Learning spatial filters from EEG signals with Graph Signal Processing methods.

In this paper, we propose to learn a spatial filter directly from Electroencephalography (EEG) signals using graph signal processing tools. We combine a graph learning algorithm with a high-pass graph filter to remove spatially large signals from the raw data. This approach increases topographical localization, and attenuates volume-conducted features. We empirically show that our method gives similar results that the surface Laplacian in the noiseless case while being more robust to noise or defective electrodes.Clinical relevance- The proposed method is an alternative to the surface Laplacian filter that is commonly used for processing EEG signals. It could be used in cases where this standard approach does not provide satisfying results (low signal-to-noise ratios due to a low number of epochs, defective electrodes). This could be particularly interesting in case of an electrode defect, as it can happen in clinical practice.

Quantitative assessment of consciousness during anesthesia without EEG data.

Assessing the depth of anesthesia (DoA) is a daily challenge for anesthesiologists. The best assessment of the depth of anesthesia is commonly thought to be the one made by the doctor in charge of the patient. This evaluation is based on the integration of several parameters including epidemiological, pharmacological and physiological data. By developing a protocol to record synchronously all these parameters we aim at having this evaluation made by an algorithm. Our hypothesis was that the standard parameters recorded during anesthesia (without EEG) could provide a good insight into the consciousness level of the patient. We developed a complete solution for high-resolution longitudinal follow-up of patients during anesthesia. A Hidden Markov Model (HMM) was trained on the database in order to predict and assess states based on four physiological variables that were adjusted to the consciousness level: Heart Rate (HR), Mean Blood Pressure (MeanBP) Respiratory Rate (RR), and AA Inspiratory Concentration (AAFi) all without using EEG recordings. Patients undergoing general anesthesia for hernial inguinal repair were included after informed consent. The algorithm was tested on 30 patients. The percentage of error to identify the actual state among Awake, LOC, Anesthesia, ROC and Emergence was 18%. This protocol constitutes the very first step on the way towards a multimodal approach of anesthesia. The fact that our first classifier already demonstrated a good predictability is very encouraging for the future. Indeed, this first model was merely a proof of concept to encourage research ways in the field of machine learning and anesthesia.

Apprenticeship Learning for a Predictive State Representation of Anesthesia.

OBJECTIVE: In this paper, we present an original decision support algorithm to assist the anesthesiologists delivery of drugs to maintain the optimal Depth of Anesthesia (DoA). METHODS: Derived from a Transform Predictive State Representation algorithm (TPSR), our model learned by observing anesthesiologists in practice. This framework, known as apprenticeship learning, is particularly useful in the medical field as it is not based on an exploratory process - a prohibitive behavior in healthcare. The model only relied on the four commonly monitored variables: Heart Rate (HR), the Mean Blood Pressure (MBP), the Respiratory Rate (RR) and the concentration of anesthetic drug (AAFi). RESULTS: Thirty-one patients have been included. The performances of the model is analyzed with metrics derived from the Hamming distance and cross entropy. They demonstrated that low rank dynamical system had the best performances on both predictions and simulations. Then, a confrontation of our agent to a panel of six real anesthesiologists demonstrated that 95.7% of the actions were valid. CONCLUSION: These results strongly support the hypothesis that TPSR based models convincingly embed the behavior of anesthesiologists including only four variables that are commonly assessed to predict the DoA. SIGNIFICANCE: The proposed novel approach could be of great help for clinicians by improving the fine tuning of the DoA. Furthermore, the possibility to predict the evolutions of the variables would help preventing side effects such as low blood pressure. A tool that could autonomously help the anesthesiologist would thus improve safety-level in the surgical room.

Selection of the Best Electroencephalogram Channel to Predict the Depth of Anesthesia.

Precise cerebral dynamics of action of the anesthetics are a challenge for neuroscientists. This explains why there is no gold standard for monitoring the Depth of Anesthesia (DoA) and why experimental studies may use several electroencephalogram (EEG) channels, ranging from 2 to 128 EEG-channels. Our study aimed at finding the scalp area providing valuable information about brain activity under general anesthesia (GA) to select the more optimal EEG channel to characterized the DoA. We included 30 patients undergoing elective, minor surgery under GA and used a 32-channel EEG to record their electrical brain activity. In addition, we recorded their physiological parameters and the BIS monitor. Each individual EEG channel data were processed to test their ability to differentiate awake from asleep states. Due to strict quality criteria adopted for the EEG data and the difficulties of the real-life setting of the study, only 8 patients recordings were taken into consideration in the final analysis. Using 2 classification algorithms, we identified the optimal channels to discriminate between asleep and awake states: the frontal and temporal F8 and T7 were retrieved as being the two bests channels to monitor DoA. Then, using only data from the F8 channel, we tried to minimize the number of features required to discriminate between the awake and asleep state. The best algorithm turned out to be the Gaussian Naïve Bayes (GNB) requiring only 5 features (Area Under the ROC Curve - AUC- of 0.93 ± 0.04). This finding may pave the way to improve the assessment of DoA by combining one EEG channel recordings with a multimodal physiological monitoring of the brain state under GA. Further work is needed to see if these results may be valid to asses the depth of sedation in ICU.