Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Tensor-based fMRI signal processing

Karakasis Paris

Full record

---

Summary

Functional magnetic resonance imaging (fMRI) is one of the most popular methods in studying the human brain. fMRI provides a non-invasive way to measure brain activity, detecting local changes of blood oxygen level density (BOLD) in the brain, over time. The purpose of fMRI signal analysis is the localization of brain areas that are related with particular tasks. The problem of fMRI signal analysis can be considered as a blind source separation problem (BSS) , which is the problem of extracting a set of source signals from a set of mixed signals, without using prior information (or with very little prior information) about the source signals or the mixing process. In this thesis, initially, we study the usage of nonnegative matrix factorization models in BSS problems, assuming the existence of delays in propagation environments with or without echo. Next, we study the usage of tensor factorization models, through PARAFAC and Nonnegative Tensor Factorization models, in BSS problems, as well as how these models can be extended under the assumption of propagation environments without echo. Finally, we use these tensor factorization models in fMRI data analysis. In our analysis, we processed fMRI data from different subjects performing the same tasks (group task related fMRI analysis) and we extracted common activation brain maps as well as common activation time signals.

Available Files

Services