Clustered cNMF for fMRI data analysis

This paper introduces a framework for the application of constrained non-negative matrix factorization (cNMF) to estimate the statistically distinct neural responses in a sequence of functional magnetic resonance images (fMRI). While an improved objective function has been defined to make the representation suitable for task-related brain activation detection, in this paper we explore various methods for better detection and efficient computation, placing particular emphasis on the initialization of the constrained NMF algorithm. The K-means algorithm performs this structured initialization and the information theoretic criterion of minimum description length (MDL) is used to estimate the number of clusters. We illustrate the method by a set of functional neuroimages from a motor activation study.

[1]  H. Sebastian Seung,et al.  Learning the parts of objects by non-negative matrix factorization , 1999, Nature.

[2]  James V. Stone.  Independent component analysis: an introduction , 2002, Trends in Cognitive Sciences.

[3]  V D Calhoun,et al.  Spatial and temporal independent component analysis of functional MRI data containing a pair of task‐related waveforms , 2001, Human brain mapping.

[4]  L. K. Hansen,et al.  On Clustering fMRI Time Series , 1999, NeuroImage.

[5]  R. Turner,et al.  Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. , 1992, Proceedings of the National Academy of Sciences of the United States of America.

[6]  Vince D. Calhoun,et al.  Complex ICA for fMRI analysis: performance of several approaches , 2003, 2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03)..

[7]  Alan C. Evans,et al.  A general statistical analysis for fMRI data , 2000, NeuroImage.

[8]  James V. Stone,et al.  Spatiotemporal Independent Component Analysis of Event-Related fMRI Data Using Skewed Probability Density Functions , 2002, NeuroImage.

[9]  H. Sebastian Seung,et al.  Algorithms for Non-negative Matrix Factorization , 2000, NIPS.

[10]  Ola Friman,et al.  Adaptive analysis of functional MRI data , 2003 .

[11]  Jie Tian,et al.  Detecting brain activations by constrained non-negative matrix factorization from task-related BOLD fMRI , 2004, SPIE Medical Imaging.

[12]  S Makeig,et al.  Analysis of fMRI data by blind separation into independent spatial components , 1998, Human brain mapping.

[13]  D. Tank,et al.  Brain magnetic resonance imaging with contrast dependent on blood oxygenation. , 1990, Proceedings of the National Academy of Sciences of the United States of America.

[14]  V D Calhoun,et al.  Independent component analysis of fMRI data in the complex domain , 2002, Magnetic resonance in medicine.

[15]  Thomas Kailath,et al.  Detection of signals by information theoretic criteria , 1985, IEEE Trans. Acoust. Speech Signal Process..

[16]  Kai-Hsiang Chuang,et al.  Model-free functional MRI analysis using Kohonen clustering neural network and fuzzy C-means , 1999, IEEE Transactions on Medical Imaging.

[17]  Stan Z. Li,et al.  Learning spatially localized, parts-based representation , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.

[18]  T. Adali,et al.  Latency (in)sensitive ICA Group independent component analysis of fMRI data in the temporal frequency domain , 2003, NeuroImage.

[19]  Christian F. Beckmann,et al.  PROBABILISTIC ICA FOR FMRI - NOISE AND INFERENCE , 2003 .

[20]  J. MacQueen.  Some methods for classification and analysis of multivariate observations , 1967 .

[21]  Vince D. Calhoun,et al.  ICA of functional MRI data: an overview. , 2003 .