Beyond binary parcellation of the vestibular cortex - A dataset.

The data-set presented in this data article is supplementary to the original publication, doi:10.1016/j.neuroimage.2018.05.018 (Kirsch et al., 2018). Named article describes handedness-dependent organizational patterns of functional subunits within the human vestibular cortical network that were revealed by functional magnetic resonance imaging (fMRI) connectivity parcellation. 60 healthy volunteers (30 left-handed and 30 right-handed) were examined on a 3T MR scanner using resting state fMRI. The multisensory (non-binary) nature of the human (vestibular) cortex was addressed by using masked binary and non-binary variations of independent component analysis (ICA). The data have been made publicly available via github (https://github.com/RainerBoegle/BeyondBinaryParcellationData).

Specifications table

Value of the data

Proposition of a functional connectivity based parcellation (fCBP) approach that addresses the multisensory (non-binary) nature of the human vestibular cortex, here vestibular.

Two variations of independent component analysis (ICA) are used: The traditional (binary) ICA approach, where each functional sub-unit must be spatially distinct (and voxels are forced to choose a sub-unit). And a variation, the multivariate (non-binary) ICA approach where functional-subunits can overlap (and voxels can to be part of multiple sub-units with their various behavioral interpretations).

This non-binary methodical approach might be able to reflect multiple signals at the same spatial location, e.g. in multiple populations of neurons or a single multisensory population.

Data

This data set aims to identify handedness-dependent organizational patterns of functional subunits within the human vestibular cortex whilst addressing its multisensory (non-binary) nature. To that end, 60 healthy volunteers (30 left-handed and 30 right-handed) were analyzed using a masked binary and non-binary fCBP (functional connectivity based parcellation) approach. This mask was data-driven (composed of whole brain independent components) and specific to the vestibular cortical system as the used independent components (ICs) were required to include vestibular reference coordinates derived from two meta-analyses of vestibular neuroimaging experiments pinpointing the vestibular cortex [2], [3].

Experimental design, materials and methods

Age- and gender-matched 30 left-handed (LH; 14 females; aged 20–65 years, mean age 26.1 ± 8.6 years) and 30 right-handed (RH; 17 females; aged 20–67 years, mean age 26.7 ± 8.3 years) healthy volunteers with a verified sound vestibular system (semicircular and otolith function) were examined on a 3T MR scanner (Magnetom Verio, Siemens Healthcare, Erlangen, Germany) using task free resting state functional MRI (fMRI) (Table 1, Table 2A, Table 2B, Table 3A, Table 3B, Table 3C).

Table 1

Overview of used behavioral Interpretations of intrinsic whole brain resting-state networks (RSN).

RSN1-20 was characterized as per Laird et al. [4]. Here, RSN 1-5 were accorded to “emotional and autonomic processes, perception”; RSN 6-9 to “mixture of functions related to motor and visuospatial integration, coordination, and execution”, RSN 10-12 to “Networks related to visual perception”, RSN10-18 to “divergent networks”. (*) RSN19-20 was defined as frequent artifacts. In addition, seven further RSN were defined using anatomical knowledge if they did not fit any of the Laird atlas RSN components. To assign a sound function to these 7 extra RSNs the maximum xyz-MNI-coordinates were entered in the Neurosynth platform (neurosynth.org) with a radius of 4 mm. The most plausible associations given by this automated synthesis with large-scale human functional neuroimaging data [5] were chosen and specified using main concept terms such as „emotional processing“. Each of the 27 RSNs was assigned to a separate color (cp. color scale), which matches the colors used for RSN-affiliations of whole brain IC maps in Fig. 2 of [1].

Abr.: A1 = primary auditory cortex; ACC = anterior cingulate cortex, BA = Broadman areal, FEF = frontal eye fields, IC = independent component; M1 = primary motor cortex; MST = medial superior temporal area; MT = middle temporal area; OFC = orbitofrontal cortex, RSN = resting-state network; S1 = primary somatosensory cortex: SMA = supplementary motor area, V1-5 = primary, secondary and tertiary visual cortices.

Table 2A

Characterization of “asymmetrical and less connected” parcels.

Masked Binary and non-binary fCBP (functional connectivity brain parcellation) resulted in 30 different parcels, which were categorized by means of “spatial symmetry”, “number of parcels to systems correlations” and “predominant anatomical landmark”. This resulted in two different types of parcels: “Asymmetrical and less connected” (Table 2A) and “symmetrical and connected” (Table 2B) voxels (V). Each of the 30 parcels (P) was assigned to a separate color (cp. color scale Fig. 3 of [1]), which was the same in both LH (left-handed) and (RH) right-handed subgroups. “Asymmetrical” parcels were highlighted in grey. Parcels were anatomically characterized using the Harvard- Oxford structural cortical atlas in bold letters [6], [7] and the Jülich histological (cyto- and myeloarchitectonic) atlas in regular letters [8], [9]. Handedness-dependency (+/-) was calculated using a laterality index. If the laterality-index (L-I) per parcel and in between LH and RH changed concordant it was termed handedness-independent (-). An inverse laterality-index was termed handedness-dependent (+).

