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

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Imaging Analysis

Background:

• Current functional magnetic resonance imaging (fMRI) analysis methods for causal relationships often rely on hypothesis-driven or linear models.
• These linear approaches may lead to inaccuracies in detecting true brain activity and effective connectivity (EC).
• There is a need for nonlinear methods to more accurately assess brain activity and EC.

Purpose of the Study:

• To introduce a novel nonlinear method, EC-BP (effective connectivity based on back-propagation neural network), for evaluating human brain EC using fMRI data.
• To assess the feasibility and accuracy of EC-BP by comparing it with Granger causality analysis (GCA) using simulated time series.
• To apply and evaluate EC-BP on real fMRI data from healthy subjects.

Main Methods:

• Development of a novel EC-BP technology utilizing a back-propagation neural network with a nonlinear model.
• Simulation of four distinct time series datasets to rigorously test EC-BP's performance against GCA.
• Application of the EC-BP method to fMRI data acquired from 60 healthy human participants.

Main Results:

• Simulated data analysis confirmed EC-BP's ability to accurately identify original causal relationships, unlike GCA which failed to detect nonlinear causality.
• Analysis of healthy subject fMRI data revealed significant differences between EC-BP and GCA in the top 50 effective connections.
• EC-BP highlighted connections involving the hippocampus and parahippocampus, whereas GCA focused on regions like the paracentral lobule, caudate, putamen, and pallidum.

Conclusions:

• The proposed EC-BP method offers a more comprehensive approach to detecting effective connectivity by capturing nonlinear dynamics in brain activity.
• EC-BP provides valuable supplementary information to traditional methods like GCA, enhancing the understanding of brain network interactions.
• This nonlinear approach promises to advance the detection and evaluation of effective connectivity in neuroimaging studies.