FMRI

From DataMining

Contents 1 Resources for a Course on Data Analysis of fMRI 2 fMRI Background 3 Machine Learning 4 Software 5 Data Sets 6 Papers 7 Homework 8 Courses at Penn 9 fMRI at the University of Pennsylvania

Resources for a Course on Data Analysis of fMRI

This page is a collection of resources for a course on using machine learning methods for the analysis of fMRI data. It assumes a background in basic linear algebra, matlab, and basic familiarity with fMRI

fMRI Background

• fMRI for newbies
• Aguirre's lectures and
• mit OpenCourseWare

Machine Learning

We will use the key concepts of machine learning from the first half of CIS520, which draws heavily on Andrew Moore's excellent notes and on a set of readings, mostly from The Elements of Statistical Learning.

• Key concepts
  • Linear algebra review
  • Matlab Tutorial
  • Linear Regression
  • Generative vs Discriminative, Naive Bayes and Logistic Regression
  • Hypothesis testing vs. regression
  • Overfitting, Bias-Variance Decomposition, Cross validation
  • multiple comparisons problem, FDR
  • L0 regularization, stepwise, stagewise and streamwise regression
  • L1 regularization (lasso, elastic net)
  • Support Vector Machines (SVMs) and perceptrons
  • PCA, SVD, ICA
  • transfer learning
  • multi-view learning (and CCA)
  • significance testing

Software

• matlab
  • logistic regresion b = glmfit(X,y,'binomial','link','logit')
• lasso (with matlab interface)
  • http://www2.imm.dtu.dk/pubdb/views/publication_details.php?id=3897 (I used this one)
  • http://www.stat.berkeley.edu/twiki/Research/YuGroup/Software (Not tested; also, lacks elastic net)
• libsvm (with a matlab interface)
• MVPA Princeton MVPA toolbox and manual
• voxbo
• we will not use spm, fsl, anfni
• CCA in R on CRAN
• CCA code

Data Sets

we should later get versions that are ready to go for non-fMRI savvy people.

• haxby data
• StarPlus CMU data
• Pittsburgh Brain Activity Interpretation Competition

Papers

• Overview of ML and fMRI
  • Machine learning classifiers and fMRI: A tutorial overview Francisco Pereira, Tom Mitchell, Matthew Botvinick. NeuroImage 23, 250–263. doi:10.1016/j.neuroimage.2008.11.00

• SVM
  • Functional magnetic resonance imaging (fMRI) “brain reading”: detecting and classifying distributed patterns of fMRI activity in human visual cortex NeuroImage Volume 19, Issue 2, June 2003, Pages 261-270
• lasso, elatstic net
  • Interpretable Classifiers for fMRI Improve Prediction of Purchases; IEEE Transactions on Neural Systems and Rehabilitation Engineering, VOL. 16, NO. 6, Dec. 2008 539
  • Interpreting single Trial data using groupwise regularisation. van Gerven, M.; Hesse, C.; Jensen, O. & Heskes, T. Neuroimage, 2009 http://www.ncbi.nlm.nih.gov/pubmed/19285139
• ICA
  • Spatially independent activity patterns in functional MRI data during the Stroop color-naming task Martin J. McKeown ... Terrence J. Sejnowski, PNAS February 3, 1998 vol. 95 no. 3 803-810

• CCA
  • Examining Associations Between FMRI AND EEG Data Using Canonical Correlation Analysis Nicolle Correa, Yi-Ou Li, and Tulay Adalı, Vince D. Calhoun http://www.csee.umbc.edu/~adali/pubs/IEEEpubs/isbi2008correa.pdf
  • Investigating the relationship between pharmacological MRI and electrophysiology using Canonical Correlation Analysis. Forum of European Neuroscience 6, 123 (07/13/ 2008) http://www.kyb.mpg.de/publication.html?publ=5336 (abstract only?)

• PLS
  • Spatial Pattern Analysis of Functional Brain Images Using Partial Least Squares A. R. MCINTOSH, F. L. BOOKSTEIN, J. V. HAXBY, AND C. L. GRADY. NEUROIMAGE 3, 143–157 (1996) ARTICLE NO. 0016 ftp://ftp.rotman-baycrest.on.ca/pub/Randy/PLS/pls_article.pdf
• significance testing
  • Detecting signals in FMRI data using powerful FDR procedures. Statistics and Its Interface Volume 1 (2008) 23–32 Martina Pavlicov´a, Thomas J. Santner and Noel Cressie www.intlpress.com/SII/p/2008/1-1/SII-1-1-A3-Pavlicova.pdf
    • Note citations to Benjamini and his co-authors
• Artificial Neural Networks
  • Polyn et al (2005) http://www.sciencemag.org/cgi/content/abstract/310/5756/1963

• Heirarchical Bayes
  • Exploratory fMRI Analysis without Spatial Normalization Lashkari and Golland
• Clustering
  • Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage. 2009 Jan 1;44(1):83-98.

• permutation testing
  • Nonparametric permutation tests for functional neuroimaging: A primer with examples. Thomas E. Nichols, Andrew P. Holmes (DOI 10.1002/hbm.1058) Human Brain Mapping 15(1) , Pages 1-25. 2002
    • Includes matlab code

• applications
  • Predicting Human Brain Activity Associated with the Meanings of Nouns Science 320, 1191 (2008). Tom M. Mitchell, Svetlana V. Shinkareva. Andrew Carlson, Kai-Min Chang, Vicente L. Malave, Robert A. Mason, Marcel Adam Just
  • Hanson, S. J., & Halchenko, Y. O. (2008). Brain reading using full brain support vector machines for object recognition: there is no "face" identification area. Neural Computation, 20(2), 486-503.

• CCA (but not fMRI)
  • Retrieving Keyword’s to an Image Query using Kernel CCA. Hardoon et al. 2004 http://eprints.pascal-network.org/archive/00000257/01/keywords_m.pdf

• general fMRI statistics/Background
  • Empirical Analyses of BOLD fMRI Statistics, 1997 E. Zarahn, G. K. Aguirre and M. D'Esposito1 doi:10.1006/nimg.1997.0263

• Brain connectivity contest
  • http://sfcweb.lrdc.pitt.edu/pbc/2009/
  • Bullmore & Sporns 2009 Review of graph theoretic analysis brain networks in

Homework

There will be weekly homework and a substantial final project, involving the application of both standard and novel ML methods to fMRI data.

Courses at Penn

• CIS 537 Biomedical Image analysis (Gee, Yushkevich)
• PSYC-745 Seminar FMRI Data Analysis (Aguirre, Epstein)
• CIS520 Machine Learning (Taskar, Ungar)

fMRI at the University of Pennsylvania

• center for neuroimaging
• psychology
• radiology
• cognitive neuroscience
• people
  • Geoff Aguirre
  • Christos Davizikos
  • John Detre
  • Russel Epstein
  • James Gee

---

Main Page for the University of Pennsylvania Datamining group