Inner-speech marker detection with EEG and possible extension to auditory-verbal hallucinations
Résumé
Auditory verbal hallucinations (AVH) are one of the most disabling symptoms for people suffering from psychiatric disorders like schizophrenia (SCZ). From a brain activity point of view, these experiences are pretty similar to "inner speech", a task that each of us regularly performs when we mentally generate sentences without speaking them aloud. In this work, we investigate to what extent detecting inner-speech from electroencephalographic (EEG) signals is feasible, with the aim of identifying the periods with and without such events. If this first objective can be achieved, it will constitute a first milestone towards the detection of hallucinatory periods in schizophrenia patients using a similar approach. To quantify the accuracy of detection, we started from already existing databases obtained from literature. We used a Filter Bank Common Spatial Patterns (FBCSP) algorithm to classify inner speech versus rest periods, and observed that the within subjects classifier’s performances were significantly higher than chance level. Since FBCSPs use spatial filters, they can be visualized to unravel the spatial discrepancies in neural activation between two tasks. We also confirmed the interpretability of the method with more obvious classifications as controls (e.g. Visual versus non visual periods). Following these encouraging results with inner-speech, we developed a protocol to record EEG during AVH in SCZ patients. One of the biggest challenges in properly detecting AVH signature, is to define a robust labelling of the hallucinatory/non-hallucinatory periods and to minimize artifacts that could potentially be correlated with the onset or offset of the episodes. In order to circumvent this issue, we designed a protocol that does not rely on subject-report during the episode, but only at the end, asking the subject to estimate the event’s duration and report it by pressing a button during approximately the same time span. We are currently testing the feasibility and reliability of such a paradigm in healthy individuals using an inner-speech task, before generalizing it to SCZ patients.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |