Title

IMPROVEMENT OF TEMPORAL RESOLUTION OF FMRI DATA FOR BRAIN DECODING

Abstract

Abstract

In this study, we aim to increase the accuracy of the mapping between
the states of the brain and problem-solving phases namely planning and
execution. To create a computational model to generate the mapping, an
fMRI dataset obtained from subjects solving the Tower of London
problem has been used. fMRI data is suitable for this problem as it provides
regional and time-varying changes in brain metabolism. However,
developing the model using fMRI data is not trivial. Generally, fMRI data has a
very large feature vector while having a small sample size due to the
scanner limitations. We propose two methods to overcome these limitations and
increase the mapping performance. Both methods have a preliminary
stage where we perform preprocessing. Preprocessing stage includes feature
selection and whitening. The proposed methods are built with
polynomial regression and neural networks utilising the spatial and temporal nature of the data.

Supervisor(s)

Supervisor(s)

EMEL VAROL

Date and Location

Date and Location

2022-02-10 14:00:00

Category

Category

MSc_Thesis