Table 2B

Characterization of “symmetrical & connected” parcels.

Masked Binary and non-binary fCBP (functional connectivity brain parcellation) resulted in 30 different parcels, which were categorized by means of “spatial symmetry”, “number of parcels to systems correlations” and “predominant anatomical landmark”. This resulted in two different types of parcels: “Asymmetrical and less connected” (Table 2A) and “symmetrical and connected” (Table 2B) voxels (V). Each of the 30 parcels (P) was assigned to a separate color (cp. color scale Fig. 3 of [1]), which was the same in both LH (left-handed) and (RH) right-handed subgroups. “Asymmetrical” parcels were highlighted in grey. Parcels were anatomically characterized using the Harvard- Oxford structural cortical atlas in bold letters [6], [7] and the Jülich histological (cyto- and myeloarchitectonic) atlas in regular letters [8], [9]. Handedness-dependency (+/-) was calculated using a laterality index. If the laterality-index (L-I) per parcel and in between LH and RH changed concordant it was termed handedness-independent (-). An inverse laterality-index was termed handedness-dependent (+).

Table 3A

Characterization of “unique voxels” within parcels, type “asymmetric and less connected”.

Masked non-binary fCBP (functional connectivity based parcellation) enabled the distinction of spatial uniqueness (Table 3A, Table 3B) and commonality (3C) of independent components that form parcels. Analog to Table 2A, Table 2B“unique voxels” (U) were left in the previous categorization in two types of “unique” voxels: Type previously “asymmetric and less connected” (Table 3A, highlighted in grey) and type previously “symmetric and connected” (Table 3B). An inverse laterality-index (L-I) was termed handedness-dependent (+), a concordant laterality-index meant handedness-independency (-). Common” voxels were defined as voxels that overlapped in between parcels. To enable visualization of the “common” voxels, 9 groups of 2-6 spatially similar parcels (“common clusters”; C) were defined and correlated to whole brain RSN.

For a depiction of “unique” voxels please view Fig. 6A in [1], and for the “common clusters” view Fig. 6B in [1]. Each of the 30 parcels (P) was assigned to a separate color (cp. color scale Fig. 3 in [1]), which was also used for the parcel’s “unique voxels”. The colors match between (left-handed) LH and right-handed (RH) subgroups. This color-code can also be seen in Table 3C, where each color represents one of the parcels included in the “common” cluster. U and C were anatomically characterized using the Harvard–Oxford structural cortical atlas in bold letters [6], [7] and the Jülich histological (cyto- and myelo-architectonic) atlas in regular letters [8], [9].

Table 3B

Characterization of “unique voxels” within parcels, type “symmetric and connected”.

Masked non-binary fCBP (functional connectivity based parcellation) enabled the distinction of spatial uniqueness (Table 3A, Table 3B) and commonality (Table 3C) of independent components that form parcels. Analog to Table 2A, Table 2B “unique voxels” (U) were left in the previous categorization in two types of “unique” voxels: Type previously “asymmetric and less connected” (Table 3A, highlighted in grey) and type previously “symmetric and connected” (Table 3B). An inverse laterality-index (L-I) was termed handedness-dependent (+), a concordant laterality-index meant handedness-independency (-). Common” voxels were defined as voxels that overlapped in between parcels. To enable visualization of the “common” voxels, 9 groups of 2-6 spatially similar parcels (“common clusters”; C) were defined and correlated to whole brain RSN.

For a depiction of “unique” voxels please view Fig. 6A in [1], and for the “common clusters” view Fig. 6B in [1]. Each of the 30 parcels (P) was assigned to a separate color (cp. color scale Fig. 3 in [1]), which was also used for the parcel’s “unique voxels”. The colors match between (left-handed) LH and right-handed (RH) subgroups. This color-code can also be seen in Table 3C, where each color represents one of the parcels included in the “common” cluster. U and C were anatomically characterized using the Harvard- Oxford structural cortical atlas in bold letters [6], [7] and the Jülich histological (cyto- and myelo-architectonic) atlas in regular letters [8], [9].

Table 3C

Characterization of “common” clusters (C).

Masked non-binary fCBP (functional connectivity based parcellation) enabled the distinction of spatial uniqueness (Table 3A, Table 3B) and commonality (Table 3C) of independent components that form parcels. Analog to Table 2A, Table 2B “unique voxels” (U) were left in the previous categorization in two types of “unique” voxels: Type previously “asymmetric and less connected” (Table 3A, highlighted in grey) and type previously “symmetric and connected” (Table 3B). An inverse laterality-index (L-I) was termed handedness-dependent (+), a concordant laterality-index meant handedness-independency (-). Common” voxels were defined as voxels that overlapped in between parcels. To enable visualization of the “common” voxels, 9 groups of 2-6 spatially similar parcels (“common clusters”; C) were defined and correlated to whole brain RSN.

For a depiction of “unique” voxels please view Fig. 6A in [1], and for the “common clusters” view Fig. 6B in [1]. Each of the 30 parcels (P) was assigned to a separate color (cp. color scale Fig. 3 in [1]), which was also used for the parcel’s “unique voxels”. The colors match between (left-handed) LH and right-handed (RH) subgroups. This color-code can also be seen in Table 3C, where each color represents one of the parcels included in the “common” cluster. U and C were anatomically characterized using the Harvard–Oxford structural cortical atlas in bold letters [6], [7] and the Jülich histological (cyto- and myelo-architectonic) atlas in regular letters [8], [9].

