Cortical Surface-Informed Volumetric Spatial Smoothing of fMRI Data via Graph Signal Processing

Abstract

Conventionally, as a preprocessing step, functional MRI (fMRI) data are spatially smoothed before further analysis, be it for activation mapping on task-based fMRI or functional connectivity analysis on resting-state fMRI data. When images are smoothed volumetrically, however, isotropic Gaussian kernels are generally used, which do not adapt to the underlying brain structure. Alternatively, cortical surface smoothing procedures provide the benefit of adapting the smoothing process to the underlying morphology, but require projecting volumetric data on to the surface. In this paper, leveraging principles from graph signal processing, we propose a volumetric spatial smoothing method that takes advantage of the gray-white and pial cortical surfaces, and as such, adapts the filtering process to the underlying morphological details at each point in the cortex.

Voxel-Wise Brain Graphs From Diffusion MRI: Intrinsic Eigenspace Dimensionality and Application to Functional MRI

“…The proposed voxel-wise graphs can be leveraged to perform whole-brain anatomically-informed spatial filtering and interpolation of fMRI data, operations that are inherent within numerous fMRI processing pipelines; e.g. spatial smoothing to enhance whole-brain fMRI activation mapping, as done using tissue-specific designs in gray matter [16], [18] and white matter [17], [31]. Moreover, it yet remains to be studied how functional connectivity (FC) and their associated measures can be extended to accurately integrate structural information.…”

“…GSP has found numerous applications across multiple domains-see e.g. [11] for a recent review, and in particular within neuroimaging, examples include: brain state decoding [12]- [14], brain signal denoising [15], brain activation mapping [16]- [18], source localization [19], diagnosing neuropathology [20], tracking fast spatiotemporal cortical dynamics [21], [22], brain fingerprinting and task decoding [23] via quantifying the degree of coupling between brain function and structure [24], identifying dynamically evolving populations of neurons [25], deciphering signatures of attention switching [26], manifesting white matter pathways that mediate cortical activity [27], and elucidating perturbations of consciousness induced by brain injury or drugs [28], [29].…”

Brain fingerprinting using fMRI spectral signatures on high-resolution cortical graphs

“…edges that connect adjacent voxels that lie on the opposite sides of narrow sulci-are pruned out by using the pial and white surfaces. We thus obtain two graphs per subject, one per hemisphere, entailing a unique harmonic basis [14] using which fMRI data can be characterized [16,17]. 1 https://surfer.nmr.mgh.harvard.edu…”

“…In particular, we show how GSPbased spectral signatures of fMRI data on cerebral hemisphere cortex (CHC) graphs [14,16] can provide informative signatures for subject identification. We decompose resting-state fMRI data using systems of spectral heat kernels [13,17], and, in turn, treat the energy retained in each filtered signal [16] as a feature. Our proposed method does indeed rely on using a brain parcellation [18], albeit not for coarsening cortical activity maps, but rather to derive voxel-wise cortical maps that are more informative than raw fMRI maps, each associated to a specific cortical region [19].…”