Encoding Biomarkers of Consciousness Under Conditions of Abnormal Cortical Dynamics

Consciousness and its absence can be assessed and quantified through different techniques. Sleep-stage classification via electroencephalography (EEG) signal analysis is the most common and effective method for this purpose. Recently, the research moved from a spectral investigation of the EEG to a complexity study. Entropy features showed to be more robust biomarkers of consciousness than frequency-based measures. To further contribute to this discussion, we utilized a Deep Learning (DL) derived function, Gradient-weighted Class Activation Mapping (Grad-CAM), to predict if this generalizability is found also in the DL's unbiased feature selection. To do so, we exploited a semi-automatic Machine Learning (ML) classification pipeline to compare the performances obtained with the original EEG and the Grad-CAM. The generalizability of the identified features was addressed by using Angelman Syndrome (AS) patients as training set and Dup15q11.2-13.1 syndrome (Dup15q) patients as test set. These disorders present opposite abnormal EEG phenotypes and are not suitable for canonical spectral-based sleep scoring. We ultimately found that the DL network recognizes entropic features as meaningful characteristics of the input signal and enhances them in the formulation of the Grad-CAM. Conversely, spectral features undergo a suppression in the final encoding of the DL classifier. Furthermore, the improvements provided by the application of Grad-CAM allowed for a boost in the ML classification performances for the systems trained over entropy features. To bring further significance to these results, a neurotypical EEG dataset should be included. This would allow for a direct comparison of the ML and DL performances while bringing more significance to the entropic analysis adaptability. Furthermore, the workflow developed in this study has the potential to be standardized and be applied in the field of explainability of DL classification outcomes.