BME 499.098/Biostat 642:
Introduction to Functional MRI

Schedule

Meeting Time: 10AM to Noon, Monday-Friday, Aug 6 through Aug 31
Meeting Place: 1528 CC Little, except as noted.

Lab Meeting Times: 1PM-3PM, only on days indicated below.
Lab Meeting Place: PC lab in the basement of East Hall, EH B250

New:
Course Evaluation

Course Syllabus & Class Goals

Week 1

Mon Aug 6: Unix
Instructor: Luis Hernandez
Lab: 1-3PM

After this class you will be able to

Understand the foundations of the Unix operating system and how it can be used as an important tool for processing functional images.
Use basic commands and their syntax, and have understanding of the concepts of:
- Shells and Kernels
- Jobs and processes
- Directory structure
- Environment variables
- File Permissions and ownership
- Networking: remote use and file transfers
Use basic Unix tools for (1) Quick File editing and (2) finding help on Unix topics.

Referencs

http://www.bme.umich.edu/~hernan/Public/FMRI_class/
Class notes
UNIX Tutorial for Beginners from University of Surrey.
A Basic UNIX Tutorial from Idaho State University.
Unix Tips, slides from Brown Bag seminar by Tom Nichols.
"Understanding Unix", by Stan Kelly-Bottle, 1994, Sybex.

(Note: the Unix lecture will provide the tools for completion of homework assignments based on real functional imaging applications.)

Tue Aug 7: Matlab
Instructor: Luis Hernandez
Lab: 1-3PM

After this class you will be able to

Use Matlab as a data processing tool, in particular in the elementary processing of imaging data.
Navigate through the Matlab environment (path , startup , editors, help, reseources)
Manipulate variables in terms of vectors, matrices.
Use elementary operations on vectors, matrices, and scalars
Examine data thru plots and images.
Distinguish data types in matlab.
Handle strings.
Store and retrieve data in different formats.
Write your own scripts and functions

Referencs

Matlab tutorials on the web:
http://www.math.siu.edu/matlab/tutorials.html
http://eng.usna.navy.mil/~mecheng/DESIGN/CAD/MATLAB/usna.html
http://www.che.lsu.edu/links/computing/tutorials/matlabt.htm

(Note: the matlab lectures will provide the tools for completion of homework assignments based on real functional imaging applications.)

Wed Aug 8: NMR Physics
Instructor: Doug Noll

After this class you will be able to

Develop the classical description of the motion of a nuclear spin in a magnetic field, including the Larmor relationship.
Describe the quantum mechanical basis of the magnetic resonance signal and its implications to signal strength and relaxation.
Use the rotating frame concept to describe excitation.
Describe signal strength for a pulse-echo experiment in terms of T1, T2, TE and TR.

Readings

Course notes on the web: http://www.eecs.umich.edu/~dnoll/BME499

Thur Aug 9: Image Formation
Instructor: Doug Noll

After this class you will be able to

Introduce the concept of a magnetic field gradient and the space-frequency relationship.
Introduce the Fourier transform and demonstrate how 1D localization can be done.
Describe phase encoding and 2D imaging methods.
Describe slice selection and pulse sequence basics.

Readings

Course notes on the web: http://www.eecs.umich.edu/~dnoll/BME499

Fri Aug 10: The BOLD Effect
Instructor: Doug Noll

After this class you will be able to

Describe magnetic susceptibility and how changes in the magnetic susceptibility of hemoglobin it can affect T2* relaxation.
Describe the metabolic and vascular physiology changes associated the neuronal activation.
Describe temporal dynamics of changes in physiology and BOLD signal.
Introduce some basic models of the BOLD effect and describe responses to long and short stimuli.

