Eric Egolf

1

• Eric Egolf
• Functional Connectivity Toolbox
• Year Long Project
• Project Advisor Professor Armen

2
Introduction

• Overview of fMRI, Functional Connectivity
• Project Description
• Algorithms and Implementations
• Timeline

3
Overview of Functional Connectivity

• fMRI measures brain activity. Measuring activity
  during cognitive tasks gives more knowledge about
  the brain.
• Applying Mathematical Algorithms that measure
  Functional Connectivity, on fMRI data, detects
  concurrent brain activity.
• A Voxel is a unit of measurement in the brain
• fMRI data gets processed and outputs
  4D(x,y,z,t) images.

4
Project Description

• Algorithms for Functional Connectivity
• Correlation, ICA, Phase Relationships, Mutual
  Information
• Implemented in Matlab
• Use existing modules, platform independent,
  matrix library
• Universal Input and Output Types

3D images
Toolbox (Matlab)
4D images
Graphical Displays
fMRI data
Processing
5
Correlation

• Correlation Coefficient
• Result Matrix of Voxels vs. Voxels
• Element in Matrix is a voxel correlated to
  another voxel through time
• Problems
• Size of Matrix
• (110592 110592 voxels)
• Interpreting Results
• Optimize Later

6
Independent Components Analysis

• ICA is a method of Blind Source Signal Separation
• Only knowing Mixed Signals extract Original
  Signals
• Applied to fMRI(extracts two parts)
• 1) Time course of voxel intensity
• 2) Component or voxels in brain that follow this
  time course
• Extracts Multiple (ComponentsTime Courses)
• View on horizontal slices of the brain

Mixed Signals
Original Signals
7

• Horizontal Slices
• 3 components overlaid
• on subjects brain
• Each component is made
• up of voxels above a
• Threshold that are
• Synchronized with a
• time course
• 1 components time
• course

8
Progress with ICA

• Extract time courses and components, store as 3D
  image.
• Overlay multiple components on structural image.
• Sort components based on time courses
  correlation to a reference function.
• GUI.

9
Timeline

• January Finish Single Subject ICA
• Febuary ICA Group Analysis
• March Optimize Correlation, Mutual Information
  as an Algorithm
• April Group Analysis