Bayesian Kernel Methods for Analysis of Functional Neuroimages

We propose an approach to analyzing functional neuroimages in which (1) regions of neuronal activation are described by a superposition of spatial kernel functions, the parameters of which are estimated from the data and (2) the presence of activation is detected by means of a generalized likelihood ratio test (GLRT). Kernel methods have become a staple of modern machine learning. Herein, we show that these techniques show promise for neuroimage analysis. In an on-off design, we model the spatial activation pattern as a sum of an unknown number of kernel functions of unknown location, amplitude, and/or size. We employ two Bayesian methods of estimating the kernel functions. The first is a maximum a posteriori (MAP) estimation method based on a reversible-jump Markov-chain Monte-Carlo (RJMCMC) algorithm that searches for both the appropriate model complexity and parameter values. The second is a relevance vector machine (RVM), a kernel machine that is known to be effective in controlling model complexity (and thus discouraging overfitting). In each method, after estimating the activation pattern, we test for local activation using a GLRT. We evaluate the results using receiver operating characteristic (ROC) curves for simulated neuroimaging data and example results for real fMRI data. We find that, while RVM and RJMCMC both produce good results, RVM requires far less computation time, and thus appears to be the more promising of the two approaches.

[1]  J B Poline,et al.  Analysis of Individual Positron Emission Tomography Activation Maps by Detection of High Signal-to-Noise-Ratio Pixel Clusters , 1993, Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism.

[2]  George Eastman House,et al.  Sparse Bayesian Learning and the Relevance Vector Machine , 2001 .

[3]  S. Strother,et al.  An evaluation of methods for detecting brain activations from PET or fMRI images , 1999, 1999 IEEE Nuclear Science Symposium. Conference Record. 1999 Nuclear Science Symposium and Medical Imaging Conference (Cat. No.99CH37019).

[4]  Nando de Freitas,et al.  An Introduction to MCMC for Machine Learning , 2004, Machine Learning.

[5]  L. K. Hansen,et al.  The Quantitative Evaluation of Functional Neuroimaging Experiments: The NPAIRS Data Analysis Framework , 2000, NeuroImage.

[6]  David J. C. MacKay,et al.  Bayesian Interpolation , 1992, Neural Computation.

[7]  Yongyi Yang,et al.  A signal-detection approach for analysis of functional neuroimages , 2001, 2001 IEEE Nuclear Science Symposium Conference Record (Cat. No.01CH37310).

[8]  S. Kay.  Fundamentals of statistical signal processing: estimation theory , 1993 .

[9]  S C Strother,et al.  Comparison of matched BOLD and FAIR 4.0T-fMRI with [15O]water PET brain volumes. , 1999, Medical physics.

[10]  Karl J. Friston,et al.  Combining Spatial Extent and Peak Intensity to Test for Activations in Functional Imaging , 1997, NeuroImage.

[11]  J. Berger.  Statistical Decision Theory and Bayesian Analysis , 1988 .

[12]  Karl J. Friston,et al.  Assessing the significance of focal activations using their spatial extent , 1994, Human brain mapping.

[13]  Jonathan Marchini,et al.  Comparing methods of analyzing fMRI statistical parametric maps , 2004, NeuroImage.

[14]  Xavier Descombes,et al.  fMRI Signal Restoration Using a Spatio-Temporal Markov Random Field Preserving Transitions , 1998, NeuroImage.

[15]  Petros Dellaportas,et al.  An Introduction to MCMC , 2003 .

[16]  Aaas News,et al.  Book Reviews , 1893, Buffalo Medical and Surgical Journal.

[17]  Miles N Wernick,et al.  Methods to detect objects in photon-limited images. , 2006, Journal of the Optical Society of America. A, Optics, image science, and vision.

[18]  Karl J. Friston,et al.  A unified statistical approach for determining significant signals in images of cerebral activation , 1996, Human brain mapping.