Readings

Course notes on the web: http://www.eecs.umich.edu/~dnoll/BME499

Week 2

Mon Aug 13: Artifacts and noise in fMRI
Instructor: Doug Noll
Lab: 1-3PM

After this class you will be able to

Describe the most common sources of additive noise in MRI.
Introduce the basic relationships between imaging parameters and noise.
Describe signal-dependent (multiplicative) sources of signal variation in MRI, including physiological noise and head movement.
Describe imaging artifacts associated with rapid image acquisition methods.
Describe imaging artifacts associated with T2*-weighted imaging.

Readings

Course notes on the web: http://www.eecs.umich.edu/~dnoll/BME499

Tue Aug 14: Hardware and Safety
Instructor: Luis Hernandez
Special Class location: White Auditorium, Cooley Hall, North Campus

After this class you will be able to

Explain the physical properties of magnetic fields as they relate to operation of a MR scanner, and safety precautions.
Know the basic hardware components of a typical MR scanner: RF Coils, Gradients, Amplifiers, Computers.
Name the different components of the functional MRI lab.

Wed Aug 15: Task Design I
Instructor: Doug Noll

After this class you will be able to

Introduce basic concepts in task design in fMRI.
Describe blocked designs and methods for optimization.
Describe event-related designs with overlapping and non-overlapping responses.
Describes issues related to linearity and superposition of the fMRI response.
Present different methods for implementing event-related designs (jittered, constant ISI, randomization schemes, etc.).

Readings

"Experimental design for brain fMRI," Aguirre and D'Esposito, pp. 369-370, in Functional MRI, Moonen and Bandettini, Eds., Springer, 1999. (Handout)
"Event-related functional MRI," Buckner and Braver, Ibib., pp. 441-450. (Handout)

Thu Aug 16: No Class!
(Workshop on Language Processing in afternoon)

Fri Aug 17: Task Design II: E-Prime
Instructor: Tor Wager
Afternoon lab only, no morning lecture. (Due to Workshop on Language Processing in morning)
Lab 1-3PM

After this class you will be able to

Decipher the intricate structure of the e-prime graphic interface.
Program a basic experiment in e-prime.
Use the e-prime template to create a safe, fun IFIS-compatible experiment for use in the scanner.
Organize data and analyze with E-DataAid.

Week 3

Mon Aug 20: Task Design III: E-Prime
Instructor: Tor Wager
Class meets in Lab

After this class you will be able to

Create e-prime scripts with and without IFIS to implement almost any experiment you can dream up.
Use basic inline code to:
(1) Jump to other points in the experiment
(2) Write output to the parallel port for control of external devices
(3) Use subject performance to control the flow of the experiment.

E-prime Homework: Eprime_class_info.doc
E-prime Homework files:
fMRIBase.es
fMRIBase.ebs
expbackground.bmp
epi.wav
Ding.wav
Homework due next Wednesday, August 29

Tue Aug 21: Basic Statistics
Instructor: Tom Nichols

After this class you will be able to

Describe the basic principles of hypothesis testing.
Write down the general linear model for (1) a t-test and (2) a correlation.
Interpret the results of a F-test in terms of extra sums of squares variation.
(1) Explain estimability, and (2) write down a model and two contrasts, one which is estimable the other which is not.
Explain the difference between fixed effects and random/mixed effects inference.

Matrix algebra review

Handouts

Lecture Notes
Matlab Notes
Pain Data:
First column, subjective pain rating
Second column, laser energy
Third column, subject number

Wed Aug 22: fMRI Stats: Temporal Modeling
Instructor: Tom Nichols

After this class you will be able to

The impact of unmodeled temporal autocorrelation on the signal (the beta's) and the noise (sigma-squared).
List three different ways of accounting for (temporal) autocorrelation.
Draw a figure illustrating fMRI noise in the frequency domain, indicating physiological sources and noting the location of the BOLD signal's energy.
Explain the relative strengths of frequency-domain versus time-domain modeling.