After Preprocessing, the data were analyzed in four major steps using a functional connectivity based parcellation (fCBP) approach: (1) independent component analysis (ICA) on a whole brain level to identify different resting state networks (RSN); (2) creation of a vestibular informed mask from four whole brain ICs that included reference coordinates of the vestibular network extracted from meta-analyses of vestibular neuroimaging experiments; (3) Re-ICA confined to the vestibular informed mask; (4) cross-correlation of the activated voxels within the vestibular subunits (parcels) to each other (P-to-P) and to the whole-brain RSN (P-to-RSN). For a flowchart of the used functional connectivity based parcellation (fCBP) methods please view Fig. 1 of [1] (Fig. 1, Fig. 2, Fig. 3, Fig. 4).

Fig. 1

Overlay of resulting 27 whole brain resting state networks (RSN). This overlay shows the spatial distribution of the 27 RSN systems. 80 dimensional whole brain ICA was performed on denoised fMRI data (LH and RH combined) using a whole brain mask. Each independent component (IC) was semi-automatically labeled to the 20 resting state network (RSN) atlas proposed by Laird et al. [4]. ICs that did not fit the Laird components (overall 7 of 80 or 8.75% of ICs) were checked visually and assigned to an anatomical label of the “Harvard-Oxford cortical structural atlas”. Here, sound behavioral interpretations to each IC (network) were determined by inserting their maximum xyz-MNI-coordinates in the large-scale, automated synthesis of human functional neuroimaging data platform Neurosynth (neurosynth.org), using a radius of 4 mm [5]. For an overview of these networks view Table 1. For an overview of the 80 dimensional whole brain ICA including their RSN attribution cp Fig. 2 in [1].

Fig. 2

Overview of 30 single parcels resulting from masked binary fCBP. To be able to compare LH and RH parcels we had to find analogous binary parcels between LH and RH. This approach was successful for interhemispheric symmetric parcels, but not for interhemispheric asymmetric parcels. Here, the parcels needed to be spatially flipped (mirrored) to correspond between LH and RH. This was done with respect to the x-axis, i.e. hemisphere-flip in x-direction in MNI-space. RH results are shown on the top row and the LH results in the lower row. Hemisphere-flips are depicted in the middle if necessary. The background colors represent the color of the parcel. The P number is indicated on the bottom right side of each overlap grouping. A more detailed depiction of spatially asymmetric and flipped parcels can be viewed in Fig. 4 of [1].

Fig. 3

P-to-RSN correlation matrix (FDR < 0.01). The x-axis (including colors) indicates the 30 parcels that resulted for LH and RH after masked binary fCBP (cp. Fig.3 in [1]). The third column represents differences between RH and LH. The y-axis (including colors) indicate the assignment to the 27 RSN systems as shown in Table 1. Note, that the number of RSN assignments to each parcel (P-to-RSN) did not differ between LH and RH. However, symmetrical parcels had significantly more RSN assignments than asymmetrical.

Fig. 4

C-to-RSN correlation matrix (FDR < 0.01). The x-axis indicates the 9 common clusters (C) that resulted for LH and RH after masked non-binary fCBP (cp. Fig. 6B in [1]). The third column represents differences between RH and LH. The y-axis (including colors) indicate the assignments of C to the 27 RSN systems as shown in Table 1. Please note, that apart from C1 “posterior insula” and C2 “inferior insula”, common clusters correlated with more than 5 RSN “systems”, which indicate manifold functionality.

All details as well as further explanations of the methods can be viewed in the original publication, https://doi.org/10.1016/j.neuroimage.2018.05.018 [1].

Subject area	Neuroscience, Vestibular system
More specific subject area	Handedness-dependent organizational patterns of (lateralized and non-lateralized) functional subunits within the human vestibular cortical network
Type of data	Tables, figures, text file, data set
How data were acquired	3T Magnetic resonance imaging (MRI) data, 32-channel head coil, T2*-weighted echo-planar imaging (EPI) sequence, T1-weighted magnetization-prepared rapid gradient echo (MP-RAGE) sequence, task-free resting state.
Data format	Analyzed, Nifti *.nii files, MATLAB *.mat files, Portable Network Graphics *.png files, Text *.txt files
Experimental factors	30 healthy right-handed (RH) volunteers and 30 age- and gender-matched healthy left-handed (LH) volunteers with a verified sound vestibular system (semicircular and otolith function)
Experimental features	The multisensory (non-binary) nature of the human (vestibular) cortex was addressed by using binary and non-binary variations of independent component analysis (ICA) to separate its functional subunits.
Data source location	Munich, Germany, Latitude 48°06′22.20″ N, Longitude 11°28′5.99″ E
Data accessibility	The analyzed data are available within this article, the used dataset can be downloaded from the GitHub Link: https://github.com/RainerBoegle/BeyondBinaryParcellationData