Deep Neural Network to Map Human Brain Activity from EEG to fMRI
Organization
Computational Neuroscience Lab
Abstract
There are different techniques to measure human brain activity. One of them, EEG, can show precisely *when* the activity occurred in time, but provides almost no information about *where* inside the brain the activity happened. Another technology, fMRI, is exactly the opposite: it shows precise 3D location of the source of activity, but cannot tell when exactly that activity took place.
We have a dataset where both EEG and fMRI data were recorded simultaneously. That allows to try to train a network (or another ML model) to predict location of fMRI activity from EEG recording.
The outcome of this work will be a model, that gets piece of EEG data as input and tells which part of brain this activity is most likely to be originating from. This would be important contribution to computational neuroscience, field of source localization and, if done well, might yield publishable result.
We have a dataset where both EEG and fMRI data were recorded simultaneously. That allows to try to train a network (or another ML model) to predict location of fMRI activity from EEG recording.
The outcome of this work will be a model, that gets piece of EEG data as input and tells which part of brain this activity is most likely to be originating from. This would be important contribution to computational neuroscience, field of source localization and, if done well, might yield publishable result.
Graduation Theses defence year
2017-2018
Supervisor
Ilya Kuzovkin
Spoken language (s)
Estonian, English
Requirements for candidates
Understanding of machine learning concepts, familiarity with neural networks, good Python/numpy programming skills, scientific attitude, ability to work independently (things will fail, you need to be able to think what to try next).
Level
Masters
Keywords
Application of contact
Name
Ilya Kuzovkin
Phone
E-mail