Electroencephalography (EEG) recordings during sleep are an important source of information about brain activity, enabling characterization of many cognitive processes related to sleep and, more broadly, to an individual's health. The signals from these recordings result from the contributions of multiple spontaneous processes, in particular, transient rhythms with well defined spectral characteristics. They are superimposed with a background signal, known as ‘aperiodic activity’. The latter is currently attracting growing interest, as it is being evaluated more precisely to establish its physiological relevance, but it also constitutes an obstacle to characterizing the intrinsic properties of the rhythms themselves.
Current methods for separating rhythmic and aperiodic activity are limited to the spectral domain and are typically applied to signals recorded from a single electrode. In this thesis, numerical tools were developed to characterize rhythmic and aperiodic activity across multiple spatial scales, ranging from intracranial recordings to sequences of scalp topographies. The first contribution consisted of developing a signal-processing method that extracts a rhythmic signal in which the aperiodic activity has been filtered out. The validation of this tool was carried out using intracranial recordings during NREM (non–rapid eye movement) sleep. The second contribution involved extending the application of this tool to scalp EEG recordings, which is necessary given that this modality is non-invasive and widely used in clinical settings. Together, these two contributions provide a tool for obtaining an inventory of rhythms without aperiodic activity. Finally, the last contribution consisted of measuring aperiodic activity in scalp topographies using microstate analysis.
Taken together, these tools provide an innovative methodological framework for disentangling and analyzing the rhythmic and aperiodic components of EEG signals. They are intended to be integrated into clinical settings to assess and characterize various pathologies and to identify reliable biomarkers.
| Date | 31 Mar 2026 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Jean-Marc Lina (Supervisor) & Rébecca Robillard (Co-supervisor) |
|---|
Foti, M.-C. (Author),
Lina (Supervisor) & Robillard (Co-supervisor),
31 Mar 2026Student thesis: Doctoral thesis › Doctorate in Engineering: Engineering