Thesis Project Form

We’ve observed that REM sleep is a strongly desynchronized state, associated with reduced large-scale network dynamics (such as phase synchronization and cross-frequency coupling) compared to NREM sleep and wakefulness, and we hypothesize that these reductions contribute to its protective role against epileptic activity. This thesis focuses on investigating whether the reduction in large-scale network dynamics that characterizes REM sleep is associated with a lower rate of interictal spikes in stereo-EEG (SEEG).

In the first phase, the student will identify and select an automatic spike detection approach that is most suitable for SEEG recordings from patients with drug-resistant epilepsy. This will include a review of existing spike detection methods, the selection of one or more candidate algorithms, and their implementation or adaptation in Python, followed by a quantitative evaluation of their performance on our SEEG data.
In the second phase, the selected spike detection method will be applied to an SEEG dataset for which network metrics (e.g. phase-locking value, phase-amplitude coupling, bistability index) are already computed for REM, N2, N3 and wakefulness. The student will quantify the IEDs in each state and examine their link to these network measures, with a specific focus on whether the REM-associated reduction in network dynamics is associated to a reduction in spike rate.

The student will learn how to analyze intracranial EEG (SEEG) data, implement and test automatic spike-detection algorithms in Python, and work with connectivity and brain criticality measures. They will gain basic experience in statistical analysis and data visualization and improve their understanding of sleep and epilepsy and their bidirectional knowledge.