Thu Aug 23: fMRI Stats: Spatial Modeling
Instructor: Tom Nichols

After this class you will be able to

Explain the difference between explicit and implicit spatial modeling.
Explain what a Familywise Error (FWE) is.
Explain the role of the maximum in the multiple comparisons problem.
Explain the impact of smoothness and search volume on corrected p-values.
Describe the assumptions neccessary for random field methods to be valid.

Lecture Notes
Quiz aka Homework

Readings

The following two papers are comprehensive reviews of functional neuroimaging (fNI) statistics. They are very dense (sorry). They are light on fMRI methods and instead focus on general issues in spatial modeling and inference in fNI.
KM Petersson, TE Nichols, J-B Poline, and AP Holmes.
Statistical limitations in functional neuroimaging I. Non-inferential methods and statistical models.
Philosophical Transactions of the Royal Society: Biological Sciences, 354:1239--1260, 1999.
KM Petersson, TE Nichols, J-B Poline, and AP Holmes.
Statistical limitations in functional neuroimaging II. Signal detection and statistical inference.
Philosophical Transactions of the Royal Society: Biological Sciences, 354:1261--1281, 1999.

Fri Aug 24: Intro to SPM
Instructor: Tom Nichols
Lab: 1-3PM

After this class you will be able to

Describe the format that SPM expect the data to be in.
Explain what SOA is and how you would specify an irregularly-spaced event-related experimental design.
Describe how to impliment a random effects analysis in SPM and list what the assumptions are for this approach to be valid.
Explain the best way is to localize activations anatomically.

Readings

SPM Manual (86 pages) PDF | Word | HTML

Week 4

Mon Aug 27: Preprocessing
Instructor: Luis Hernandez
Lab: 1-3PM

After this class you will be able to

Understand the basic assumptions in fMRI analysis, and how they are violated.
Understand motion correction through rigid body transformations
Understand the problem of slice timing in Event related fMRI experiments correction through interpolation.
Explore the pros and cons of filtering data in the temporal and the spatial domains.

Lab exercise II

Tue Aug 28: Inference & Combining Results
Instructor: Tom Nichols
Lab: 1-3PM

After this class you will be able to

Explain exactly how to interpret a cluster that has "corrected significance", versus one that only has "uncorrected significance".
Identify the most important peice of information in the footer of SPM output.
Explain the difference between a fixed effects model and a random effects model, and the conclusions you can drawn from their results.

Wed Aug 29: Advanced Inference
Instructor: Tom Nichols
Lab: 1-3PM

After this class you will be able to

Explain exactly how to interpret a FWE-corrected p-value versus a FDR-corrected p-value.
Explain the basic principles of the permutation approach to hypothesis testing & how these are applied to fMRI.

Thu Aug 30: Optimizing Power
Instructor: Tom Nichols & Tor Wager
Lab: 1-3PM

After this class you will be able to

Explain the difference between estimating the hemodynamic response and detecting a signal
Explain why block designs are optimal for [ estimating the HRF | signal detection ] and why event-related designs are optimal for [ estimating the HRF | signal detection ].
The role of autocorrelation in designing fMRI experiments.

Lecture notes

Fri Aug 31: Presentations

Students will present the data they collected and analyzed.

Course Evaluation: If you cannot attend, please fill this out and drop it in an interoffice envelope and send to
    Tom Nichols
    Biostatistics
    School of Public Health

Handouts

Correct onset times for event-related studies: ERtimes.mat
Subj 1's times were lost
Subj 2's times are in matlab variables sub9991.faces and sub9991.places
Subj 3's times are in matlab variables sub9992.faces and sub9992.places
Subj 4's times are in matlab variables sub9993.faces and sub9993.places
Subj 5's times are in matlab variables sub9995.faces and sub9994.places
Subj 6's times are in matlab variables sub9995.faces and sub9995.places
For all the gory detail, see Event_times.xls

Last modified: Wed Jun 12 17:08:50 EDT 2002

Tom Nichols nichols@umich.edu
UM Biostatistics