Paradigm free mapping: detection and characterization of single trial fMRI BOLD responses without prior stimulus information
Publication Date
October 15, 2010
Creator
Abstract
The increased contrast to noise ratio available at Ultrahigh (7T) Magnetic Resonance Imaging (MRI) allows mapping in space and time the brain's response to single trial events with functional MRI (fMRI) based on the Blood Oxygenation Level Dependent (BOLD) contrast. This thesis primarily concerns with the development of techniques to detect and characterize single trial event-related BOLD responses without prior paradigm information, Paradigm Free Mapping, and assess variations in BOLD sensitivity across brain regions at high field fMRI.
Based on a linear haemodynamic response model, Paradigm Free Mapping (PFM) techniques rely on the deconvolution of the neuronal-related signal driving the BOLD effect using regularized least squares estimators. The first approach, named PFM, builds on the ridge regression estimator and spatio-temporal t-statistics to detect statistically significant changes in the deconvolved fMRI signal. The second method, Sparse PFM, benefits from subset selection features of the LASSO and Dantzig Selector estimators that automatically detect the single trial BOLD responses by promoting a sparse deconvolution of the signal. The third technique, Multicomponent PFM, exploits further the benefits of sparse estimation to decompose the fMRI signal into a haemodynamical component and a baseline component using the morphological component analysis algorithm.
These techniques were evaluated in simulations and experimental fMRI datasets, and the results were compared with well-established fMRI analysis methods. In particular, the methods developed here enabled the detection of single trial BOLD responses to visually-cued and self-paced finger tapping responses without prior information of the events. The potential application of Sparse PFM to identify interictal discharges in idiopathic generalized epilepsy was also investigated. Furthermore, Multicomponent PFM allowed us to extract cardiac and respiratory fluctuations of the signal without the need of physiological monitoring.
To sum up, this work demonstrates the feasibility to do single trial fMRI analysis without prior stimulus or physiological information using PFM techniques.
Item Type
ethesis
Thesis Type
PhD
Supervisors
Subjects (LC)
Associated Schools / Departments
School of Computer Science (UK)
eprints ID
11488
UoN Repository URI
Except where otherwise noted, this item's license is described as
File(s)![Thumbnail Image]()
Name
thesis_CesarCaballero_final.pdf
Type
Full-text
Size
9.02 MB
Format
Adobe PDF
Checksum ()