![[2508.06118] Ensemble-based graph representation of fMRI data for cognitive brain state classification](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2508.06118] Ensemble-based graph representation of fMRI data for cognitive brain state classification
Summary
This article presents an ensemble-based graph representation method for classifying cognitive brain states using fMRI data, achieving high accuracy in various paradigms.
Why It Matters
Understanding cognitive brain states is crucial for neuroscience and clinical applications. This research offers a novel approach that enhances classification accuracy and interpretability, potentially improving diagnostic tools and therapeutic strategies in mental health.
Key Takeaways
- Proposes a new ensemble-based graph representation for fMRI data.
- Achieves binary classification accuracies of 97.07-99.74% across tasks.
- Outperforms conventional correlation graphs in classification accuracy.
- Supports interpretability at both connection and region levels.
- Can be extended to multiclass decoding and other neuroimaging modalities.
Quantitative Biology > Neurons and Cognition arXiv:2508.06118 (q-bio) [Submitted on 8 Aug 2025 (v1), last revised 20 Feb 2026 (this version, v2)] Title:Ensemble-based graph representation of fMRI data for cognitive brain state classification Authors:Daniil Vlasenko, Vadim Ushakov, Alexey Zaikin, Denis Zakharov View a PDF of the paper titled Ensemble-based graph representation of fMRI data for cognitive brain state classification, by Daniil Vlasenko and 3 other authors View PDF HTML (experimental) 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 % ...
