AnalysisSpectralEEG

The alpha-band of the EEG appears as a continuous line in the frequency domain, however wavelet analysis reveals that it is in fact fragmented. We propose to model this fragmentation using stochastic differential equation and extract patterns to predict in real-time from the EEG signal, possible transitions into deep brain states. We propose to develop a stochastic model based on two-dimensional Ornstein-Ulhenbeck processes that can reproduce a fragmented oscillatory band around a main frequency. This model will then be used to compare patterns that can be extracted from real-time segmentation approach of the EEG signal to extract predictive motifs. Using a spectral analysis based on spectrogram, wavelet and EMD decompositions, we will study the time-frequency fragmentation of the dominant frequency band of the EEG signal. To better characterize the brain states and improve both the signal processing and modeling part, we will investigate how anaesthetics can modify neuronal network activity and induce depression of activity: we will record neuronal rhythms in ex-vivo thalamocortical slices by using multi-electrode array (MEA) techniques. We will also use an in vivo physiological approach by inserting Neuropixels probes to measure simultaneously hundreds of parallel synaptic and evaluate synaptic plasticity during anesthesia by targeting the local thalamo-cortical axons. We will use behaviour test in rodent to correlate the statistics contained in the EEG with possible post-anesthetic complications. The real-time algorithms will then be developed in collaboration with a start-up company dedicated to predict the brain dynamics combining transient motifs of the EEG signal to classify brain states. Finally, we will focus on predicting in real-time the evolution of the dose to control the depth of anesthesia that we will test during rodents and human surgery within clinical tests. To conclude, the present method will give a novel tool to predict Brain during general anesthesia.