[Paper Review] Measuring the functional connectome "on-the-fly": towards a new control signal for fMRI-based brain-computer interfaces

Ricardo Pio Monti, Romy Lorenz+3 more|arXiv (Cornell University)|Feb 8, 2015

Functional Brain Connectivity Studies47 references3 citations

TL;DR

This paper proposes rt-SINGLE, a real-time algorithm for estimating dynamic functional connectivity networks from fMRI data by adapting the SINGLE method for sparse, temporally homogeneous networks. It enables rapid detection of task- and stimulus-induced changes in brain network structure, demonstrating feasibility for next-generation fMRI-based brain-computer interfaces using whole-brain connectivity as a control signal.

ABSTRACT

There has been an explosion of interest in functional Magnetic Resonance Imaging (MRI) during the past two decades. Naturally, this has been accompanied by many major advances in the understanding of the human connectome. These advances have served to pose novel challenges as well as open new avenues for research. One of the most promising and exciting of such avenues is the study of functional MRI in real-time. Such studies have recently gained momentum and have been applied in a wide variety of settings; ranging from training of healthy subjects to self-regulate neuronal activity to being suggested as potential treatments for clinical populations. To date, the vast majority of these studies have focused on a single region at a time. This is due in part to the many challenges faced when estimating dynamic functional connectivity networks in real-time. In this work we propose a novel methodology with which to accurately track changes in functional connectivity networks in real-time. We adapt the recently proposed SINGLE algorithm for estimating sparse and temporally homo- geneous dynamic networks to be applicable in real-time. The proposed method is applied to motor task data from the Human Connectome Project as well as to real-time data ob- tained while exploring a virtual environment. We show that the algorithm is able to estimate significant task-related changes in network structure quickly enough to be useful in future brain-computer interface applications.

Motivation & Objective

• To develop a real-time method for estimating dynamic functional connectivity networks in fMRI, moving beyond single-region ROI-based neurofeedback.
• To address the challenge of estimating temporally evolving, sparse functional networks quickly enough for closed-loop brain-computer interface (BCI) applications.
• To validate the method on both task-based fMRI data and real-time virtual environment exploration, demonstrating sensitivity to functional network changes.
• To provide a computationally efficient solution that reports updated functional connectivity networks at each time point, suitable for real-time feedback.

Proposed method

• Adapts the SINGLE algorithm—originally for offline, sparse, temporally homogeneous network estimation—to operate in real-time using recursive estimation with adaptive forgetting factors.
• Employs recursive updates for sample mean and covariance matrix using a weighted moving average with a forgetting factor that adjusts over time.
• Uses a log-likelihood-based objective function with a non-differentiable penalty term, approximated via derivative calculation with respect to the forgetting factor.
• Applies the Sherman-Woodbury formula to efficiently update the inverse covariance matrix without full re-inversion at each step.
• Derives analytical gradients for the adaptive forgetting factor by differentiating the log-likelihood with respect to the forgetting parameter, enabling online optimization.
• Implements a block-wise updating strategy to maintain computational efficiency while ensuring accurate tracking of network changes.

Experimental results

Research questions

• RQ1 Can a real-time algorithm accurately estimate dynamic functional connectivity networks from fMRI data with sufficient speed for BCI applications?
• RQ2 How well can the proposed method track rapid changes in functional connectivity induced by motor tasks or environmental stimuli?
• RQ3 Does the rt-SINGLE algorithm maintain accuracy in estimating sparse, temporally homogeneous networks under real-time constraints?
• RQ4 Can the method detect functionally relevant network reconfigurations in response to external stimuli, such as changes in visual environment brightness?
• RQ5 What is the computational cost of the algorithm, and is it feasible for deployment in real-time neurofeedback systems?

Key findings

• The rt-SINGLE algorithm successfully detected task-related changes in functional connectivity during motor tasks using data from the Human Connectome Project, confirming its sensitivity to neural network dynamics.
• In a real-time virtual environment study, the algorithm accurately reported a network of edges activated during daylight conditions, demonstrating responsiveness to external stimuli.
• Simulations showed that rt-SINGLE effectively tracks changes in covariance structure over time, maintaining high accuracy in dynamic network estimation.
• The method achieved real-time performance, with network estimates reported at each time point, satisfying the low-latency requirement for closed-loop BCI systems.
• The computational cost was found to be manageable, with efficient recursive updates enabling real-time operation without full matrix inversion at each step.
• The algorithm outperformed standard windowed correlation methods in detecting subtle, task-modulated connectivity changes, particularly in sparse network configurations.