Tutorial: Computing dynamic functional connectivity and task-modulation based features from fMRI data for classification. 

Functional Magnetic Resonance Imaging (fMRI) can reveal brain's activity in an indirect manner by measuring blood-oxygenation level. Temporal signal change of different brain regions can be studied using predefined anatomical brain atlases or by using a data-driven blind-source separation method such as independent component analysis (ICA) which can reveal statistically independent functional networks in the brain. If there is a task or stimulus that the participant was engaged during the experiment, related brain activation maps can be computed. If there is no task or stimulus, which is usually called "resting-state", then the interaction of the temporal signal changes between different brain regions or networks, which is called functional connectivity (FC), can be studied. Until recently, FC had been measured in a static manner, characterizing the interaction with only a single number. However, brain's functional connectivity is highly dynamic; this necessitates computation of dynamic functional connectivity metrics from resting-state fMRI data, which can reveal more information about the highly dynamic behavior of the brain networks [1-5]. Moreover, dynamic FC metrics each can be used as features for classification algorithms for classification of different groups of participants, and it has recently been shown to improve classification accuracy [6]. In this tutorial, we will briefly introduce fMRI, ICA, computation of brain activation maps, and functional connectivity. We will present the methodology on how to compute dynamic functional network connectivity based features from the fMRI data. The methodology of extracting dynamic functional connectivity metrics from resting-state fMRI data, and modulation metrics of dynamic functional connectivity if a task or stimulus is present, will be introduced. A classification example using DFC-based features will also be demonstrated.

Presenter: Dr. Unal Zak, University of Houston, Texas.

Program Outline: This tutorial will be two hours in duration, presented on Thursday, 14 June 2018. 
  • Introduction to fMRI, independent component analysis, and brain networks. 
  • Discussion of static and dynamic brain functional connectivity (FC) methods. 
  • Discussion of classification using FC-based features. 
  • Demo of classification using a fMRI dataset. 
  • Topics of ongoing and future research.
Suggested Prerequisites and Materials