Institutional Repository
Technical University of Crete
EN  |  EL

Search

Browse

My Space

EEG source localization on different realistic brain anatomies using deep learning techniques

Kolomvaki Afroditi

Full record


URI: http://purl.tuc.gr/dl/dias/769A274A-D943-4C61-9DF3-BF88EE137E5D
Year 2024
Type of Item Diploma Work
License
Details
Bibliographic Citation Afroditi Kolomvaki, "EEG source localization on different realistic brain anatomies using deep learning techniques", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2024 https://doi.org/10.26233/heallink.tuc.99945
Appears in Collections

Summary

Electroencephalography (EEG) is a technique used to record the electrical activity of the brain and is commonly used in medical settings to diagnose and monitor conditions such as epilepsy, sleep disorders, and brain injuries. It involves placing electrodes on the scalp, which detect the electrical signals produced by the brain. One of the main challenges in interpreting these EEG signals is to identify the underlying neural sources responsible for generating the measured scalp potentials. This is known as EEG source analysis or EEG inverse problem. Various numerical methods exist to address this inverse problem, but they require considerable computational time and often depend heavily on prior assumptions. Recently, neural networks have been suggested as a solution, but their training often relies on suboptimal forward modeling and they struggle to localize EEG signals across different brain anatomies and multiple brain electrical activations. In this study, we introduce a Convolutional Neural Network (CNN) architecture that is independent of the brain source space model and trained using realistic head models calibrated for skull conductivity. It is capable of solving the inverse problem for up to three active brain sources on different realistic brain anatomies. The results indicate that our CNN outperforms traditional numerical methods like sLORETA.

Available Files

Services

Statistics