Enhanced brain efficiency network by integrating the new causality with fMRI and its application for Alzheimer’s Disease study

Abstract

The analysis of influence flow is crucial for the topological characterization of functional brain efficiency networks. Directional interaction among areal BOLD signals can be modeled by the causality method. Despite being widely used for studying Alzheimer's Disease in the brain, the Granger Causality (GC) has been shown to be unable to reveal true causality relationships through mathematical proof and simulation in EEG experiments. To evaluate its effectiveness in fMRI studies, first, we integrated a novel causality method called the New Causality (NC) with fMRI data. Both strong and weak causal impacts between stochastic processes were simulated and tested by GC and NC methods. Additionally, 1,893 patients in different stages of progression toward Alzheimer s disease were acquired and analyzed through the causality-based connectivity study. Finally, machine learning was employed to explore the performance in classification under these two methods. Simulation results show that compared to the GC, the NC method is more sensitive and reasonable to address causality relationships, especially for those weak causal impacts. Both the brain efficiency network and the classification performance can be enhanced through the NC introduction. Furthermore, it provides additional evidence supporting the critical involvement of the middle insular cortex, along with the temporal, parietal, and frontal lobes, in consciousness and functional diversion, with the help of NC integration.