CSCI 6906: Foundations of Computational Neuroimaging 2011

◇◇◇

Instructors: Dr. Thomas Trappenber, Dr. Ryan D'Arcy, Jason Satel (MCS), Dr.Tim Bardouille, Dr. Xiaowei Song

Office:Room 4216, in Mona Campbell Building on Coburg RD
email: tt@cs.dal.ca
Office hour: send email

Outline

Neuroimaging technologies have developed dramatically over the last decade and have increased the need for advanced data analysis techniques. This course is an introduction to some neuroimaging and related data analysis techniques, and covers both theory and hands-on examples. We focus in particular on data from EEG, MEG and fMRI. The course covers the physical basis of the signals, common acquisition techniques, and advanced signal processing paradigms such as independent component analysis and causal modeling. The course includes an introduction to the Matlab programming environment that will be used for some hands-on exercises, and a primer on probability theory on which parts of the theory is based.

Syllabus

Notes:

will be posted here; please check frequently

Assignments:

1. The file signals.mat contains signals (time series) of 32 channels. Plot all 32 time series in a single figure with an offsets for each channel so that the data don't overlap. Send your matlab program, that reads and displays the data when executed, to prof6906@cs.dal.ca with subject line A1 by Jan. 18.

2a. Write a function dft to implement a discrete Fourier transform. Use this function to calculate the frequency spectrum of a square pulse function which is zero unless 20<t<30. Compare your resuts with the matlab implementation of function fft(). Apply a low pass filter and plot the resulting frequency spectrum and the filtered time signal by using the inverse FFT.

b. Plot the frequency and phase spectrum for one channel of a recorded EEG signal.

c. Include a short explanation (script) how you extracted the data that you are using in your analysis.

Send your results with a brief decription to prof6906 with subject line A2 before class on Feb. 15.

3a In file signal2 are a signal with 4 channels. Analyse the data with FFT, DWT, HHT, ICA. Show and discuss your results briefly.

b. Repeat the analysis above with some channels of EEG signals.

Send your results with a brief decription to prof6906 with subject line A2 before class on Tuesday March 1.

Resources:

Brief Introduction to Matlab

Signal file for first hands-on with matlab

Intro slides from Ryan D'Arcy

EEG module (by Jason Satel): Slides for EEG intro lecture, EEG setup script, Data analysis script

filter demo applet

Hilbert transform of EEG

Blind source separtion audio example

ICA demo

Papers

MEG lecture by Tim Bardouille

Links to matlab implementations of the methods we talked about:

FFT: there is a standard matlab implementation

HHT: Implementation of the intrinsic mode functions a la Huang and uses Hilbert transform from Matlab Signal Processing Toolbox

DWT: Need to look for a good package. Let me know if you find one.

FastICA: Very popular package for ICA

ICALAB: Have not tried this but look interesting

SVM: LibSVM is very popular

Some more papers and fun links

Recent review paper on machine learning in neuroimaging from Berlin group

Another overview by the Berlin group

http://nowpossible.com/bci.htm

Video lecture by Klaus-Robert Muller

The only computer game I every played and programmed in High School

Plegarism:

Please familiarize yourself with the university policy on Intellectual Honesty. Every suspected case will be reported.

Preliminary Schedule