Abstract— This paper provides a Bayesian method of analyzing functional magnetic resonance imaging (fMRI) data. Usually fMRI signals are noisy and need efficient algorithms to estimate or detect the signals accurately. Using a Bayesian frame work we have used two different priors: sparsity and clusterdness in the fMRI data by using a general linear model (GLM), which is used as a main tool in fMRI studies. This enhances the effectiveness of the model to help analyze the data better. So in this work we have built the Bayesian framework needed first. Then, we have applied our analysis on synthetic data that we made and that are well known, and the results show that clustered compressive sampling has given better results compared to the using of only sparse prior and/ or to the analysis without considering the two priors. Later we have applied it on fMRI data and the results are much better in terms of signal to noise ratio (SNR) and intensity of images.
Index Terms— Bayesian framework, sparse prior, clustered prior, posterior, MAP, compressive sensing, LASSO, clustered LASSO, GLM, fMRI data.
S. A. Tesfamicael is with Sør-Trondlag University College (HIST-ALT). He is also with the department of Electronics and Telecommunication (IET) at the Norwegian University of Science and Technology (NTNU), Trondheim, Norway (e-mail: solomont@hist.no, tesfamic@iet.ntnu.no).
F. Barzideh is with the Norwegian University of Science and Technology (NTNU), Trondheim, Norway (e-mail: faraz.barzideh@gmail.com ).
[PDF]
Cite: Solomon A. Tesfamicael and Faraz Barzideh, " Clustered Compressed Sensing in fMRI Data Analysis Using a Bayesian Framework," International Journal of Information and Electronics Engineering vol. 4, no. 2, pp. 74-80, 2014.