Causal Discovery Methods for Effective Brain Connectivity Using fNIRS

Yaru Zhou; Samuel Montero Hernandez; Howard Bowman

Causal Discovery Methods for Effective Brain Connectivity Using fNIRS

Yaru Zhou
, Samuel Montero Hernandez^*
, Howard Bowman

^*Corresponding author for this work

Research output: Contribution to conference (unpublished) › Abstract › peer-review

Abstract

Estimating effective connectivity (EC) from functional near-infrared spectroscopy (fNIRS) remains underexplored. This study evaluates four causal discovery algorithms—Granger causality (GC), vector autoregressive (VAR)-based GC, Greedy Equivalence Search (GES), and Peter-Clark (PC)—on simulated neural time series. Algorithm performance generally increases with network size. With respect to noise, GC and VAR-GC decline while GES and PC improve with higher noise. Overall, GC and VAR-based GC outperform GES and PC across simulation conditions. Future work will develop causal discovery methods to estimate EC from deconvolved fNIRS signals.

Original language	English
Publication status	Published - 15 May 2026
Event	IX Biennial Meeting of the Society for functional Near-InfraRed Spectroscopy - University of Macau, Macau, China Duration: 16 Oct 2026 → 19 Oct 2026 https://fnirs2026.fnirs.org/

Conference

Conference	IX Biennial Meeting of the Society for functional Near-InfraRed Spectroscopy
Abbreviated title	fNIRS 2026
Country/Territory	China
City	Macau
Period	16/10/26 → 19/10/26
Internet address	https://fnirs2026.fnirs.org/

Keywords

fNIRS
Causal discovery
Effective connectivity
Brain connectivity
Neural time series

Cite this

@conference{620cf09d12bf446db0a35f79c45d8306,

title = "Causal Discovery Methods for Effective Brain Connectivity Using fNIRS",

abstract = "Estimating effective connectivity (EC) from functional near-infrared spectroscopy (fNIRS) remains underexplored. This study evaluates four causal discovery algorithms—Granger causality (GC), vector autoregressive (VAR)-based GC, Greedy Equivalence Search (GES), and Peter-Clark (PC)—on simulated neural time series. Algorithm performance generally increases with network size. With respect to noise, GC and VAR-GC decline while GES and PC improve with higher noise. Overall, GC and VAR-based GC outperform GES and PC across simulation conditions. Future work will develop causal discovery methods to estimate EC from deconvolved fNIRS signals.",

keywords = "fNIRS, Causal discovery, Effective connectivity, Brain connectivity, Neural time series",

author = "Yaru Zhou and \{Montero Hernandez\}, Samuel and Howard Bowman",

year = "2026",

month = may,

day = "15",

language = "English",

note = "IX Biennial Meeting of the Society for functional Near-InfraRed Spectroscopy, fNIRS 2026 ; Conference date: 16-10-2026 Through 19-10-2026",

url = "https://fnirs2026.fnirs.org/",

}

TY - CONF

T1 - Causal Discovery Methods for Effective Brain Connectivity Using fNIRS

AU - Zhou, Yaru

AU - Montero Hernandez, Samuel

AU - Bowman, Howard

PY - 2026/5/15

Y1 - 2026/5/15

N2 - Estimating effective connectivity (EC) from functional near-infrared spectroscopy (fNIRS) remains underexplored. This study evaluates four causal discovery algorithms—Granger causality (GC), vector autoregressive (VAR)-based GC, Greedy Equivalence Search (GES), and Peter-Clark (PC)—on simulated neural time series. Algorithm performance generally increases with network size. With respect to noise, GC and VAR-GC decline while GES and PC improve with higher noise. Overall, GC and VAR-based GC outperform GES and PC across simulation conditions. Future work will develop causal discovery methods to estimate EC from deconvolved fNIRS signals.

AB - Estimating effective connectivity (EC) from functional near-infrared spectroscopy (fNIRS) remains underexplored. This study evaluates four causal discovery algorithms—Granger causality (GC), vector autoregressive (VAR)-based GC, Greedy Equivalence Search (GES), and Peter-Clark (PC)—on simulated neural time series. Algorithm performance generally increases with network size. With respect to noise, GC and VAR-GC decline while GES and PC improve with higher noise. Overall, GC and VAR-based GC outperform GES and PC across simulation conditions. Future work will develop causal discovery methods to estimate EC from deconvolved fNIRS signals.

KW - fNIRS

KW - Causal discovery

KW - Effective connectivity

KW - Brain connectivity

KW - Neural time series

UR - https://fnirs.org/resources/publications/nirs-niri-publications/

UR - https://fnirs.org/events/conferences/

M3 - Abstract

T2 - IX Biennial Meeting of the Society for functional Near-InfraRed Spectroscopy

Y2 - 16 October 2026 through 19 October 2026

ER -

Causal Discovery Methods for Effective Brain Connectivity Using fNIRS

Abstract

Conference

Keywords

Fingerprint

Cite this