Functional Mesh Learning for pattern analysis of cognitive processes

Abstract

We propose a statistical learning model for classifying cognitive processes based on distributed patterns of neural activation in the brain, acquired via functional magnetic resonance imaging (fMRI). In the proposed learning machine, local meshes are formed around each voxel. The distance between voxels in the mesh is determined by using functional neighborhood concept. In order to define functional neighborhood, the similarities between the time series recorded for voxels are measured and functional connectivity matrices are constructed. Then, the local mesh for each voxel is formed by including the functionally closest neighboring voxels in the mesh. The relationship between the voxels within a mesh is estimated by using a linear regression model. These relationship vectors, called Functional Connectivity aware Local Relational Features (FC-LRF) are then used to train a statistical learning machine. The proposed method was tested on a recognition memory experiment, including data pertaining to encoding and retrieval of words belonging to ten different semantic categories. Two popular classifiers, namely k-Nearest Neighbor and Support Vector Machine, are trained in order to predict the semantic category of the item being retrieved, based on activation patterns during encoding. The classification performance of the Functional Mesh Learning model, which range in 62-68% is superior to the classical multi-voxel pattern analysis (MVPA) methods, which range in 40-48%, for ten semantic categories.

A Comparison of Functional Networks Derived From Representational Similarity, Functional Connectivity, and Univariate Analyses

“…Specifically, networks built from second-order RSA and multivariate functional connectivity could also be compared (Coutanche and Thompson-Schill, 2014;Anzellotti and Coutanche, 2018). Other analyses than RSA exist that can pick up on more complex spatial relations between ROIs (e.g., Haxby et al, 2011;Ozay et al, 2012;Firat et al, 2013;Onal et al, 2017). Although RSA is one of the most commonly used approaches in cognitive neuroscience in comparing neural spaces across different brain regions, uncovering the complexity of the spatial relationships captured through more advanced approaches may help to compare the three methodologies in depth.…”

Modeling the Brain Connectivity for Pattern Analysis

“…In order to form the local meshes with respect to spatially p-nearest neighborhood, p number of voxels having the smallest Euclidean distance between the coordinates of the seed voxel and its neighbors are selected in the mesh of the voxels, {v(t i ,s k )} p k=1 [11]. On the other hand Firat et al [12] defined the p-nearest neighborhood functionally, where p-nearest neighbors are selected based on the functional connectivity between the seed voxel and the surrounding voxels. In this approach p-nearest neighbors {v(t i ,s k )} p k=1 of a seed voxel are the ones where the Pearson correlations with the seed voxel are the highest p voxels.…”

“…These features, called Mesh Arc Descriptors, are shown to discriminate the cognitive states better than voxel intensities. In a further study, Firat et al [12] model the relationships among functionally close neighbors and use the arc weights of functionally connected voxels to train a classifier. These studies show a significant improvement on the performance of the algorithms developed for cognitive state classification.…”

Analyzing Complex Problem Solving by Dynamic Brain Networks

“…Then, we solved the systems of linear equations using various regression techniques. Our team applied Levinson-Durbin recursion in order to estimate the edge weights of each local star mesh, where the nodes are the neighboring regions of the seed brain region (Fırat et al, 2013;Alchihabi et al, 2018). We also used ridge regression to estimate edge weights while constructing the local mesh networks across windows of time series of fMRI recordings (Onal et al, 2015(Onal et al, , 2017.…”