Integrating group-wise functional brain activities via point processes

Studying functional brain activities based on analyzing BOLD signals derived from fMRI data has received significant interest in the neuroimaging field. However, there exists considerable variability of BOLD signals for the corresponding brain region of interest (ROI) across different subjects. To extract more reliable and representative information from BOLD signals, in this paper, we propose a novel stochastic group-wise task BOLD information assessment framework. First, each BOLD signal is characterized as a point process. Second, a state-space generalized linear model is built to integrate group-wise point processes of the corresponding ROI across subjects. Third, a dynamics rate function is proposed to assess the stochastic group-wise BOLD information. Our experimental results based on working memory task fMRI data demonstrate that the resulting stochastic group-wise BOLD information is more accurate and informative than the original BOLD signals in each individual subject in terms of more robust response to the task stimulus.