swpc-directed-functional-connectivity

name	swpc-directed-functional-connectivity
description	Sliding-window prediction correlation (SWpC) for time-varying directed functional connectivity. Embeds directional LTI models within sliding windows to estimate time-resolved information flow in brain networks, going beyond undirected correlation.

SWpC: Sliding-Window Prediction Correlation for Directed Functional Connectivity

Paper: arXiv:2602.16004 Authors: Nan Xu, Xiaodi Zhang, Wen-Ju Pan, et al. Categories: q-bio.NC, cs.LG Year: 2026

Overview

SWpC estimates time-varying directed functional connectivity in brain networks. Unlike traditional sliding-window correlation (SWC) which captures undirected associations, SWpC resolves directional interactions by embedding a directional LTI model within each sliding window.

Key Concepts

Limitations of SWC

• Captures time-varying undirected associations only
• Cannot resolve directionality of information flow

SWpC Method

• Embeds directional LTI model within each sliding window
• Two complementary measures:
  1. Strength: Prediction correlation
  2. Duration: Window-wise duration of information transfer

Methodology

Algorithm

1. Sliding window for time series segmentation
2. LTI model fitting within each window between all region pairs
3. Duration estimation across windows
4. Output: Time-varying directed connectivity matrices

Validation

• Multimodal: Concurrent LFP and fMRI BOLD
• Task fMRI: HCP motor task data
• Clinical: Post-concussion vestibular dysfunction

Applications

• Task-based fMRI analysis
• Clinical neuroscience: brain-state shifts detection
• Brain-computer interfaces
• Network neuroscience

Key Insights

1. Direction matters in brain connectivity
2. Two complementary measures: strength and duration
3. Multimodal consistency across LFP and BOLD
4. Clinical utility: improved healthy vs patient discrimination

References

• Xu, N., Zhang, X., Pan, W.-J., et al. (2026). SWpC. arXiv:2602.16004.