Large Scale Functional Connectivity for Brain Decoding

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 method, local meshes are formed around each voxel. The distance between voxels in the mesh is determined by using a functional neighbourhood concept. In order to define the functional neighbourhood, 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 neighbouring 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 knearest neighbour (k-nn) and Support Vector Machine (SVM), 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% − 71% is superior to the classical multi-voxel pattern analysis (MVPA) methods, which range in 40% − 48%, for ten semantic categories.

Modeling the Sequence of Brain Volumes by Local Mesh Models for Brain Decoding

“…In the second case, four classes correspond to kitchen utensil, furniture, fear and disgust, where kitchen utensil and furniture belong to the neutral category, and fear and disgust belong to the emotional category (see Table III). 5) 73( 7) 82( 12) 81( 8) 79( 5) 88( 9) 78( 10) 81( 13) 79( 9) 72( 5) 84( 14) 66( 8) 77.6 SLM 72(5) 75( 5) 78( 12) 75( 6) 78( 5) 74( 6) 88( 5) 78( 11) 81( 11) 82( 14) 81( 15) 70( 11) 66 (15) 76.8 FMM-mean 50(8) 61( 5) 66( 5) 63( 5) 62( 10) 64( 5) 82( 9) 63( 6) 75( 7) 72( 5) 60( 5) 63( 5) 54( 5) 64.2 FMM-peak 66( 15) 67( 9) 69( 9) 78( 14) 72( 12) 63( 11) 76( 6...…”

“…In order to extract the temporal information from the fMRI data, various studies [13], [14], [15], [16], [17] compute the pairwise correlations between the responses of voxels or brain regions. Among these studies, Pantazatos et al [13] model temporal relationships by using pairwise correlations between brain regions as features for brain decoding.…”

“…Also, edge weights of their proposed brain graphs are generally selected using the correlation between pairs of selected voxels, and vertices or edge weights of these brain graphs are used as features for cognitive state classification [18]. Contrary to the coarse-level functional connectivity, Firat et al [15] use pairwise correlations of voxels as features while Baldassano et al [16] extract the functional connectivity structures by modeling pairwise connectivity among voxels. Also, a minimum spanning tree of a graph structure defined using a functional connectivity matrix is employed using all voxels for each stimulus for brain decoding [17].…”