Surface-based Neuroimaging Analyses

Introduction

Neuroimaging data is commonly represented in volumetric space, as brain images are stored as 3D grids of voxels. Surface-based analysis offers a complementary approach that represents the brain's cortex as a 2D sheet, which can be better suited for data visualization and capturing brain topology.

This page provides a high-level overview of working with surface neuroimaging data where we will cover:

• How the cortical surface is represented geometrically
• The GIFTI and CIFTI file formats
• Connectome Workbench for processing and visualizing surface data
• Working with surface data in Python and R

Surface Meshes

The cortical surface is a highly folded sheet of gray matter, such that distances in voxel space may not correspond well with geodesic distance (i.e., the distance along the surface).

• A sulcal bank and the facing gyral wall may be adjacent in voxel space but topologically distant on the surface
• To properly capture cortical topology, it is common to represent the cortical surface as a triangular mesh.

Vertices, Edges, and Faces

A surface mesh is defined by:

• Vertices: a set of $N$ points in 3D space, each with coordinates $(x, y, z)$
• Faces (triangles): triplets of vertex indices that define the vertices that are connected by edges to form a triangle

Common surface meshes include those from the Human Connectome Project (HCP) that contain 32k (standard) or 164k (high-resolution) vertices per hemisphere in fsLR space.

Surface Types

During preprocessing, multiple surface types are generated for each hemisphere:

Metric data (e.g., cortical thickness, fMRI activation) is associated with vertices on the midthickness surface and can be projected onto any surface type for visualization.

File Formats: GIFTI and CIFTI

GIFTI

GIFTI (Geometry Format for the Neuroimaging Informatics Technology Initiative) is a standard format for storing surface geometry and surface-mapped data. A GIFTI file (.gii) contains one or more data arrays, each tagged with an intent code indicating what the data represents.

Common GIFTI file types:

A .surf.gii file stores the mesh geometry of vertex coordinates and triangle indices for one hemisphere. Typically, a .func.gii or .shape.gii file stores one or more scalar maps aligned to the same vertex set. See the GIFTI specification for more details on GIFTI file specifications.

CIFTI

CIFTI (Connectivity Informatics Technology Initiative) extends previous formats to represent data jointly across surface vertices and subcortical voxels.

Some common CIFTI file types include:

The term grayordinate refers to a single unit in the CIFTI brain format, either a cortical surface vertex or a subcortical gray matter voxel. A standard 91k dense file contains ~91,000 grayordinates (59,412 cortical vertices + ~31,000 subcortical voxels).

Connectome Workbench

Connectome Workbench is a useful tool for working with surface and CIFTI data. It provides:

• wb_view: a GUI for interactive visualization
• wb_command: a command-line tool with numerous operations for processing surface and CIFTI files

Installation

Connectome Workbench can be downloaded from the HCP website

Visualizing Surface Data with wb_view

wb_view is a powerful GUI for visualizing surface and CIFTI data. It supports simultaneous display of multiple surfaces, overlays, and volume slices.

Loading Data in the GUI

1. Open files: Use File → Open File to load .surf.gii file that encodes the geometry of the surface.
2. Overlay toolbox: Add overlays by opening .shape.gii or .dscalar.nii files in the same atlas space as the .surf.gii file. Adjust the presentation of the overlays by using the overlay toolbox at the bottom of wb_view.

See the Workbench Connectome tutorial for much more information on how to use wb_view.

Common Workflows

wb_command is a command-line interface that can be used for numerous pre-processing steps when working with surface data. Examples include:

• Smoothing a metric file using wb_command -metric-smoothing
• Separating a CIFTI file into two hemispheres with wb_command -cifti-separate
• Resampling to a different resolution with wb_command -metric-resample

The wb_command reference contains documentation for the many use cases of wb_command.

Working with Surface Data in Python

nibabel

nibabel is the main Python library for reading neuroimaging file formats, including GIFTI and CIFTI.

import nibabel as nib
import numpy as np

# Load a GIFTI surface geometry file
surf = nib.load("sub-001.L.midthickness.32k_fs_LR.surf.gii")
vertices = surf.darrays[0].data 
faces    = surf.darrays[1].data 

# Load a GIFTI metric (scalar) file
metric = nib.load("sub-001.L.thickness.shape.gii")
thickness = metric.darrays[0].data

# Load a CIFTI scalar map
img = nib.load("sub-001.dscalar.nii")

Neuromaps

neuromaps provides tools for fetching and transforming brain annotation maps across different spaces, with strong CIFTI/GIFTI support:

from neuromaps import datasets, transforms

# Fetch a reference annotation
annotation = datasets.fetch_annotation(source='abagen', desc='genepc1')

# Resample between surface spaces (fsaverage → fsLR)
resampled = transforms.fsaverage_to_fslr(annotation, target_density='32k')

Working with Surface Data in R

ciftiTools

ciftiTools is the most widely used R package for reading, writing, and visualizing CIFTI files.

Setup

#| eval: false
install.packages("ciftiTools")
library(ciftiTools)
ciftiTools.setOption("wb_path", "/path/to/workbench/bin_macosx64/wb_command")

Reading and exploring CIFTI files

#| eval: false
# Read a dense timeseries
xii <- read_cifti("sub-01_bold.dtseries.nii")
# Access left cortex data only
lh_data <- xii$data$cortex_lh   
# Resample to a coarser resolution
xii_32k <- resample_cifti(xii, resamp_res = 32492)

For more information on how to use the ciftiTools package for surface-based analysis, see the ciftiTools website.

gifti (R package)

The gifti package provides basic support for analyzing GIFTI files in R:

```r
#| eval: false
install.packages("gifti")
library(gifti)

# Read a surface geometry file
surf <- read_gifti("sub-01.L.midthickness.32k_fs_LR.surf.gii")
vertices <- surf$data$pointset 
faces    <- surf$data$triangle 

# Read a metric file
metric <- read_gifti("sub-01.L.thickness.shape.gii")
thickness <- metric$data[[1]]
```

Learning Resources

• Andy's Brain Book: FreeSurfer Tutorial: accessible intro to surface data in FreeSurfer
• Connectome Workbench Tutorial (HCP): comprehensive guide to wb_view
• nibabel CIFTI documentation
• neuromaps documentation: tutorials on surface annotation, transformation, and statistical comparison
• ciftiTools vignettes: examples of common pipelines using ciftiTools workflow in R