Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Analysis of electroencephalography in epilepsy after transcranial brain stimulation using connectivity models and machine learning

Tsipouraki Alexandra

Full record


URI: http://purl.tuc.gr/dl/dias/0C79044D-49DB-493F-A07A-C2CF307F59BB
Year 2024
Type of Item Diploma Work
License
Details
Bibliographic Citation Alexandra Tsipouraki, "Analysis of electroencephalography in epilepsy after transcranial brain stimulation using connectivity models and machine learning", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2024 https://doi.org/10.26233/heallink.tuc.99099
Appears in Collections

Summary

Epilepsy constitutes a neurological disorder affecting approximately 50 million individuals globally, significantly impacting their quality of life.Conventionally, epilepsy symptoms are managed through the administration of antiepileptic drugs or, where feasible, through surgical intervention. However, the frequent cases of refractory focal epilepsy have highlighted the necessity for new, personalized therapeutic approaches. Among these, transcranial Direct Current Stimulation (tDCS) has emerged as a promising potential solution. The present work, focuses on the study of EEG recordings deriving from a proof-of-principle N-of-1 trial study, whose scope was to investigate the effects of multi-channel (mc-) tDCS application on a patient with refractory focal epilepsy. A double-blind sham-controlled stimulation experiment was conducted in a two-week long stimulation trial. Distributed Constrained Maximum Intensity (D-CMI)-based-mc-tDCS and sham stimulation were applied twice every week-day for 20 minutes each. EEG data, was recorded for 1 hour before and after stimulation. Experts, marked a highly significant reduction in interictal spike frequency after the stimulation process, while this was not the case for sham.Our purpose, is to evaluate EEG connectivity patterns, using generalized Partial Directed Coherence (gPDC) before and after stimulation and sham procedures accordingly. The raw EEG recordings are segmented into 3-second long sub-signals, to which then gPDC is applied and studied. We further proceed to the extraction of connectivity and statistical features from this analysis, and provide this information to Machine Learning models, in order to verify and validate our connectivity findings. The final results are promising; the connectivity analysis performed on the EEG data validated the results which had already derived from the trial, the epileptogenic zone was confirmed, as also was the reduction of IEDs after the tDCS. Finally, the ML models' results validated the robustness of our connectivity study, highlighted by the decrease in class separability after the stimulation process but not after sham. Τhis research, contributes to gaining a deeper understanding of the neural mechanisms underlying epilepsy, utilizing a non-invasive modality, as is EEG. The ability to gather and analyze more extensive EEG data sets over longer periods, can enhance in the future the depth and reliability of our findings, offering a richer understanding of the effects of Transcranial Direct Current Stimulation (tDCS) on epilepsy.

Available Files

Services

Statistics