[19]  Karl J. Friston,et al.  Classical and Bayesian Inference in Neuroimaging: Theory , 2002, NeuroImage.

[20]  K J Worsley,et al.  An overview and some new developments in the statistical analysis of PET and fMRI data , 1997, Human brain mapping.

[21]  Ahmad Abu-Naser,et al.  Methods of Detecting Objects in Photon-Limited Images , 2005 .

[22]  Heekuck Oh,et al.  Neural Networks for Pattern Recognition , 1993, Adv. Comput..

[23]  Alan C. Evans,et al.  A Three-Dimensional Statistical Analysis for CBF Activation Studies in Human Brain , 1992, Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism.

[24]  Karl J. Friston,et al.  Posterior probability maps and SPMs , 2003, NeuroImage.

[25]  B. Everitt,et al.  Mixture model mapping of brain activation in functional magnetic resonance images , 1999, Human brain mapping.

[26]  Thomas E. Nichols,et al.  Statistical limitations in functional neuroimaging. I. Non-inferential methods and statistical models. , 1999, Philosophical transactions of the Royal Society of London. Series B, Biological sciences.

[27]  C. Geyer,et al.  Simulation Procedures and Likelihood Inference for Spatial Point Processes , 1994 .

[28]  R. Adler.  The Geometry of Random Fields , 2009 .

[29]  N. L. Johnson,et al.  Multivariate Analysis , 1958, Nature.

[30]  R. Adler,et al.  The Geometry of Random Fields , 1982 .

[31]  Michael E. Tipping,et al.  Fast Marginal Likelihood Maximisation for Sparse Bayesian Models , 2003 .

[32]  Karl J. Friston.  Imaging neuroscience: principles or maps? , 1998, Proceedings of the National Academy of Sciences of the United States of America.

[33]  Alan C. Evans,et al.  Searching scale space for activation in PET images , 1996, Human brain mapping.

[34]  Guillaume Stawinski,et al.  Reversible jump Markov chain Monte Carlo for Bayesian deconvolution of point sources , 1998, Optics & Photonics.

[35]  Nikolas P. Galatsanos,et al.  Relevance vector machine analysis of functional neuroimages , 2004, 2004 2nd IEEE International Symposium on Biomedical Imaging: Nano to Macro (IEEE Cat No. 04EX821).

[36]  N. Hartvig,et al.  A Stochastic Geometry Model for Functional Magnetic Resonance Images , 2002 .

[37]  R. Nowak,et al.  Generalized likelihood ratio detection for fMRI using complex data , 1999, IEEE Transactions on Medical Imaging.

[38]  Fuqiang Zhao,et al.  Spatial specificity of cerebral blood volume-weighted fMRI responses at columnar resolution , 2005, NeuroImage.

[39]  Christopher M. Bishop,et al.  Neural networks for pattern recognition , 1995 .

[40]  Carl E. Rasmussen,et al.  Pruning from Adaptive Regularization , 1994, Neural Computation.

[41]  S. Strother,et al.  Quantitative Comparisons of Image Registration Techniques Based on High‐Resolution MRI of the Brain , 1994, Journal of computer assisted tomography.

[42]  Jan Sijbers,et al.  Generalized likelihood ratio tests for complex fMRI data: a Simulation study , 2005, IEEE Transactions on Medical Imaging.

[43]  N V Hartvig,et al.  Spatial mixture modeling of fMRI data , 2000, Human brain mapping.

[44]  Stephen C. Strother,et al.  An evaluation of methods for detecting brain activations from functional neuroimages , 2002, Artif. Intell. Medicine.

[45]  P. Green.  Reversible jump Markov chain Monte Carlo computation and Bayesian model determination , 1995 .

[46]  Lars Kai Hansen,et al.  The Quantitative Evaluation of Functional Neuroimaging Experiments: The NPAIRS Data Analysis Framework , 2000, NeuroImage.