Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Causality-based Subject and Task Fingerprints using fMRI Time-series Data

Dachuan Song, Li Shen, Duy Duong-Tran, Xuan Wang

TL;DR

This paper addresses identifying individuals and cognitive tasks from fMRI time-series by introducing causality-based fingerprints. It proposes a two-timescale linear state-space model to extract spatio-temporal causal signatures, with subject fingerprints derived from modal decomposition and projection, and task fingerprints via a Graph Neural Network applied to the causal signature matrices. Empirical results on the Human Connectome Project data show that the approach achieves about 80% accuracy for subject identification in resting-state data and up to around 90% average accuracy for task identification across eight tasks, with resting-state achieving 100% for several tasks, outperforming non-causal baselines. The work advances explainable, directional brain interaction representations with potential clinical impact in neurodegenerative disease diagnostics and real-time brain-state monitoring.

Abstract

Recently, there has been a revived interest in system neuroscience causation models due to their unique capability to unravel complex relationships in multi-scale brain networks. In this paper, our goal is to verify the feasibility and effectiveness of using a causality-based approach for fMRI fingerprinting. Specifically, we propose an innovative method that utilizes the causal dynamics activities of the brain to identify the unique cognitive patterns of individuals (e.g., subject fingerprint) and fMRI tasks (e.g., task fingerprint). The key novelty of our approach stems from the development of a two-timescale linear state-space model to extract 'spatio-temporal' (aka causal) signatures from an individual's fMRI time series data. To the best of our knowledge, we pioneer and subsequently quantify, in this paper, the concept of 'causal fingerprint.' Our method is well-separated from other fingerprint studies as we quantify fingerprints from a cause-and-effect perspective, which are then incorporated with a modal decomposition and projection method to perform subject identification and a GNN-based (Graph Neural Network) model to perform task identification. Finally, we show that the experimental results and comparisons with non-causality-based methods demonstrate the effectiveness of the proposed methods. We visualize the obtained causal signatures and discuss their biological relevance in light of the existing understanding of brain functionalities. Collectively, our work paves the way for further studies on causal fingerprints with potential applications in both healthy controls and neurodegenerative diseases.

Causality-based Subject and Task Fingerprints using fMRI Time-series Data

TL;DR

This paper addresses identifying individuals and cognitive tasks from fMRI time-series by introducing causality-based fingerprints. It proposes a two-timescale linear state-space model to extract spatio-temporal causal signatures, with subject fingerprints derived from modal decomposition and projection, and task fingerprints via a Graph Neural Network applied to the causal signature matrices. Empirical results on the Human Connectome Project data show that the approach achieves about 80% accuracy for subject identification in resting-state data and up to around 90% average accuracy for task identification across eight tasks, with resting-state achieving 100% for several tasks, outperforming non-causal baselines. The work advances explainable, directional brain interaction representations with potential clinical impact in neurodegenerative disease diagnostics and real-time brain-state monitoring.

Abstract

Recently, there has been a revived interest in system neuroscience causation models due to their unique capability to unravel complex relationships in multi-scale brain networks. In this paper, our goal is to verify the feasibility and effectiveness of using a causality-based approach for fMRI fingerprinting. Specifically, we propose an innovative method that utilizes the causal dynamics activities of the brain to identify the unique cognitive patterns of individuals (e.g., subject fingerprint) and fMRI tasks (e.g., task fingerprint). The key novelty of our approach stems from the development of a two-timescale linear state-space model to extract 'spatio-temporal' (aka causal) signatures from an individual's fMRI time series data. To the best of our knowledge, we pioneer and subsequently quantify, in this paper, the concept of 'causal fingerprint.' Our method is well-separated from other fingerprint studies as we quantify fingerprints from a cause-and-effect perspective, which are then incorporated with a modal decomposition and projection method to perform subject identification and a GNN-based (Graph Neural Network) model to perform task identification. Finally, we show that the experimental results and comparisons with non-causality-based methods demonstrate the effectiveness of the proposed methods. We visualize the obtained causal signatures and discuss their biological relevance in light of the existing understanding of brain functionalities. Collectively, our work paves the way for further studies on causal fingerprints with potential applications in both healthy controls and neurodegenerative diseases.
Paper Structure (11 sections, 8 equations, 5 figures, 2 tables)

This paper contains 11 sections, 8 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods: Causal Fingerprint
  3. A Two-timescale State-space Model for Causal Dynamics
  4. Subject Causal Fingerprint via Modal Decomposition and Projection
  5. Task Causal Fingerprint via Graph Neural Network
  6. Results
  7. Dataset and Modeling.
  8. Subject Causal Fingerprinting
  9. fMRI Task Causal Fingerprinting.
  10. Visualization and Discussions
  11. Conclusion

Figures (5)

  • Figure 1: (a) Brain Activity (time-series data) captured using Schaefer Parcellation. (b) Causal Dynamics modeling and data-driven parameter identification.
  • Figure 2: Five-Layer GNN for Task Fingerprinting.
  • Figure 3: (a) Functional Areas of the Brain. (b-d) Ten Schaefer Parcellation Areas (blue dots) that are Chosen as System Inputs: Two areas are associated with the prefrontal cortex; Two areas are associated with the pre-motor cortex; Two areas are associated with the somatic cortex; Two areas are associated with the vision cortex; Two areas are associated with the hearing cortex. We assume these are areas likely to 'initiate' certain brain activities.
  • Figure 4: Left: Subject Fingerprinting based on CM+MD&P: The identification accuracy is compared with (black line) using the FC+CoR method. Right: Task Fingerprinting based on CM+GNN: The figure shows the correct/incorrect classification for the same/different fMRI tasks.
  • Figure 5: Visualization of Brain Regions with Strongest Connection Weights, Showing Association Strengths Across Multiple Cognitive Tasks (Importance Scores Indicated by Color-Bar). LH refers to the left hemisphere, and RH refers to the right hemisphere of the brain. Arrows indicate the directed relation about which areas cause impacts to other areas. Note that for ease of visualization, we use two colors and set the threshold to highlight only the most active areas of the brain and explain their directional relations.

Theorems & Definitions (2)

  • definition 1
  • Remark 1