Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Ensemble-based graph representation of fMRI data for cognitive brain state classification

Daniil Vlasenko, Vadim Ushakov, Alexey Zaikin, Denis Zakharov

TL;DR

An ensemble-based graph representation in which each edge weight encodes state evidence as the difference between posterior probabilities of two states, estimated by an ensemble of edge-wise probabilistic classifiers from simple pairwise time-series features is proposed.

Abstract

fMRI is a non-invasive technique for investigating brain activity, offering high-resolution insights into neural processes. Understanding and decoding cognitive brain states from fMRI depends on how functional interactions are represented. We propose an ensemble-based graph representation in which each edge weight encodes state evidence as the difference between posterior probabilities of two states, estimated by an ensemble of edge-wise probabilistic classifiers from simple pairwise time-series features. We evaluate the method on seven task-fMRI paradigms from the Human Connectome Project, performing binary classification within each paradigm. Using compact node summaries (mean incident edge weights) and logistic regression, we obtain average accuracies of 97.07-99.74 %. We further compare ensemble graphs with conventional correlation graphs using the same graph neural network classifier; ensemble graphs consistently yield higher accuracy (88.00-99.42 % vs 61.86-97.94 % across tasks). Because edge weights have a probabilistic, state-oriented interpretation, the representation supports connection- and region-level interpretability and can be extended to multiclass decoding, regression, other neuroimaging modalities, and clinical classification.

Ensemble-based graph representation of fMRI data for cognitive brain state classification

TL;DR

An ensemble-based graph representation in which each edge weight encodes state evidence as the difference between posterior probabilities of two states, estimated by an ensemble of edge-wise probabilistic classifiers from simple pairwise time-series features is proposed.

Abstract

fMRI is a non-invasive technique for investigating brain activity, offering high-resolution insights into neural processes. Understanding and decoding cognitive brain states from fMRI depends on how functional interactions are represented. We propose an ensemble-based graph representation in which each edge weight encodes state evidence as the difference between posterior probabilities of two states, estimated by an ensemble of edge-wise probabilistic classifiers from simple pairwise time-series features. We evaluate the method on seven task-fMRI paradigms from the Human Connectome Project, performing binary classification within each paradigm. Using compact node summaries (mean incident edge weights) and logistic regression, we obtain average accuracies of 97.07-99.74 %. We further compare ensemble graphs with conventional correlation graphs using the same graph neural network classifier; ensemble graphs consistently yield higher accuracy (88.00-99.42 % vs 61.86-97.94 % across tasks). Because edge weights have a probabilistic, state-oriented interpretation, the representation supports connection- and region-level interpretability and can be extended to multiclass decoding, regression, other neuroimaging modalities, and clinical classification.

Paper Structure

This paper contains 23 sections, 2 equations, 5 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Results
  3. Discussion
  4. Methods
  5. Author contributions statement
  6. Acknowledgements
  7. Data availability
  8. Additional information

Figures (5)

  • Figure 1: Pipeline of ensemble graph construction and classification.
  • Figure 2: Architecture of the GNN used to compare the efficiency of ensemble and correlation graphs in a classification task.
  • Figure 3: Bootstrap percentile-t confidence intervals for run-level metric differences between ensemble-based and correlation-based graphs.
  • Figure 4: Averaged edge weight matrices of ensemble graphs for "story" (left) and "math" (right) brain states in the language processing experiment.
  • Figure 5: Application of the Fruchterman-Reingold force-directed algorithm to the averaged edge weight matrices of ensemble graphs for "story" (left) and "math" (right) brain states in the language processing experiment.