Abstract:Understanding how neural populations interact is crucial to understand brain function. Most common approaches to infer neural interactions are based on Granger causality (GC) analyses and effective connectivity (EC) models of neural time series. However, an in-depth investigation of the similarity and complementarity of these approaches is currently lacking. GC and EC are classically thought to provide complementary information about the interdependence between neural signals. Whereas GC quantifies the amount of predictability between time series and it is interpreted as a measure of information flow, EC quantifies the amount and sign of the interaction, and it is often interpreted as the causal influence that a neural unit exert over another. Here, we show that, in the context of functional magnetic resonance imaging (fMRI) data analysis and first-order autoregressive models, GC and EC share common assumptions and are mathematically related. More precisely, by defining a ‘corrected’ version of GC accounting for unequal noise variances affecting the source and target node, we show that the two measures are linked by an approximately quadratic relation, where positive or negative values of EC are associated with identical values of GC. While the relation is obtained in limit of infinite sampling time, we use simulations to show that it can be observed in finite data samples as classically observed in neuroimaging studies, provided sufficiently long sampling, multiple sessions or group averaging. Finally, we compare the GC and EC analyses on fMRI data from the Human Connectome Project, and obtain results consistent with simulation outcomes. While GC and EC analyses do not provide reliable estimates at the single subject or single connection level, they become stable at the group level (more than approximately 20 subjects), where the predicted relation between GC and EC can be clearly observed from the data. To conclude, our study provides a common mathematical framework to make grounded methodological choices in the reconstruction and analysis of directed brain networks from neuroimaging time series.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to explore the similarities and complementarities between Granger Causality (GC) and Effective Connectivity (EC) in the analysis of functional magnetic resonance imaging (fMRI) data. Specifically, the author hopes to evaluate the consistency and differences between these two methods in reconstructing and analyzing directed brain networks obtained from neuroimaging time series by establishing a common mathematical framework. The paper focuses on the following two aspects in particular: 1. **Topological structure**: Whether there are directional and causal interactions. 2. **Asymmetry**: Whether the connection strength in one direction is stronger than that in the opposite direction. ### Background and motivation - **Functional Connectivity (FC)**: Traditional functional connectivity methods are based on undirected functional connectivity (UFC), such as the correlation of BOLD signals. Although they are statistically reliable and the results are reproducible, they cannot provide information on the directionality and asymmetry of the interactions between brain regions. - **Granger Causality (GC)**: Based on the Wiener - Granger principle, it infers directional interactions by predicting whether one time series is helpful in predicting another time series. GC usually assumes a linear multivariate autoregressive (MAR) model as the generation process of BOLD time series. - **Effective Connectivity (EC)**: Assuming that the data is generated by a continuous dynamic system, EC weights quantify the strength and sign of directional interactions between brain regions. EC models are usually based on a system of linear stochastic differential equations. ### Research methods - **Theoretical analysis**: The author deduced the mathematical relationship between GC and EC under the assumption that the time series are generated by linear network dynamics and Gaussian input. In particular, they defined a "modified" version of GC to take into account the case where the noise variances of the source and target nodes are not equal. - **Numerical simulation**: By simulating the MOU process, the quadratic relationship derived theoretically was verified, and the performance of these relationships in finite - sample data was explored. - **Application to real - data**: Using resting - state fMRI data from the Human Connectome Project (HCP), the results of GC and EC analyses were compared. ### Main findings - **Theoretical relationship**: In the limit of infinite sampling time, there is an approximate quadratic relationship between GC and EC. The positive and negative values of EC correspond to the same GC value. - **Numerical simulation**: In finite - sample data, this relationship can still be observed as long as the sampling time is long enough or multiple sessions or group averaging are carried out. - **Real - data**: At the individual level or for a single connection, the estimates of GC and EC analyses are unreliable; but at the group level (more than about 20 subjects), the predictive relationship between GC and EC can be clearly observed from the data. ### Conclusion This study provides a common mathematical framework that enables researchers to make informed method selections when reconstructing and analyzing directed brain networks obtained from neuroimaging time series. This helps to better understand the dynamic characteristics of brain functional networks.

Directed neural interactions in fMRI: a comparison between Granger Causality and Effective Connectivity

Is Granger Causality a Viable Technique for Analyzing fMRI Data?

Causality analysis of neural connectivity: critical examination of existing methods and advances of new methods.

Nonlinear effective connectivity measure based on adaptive Neuro Fuzzy Inference System and Granger Causality

Minimum-phase property of the hemodynamic response function, and implications for Granger Causality in fMRI

Kernel Granger Causality Mapping Effective Connectivity on FMRI Data.

Nonlinear Connectivity by Granger Causality.

Multivariate Granger causality unveils directed parietal to prefrontal cortex connectivity during task-free MRI

A New Method on Construction of Brain Effective Connectivity Based on Functional Magnetic Resonance Imaging

A Conditional Granger Causality Model Approach for Group Analysis in Functional Magnetic Resonance Imaging

Recovering Directed Networks in Neuroimaging Datasets Using Partially Conditioned Granger Causality

Comparison of Functional Network Connectivity and Granger Causality for Resting State fMRI Data.

Investigation of Effective Connectivity in the Motor Cortex of Fmri Data Using Granger Causality Model

Effective Connectivity of Brain Network Based on Granger Causality and PCA

Different Features of Metabolic Connectivity Map and Granger Causality Method in Revealing Directed Dopamine Pathways: A Pilot Study Based on Integrated PET/MR

Analyzing coherent brain networks with Granger causality.

Kernel Granger Causality Mapping Effective Connectivity: A Resting fMRI Study

Novel Machine Learning Approaches for Improving the Reproducibility and Reliability of Functional and Effective Connectivity from Functional MRI

Detecting connectivity in EEG: A comparative study of data-driven effective connectivity measures

Kernel Granger Causality Based on Back Propagation Neural Network Fuzzy Inference System on fMRI Data

Exploring intricate connectivity patterns for cognitive functioning and neurological disorders: incorporating frequency-domain NC method into fMRI analysis