Introduction to Functional Magnetic Resonance Imaging fMRI Research

1
Introduction to Functional Magnetic Resonance
Imaging (fMRI) Research

• By Cinly Ooi

2
Contents

• Different kinds of MRI Imaging
• The Brain Mapping Unit, Department of Psychiatry,
  University of Cambridge
• fMRI research cycle
• How fMRI data are analyzed
• Areas of research current and future development
• Computing
• A few words for new comers to fMRI research

3
Magnetic Resonance Imaging

• Structural MRI (MRI)
• functional MRI (fMRI)

Common factor 4D images are acquired
4
Structural vs functional MRI

• Structural MRI (MRI)
• Used for studying the spatial relationship of
  voxels
• End result is a 3D image
• The time dimension is collapsed to create a
  clearer 3D image. Generally, the longer the
  acquisition, the clearer the image.
• Functional MRI (fMRI)
• Used for studying the temporal relationship of
  individual voxel
• End result is a 4D image
• The shorter the time of acquiring each 3D image
  (repetition time), the better the temporal
  resolution.

5
Brain Mapping Unit

• Part of Department of Psychiatry, University of
  Cambridge.
• Established in Mid 1990s.
• Currently have about 22 staffs and postgraduate
  students.
• Research Interest
• use fMRI techniques to study psychiatric problem.
• Develop new ways of analyzing fMRI data.
• Get fMRI data from Wolfson Brain Imaging Centre
  (WBIC) 3T MRI Machine from Bruker GmbH and
  P.E.T. Scanner.
• We only study human subjects.

6
fMRI Activities

• Method studies
• Creating/improving fMRI data analysis techniques
• Combine signal processing/image processing
  techniques and Statistics with current
  understanding of how the brain functions.
• Psychiatry
• Conduct Patients/Normal subject fMRI experiments
  to understand psychiatric illnesses.

7
Standard Experiment Cycle
Start
8
Data Acquisition
Picture of MRI Machine
MRI machine image downloaded from the internet
and is not the one used at WBIC
fMRI images modified from kircher et. al. 01
Please note that images used on this and
subsequent pages are all artificially created.
9
Data Analysis
Voxel Signal Extraction
Model Fitting
Convolution
10
Statistical Map Generation
Statistical Map Generation
11
Group Map Generation

• Shows brain activation for each drug condition
• Its an attempt to remove individual-specific
  results from group of experiment under the same
  condition

12
Group Mapping
Subject 1s Statistics Map
Subject 2s Statistics Map
Group Activation Map
Subject 3s Statistics Map
Subject 2s Statistics Map
13
Contrast between conditions

• Here, statistical operations are performed on
  group maps.
• The aim is to compare brain activation between
  different drug conditions, e.g. between drug A
  and placebo condition

14
Contrast between groups
Condition 2 Group Statistics Map
Condition 1 Group Statistics Map
Brain area activated by Condition 1 and Not
Condition 2
15
Areas of Research

• All areas, from MRI machine design, experiment
  designs to final result interpretation, are
  constantly evolving.
• In methodology, more active areas of research
  are
• Statistical Maps Generation
• Group maps and contrast maps generations
• Model fitting

16
New Directions

• Creation of fMRI Data Repositories fMRI Center
  in the US and Neurogenerator in Europe
• Fitting multiple time series at the same time
  either independently of each other (parallel
  processing), or to pass information between them
  (symbiotic evolution, better)
• Exploiting improvement in computing technology
  computer clusters, XML, computing speed

17
Computing Requirement

• Individual Map
• Each 4D file from Scanner 500MB approx
• Intermediate data 500MB x 5
• Processing time 1 hour
• Best done with a script (at least 36 map per exp)
• Group Map
• 20 minutes per individual map
• 30 minutes to generate group map
• Contrast Map
• 5 minutes

18
Graphical User Interface

• Most researchers do not like to do Graphical User
  Interface waste of research time
• Unfortunately end users like GUI.
• Currently no GUI ALMOST means no user beyond a
  small closely supported groups

19
Problems with fMRI software

• File formats problem
• Different software uses different file format
• No interoperability
• Even though the inputs are almost identical
• Difficult to switch processing from one software
  package to another
• No comparison between different softwares
• This is stuff that can start a religious war
• Like most academic softwares, no thorough testing

20
Advice to new comers to fMRI software

• Learn computing skills
• Most fMRI analysis software are written for
  Linux/Unix Scripts, C/C, Matlab, Java/Motif
• The more computing skills you have, the more
  productive you become.
• Understand the software you are using!
• Do not use them as blackboxes!
• Test them thoroughly Regression test,
  functionality test
• Document your source code
• the researcher himself is to be blamed for using
  the wrong software, not the software developers!

21
The End

• This presentation can be downloaded from
• http//www-bmu.psychiatry.cam.ac.uk/SEMINARS/past
  .html
• Presenters contact details
• Cinly Ooi
• Box 255, Brain Mapping Unit,
• University Department of Psychiatry,
• Addenbrookes Hospital
• Cambridge CB2 2QQ
• United Kingdom
• Tel [phone]
• Email co224_at_cam.ac.uk
• Webpage http//www-staff.psychiatry.cam.ac.uk/co
  224