Supramarginal gyrus

Right fronto-parietal network RFPN Right superior lateral occipital cortex, right angular gyrus, right middle frontal gyrus, left superior lateral occipital cortex

Left fronto-parietal network LFPN Left superior lateral occipital cortex, right angular gyrus, left middle frontal gyrus, right superior lateral occipital cortex

Dorsal attention network DAN Superior parietal lobule, supramarginal gyrus, superior lateral occipital cortex

81

Figure 4.1. Spatial maps of the 27 resting-state networks (RSNs) obtained from the

independent healthy control group. RSN maps are thresholded at 3<z<12. Images are shown in radiological convention, i.e. the left side of the image corresponds to the right hemisphere.

82