CausalMamba: Scalable Conditional State Space Models for Neural Causal Inference

TL;DR

CausalMamba is a scalable framework that improves neural causal inference from fMRI data by decomposing the problem into BOLD deconvolution and causal graph inference, outperforming existing methods in accuracy and fidelity.

Contribution

It introduces a novel two-stage approach combining BOLD deconvolution and a Conditional Mamba architecture for scalable neural causal inference from fMRI data.

Findings

Achieves 37% higher accuracy than DCM on simulated data.

Recovers neural pathways with 88% fidelity in real fMRI data.

Detects dynamic brain hub reconfiguration during working memory tasks.

Abstract

We introduce CausalMamba, a scalable framework that addresses fundamental limitations in fMRI-based causal inference: the ill-posed nature of inferring neural causality from hemodynamically distorted BOLD signals and the computational intractability of existing methods like Dynamic Causal Modeling (DCM). Our approach decomposes this complex inverse problem into two tractable stages: BOLD deconvolution to recover latent neural activity, followed by causal graph inference using a novel Conditional Mamba architecture. On simulated data, CausalMamba achieves 37% higher accuracy than DCM. Critically, when applied to real task fMRI data, our method recovers well-established neural pathways with 88% fidelity, whereas conventional approaches fail to identify these canonical circuits in over 99% of subjects. Furthermore, our network analysis of working memory data reveals that the brain strategically shifts its primary causal hub-recruiting executive or salience networks depending on the stimulus-a sophisticated reconfiguration that remains undetected by traditional methods. This work provides neuroscientists with a practical tool for large-scale causal inference that captures both fundamental circuit motifs and flexible network dynamics underlying cognitive function.