Date of Award

Spring 4-1-2012

Document Type


Degree Name

Master of Science (MS)


Electrical, Computer & Energy Engineering

First Advisor

Francois G. Meyer

Second Advisor

Shannon Hughes

Third Advisor

Tor Wager


In this thesis we propose the use of Sparse Principal Component Analysis to recover neuronal areas in Brain Imaging. We work with functional magnetic resonance imaging data focusing our attention on the dimensionality reduction stage to represent the neuronal activation within the components that contain the maximum temporal variance, tightly related with the hemodynamic response of the neurons. The motivation for the sparse representation follows the idea of the massive modularity definition of the mind where "different neural circuits are specialized for solving adaptive problems''.

The results show that the new sparse low dimensional basis (Eigenbrains) generated through novel unsupervised algorithms, such as Augmented Sparse Principal Component Analysis, perform competitively in terms of neuronal activity prediction. We push the limits of the brain understanding by describing a neuronal network through each Eigenbrain component and defining a prediction neuronal model using a linear combination of them.