Abstract:Despite its complexity, whole-brain neural activity spontaneously organizes into a limited range of states, indicating that a low-dimensional representation or embedding might be sufficient to capture much of its macroscale dynamics. Clinical practice makes use of this notion by classifying EEG into discrete states, for example in the wake-sleep cycle or along the ictal- interictal continuum. Such classification into distinct classes has clinical value and simplifies EEG interpretation (by humans and, more recently, also by artificial intelligence [AI]) but misses subtle differences, mixtures and transitions between states. In other words, it does not adequately reflect the true underlying, low-dimensional continuum of states. The lack of a continuum EEG map limits more realistic characterizations of EEG states and challenges prognostication in patients with disorders of consciousness due to their diverse patterns, etiologies and underlying pathophysiologies. We here use data across the health-disease continuum (wake [W], sleep [REM, N1, N2, N3], ictal-interictal-continuum [lateralized and generalized periodic discharges (LPD, GPD), lateralized and generalized rhythmic delta activity (LRDA, GRDA)], seizures [SZ], burst suppression [BS]; 20,043 patients, 288,986 EEG segments) and a deep neural network that optimises a low-dimensional embedding space. This universal map of EEG (UM-EEG) is highly semantic, confined and captures meaningful distances as well as transitions between EEG states. Classification and clustering of unseen EEG are easily implemented using standard techniques on the map that match performance of current AI algorithms while extending it to the currently largest set of classes across the full health-disease continuum (mean AUROCs and 95% CI one-vs-all classification: W 0.94 [0.94, 0.95], N1 0.85 [0.85, 0.86], REM 0.92 [0.91, 0.92], N2 0.91 [0.90, 0.91], N3 0.98 [0.97, 0.98], GRDA [0.97 [0.96, 0.97], LRDA 0.97 [0.96, 0.97], SZ 0.87 [0.83, 0.91], GPD 0.99 [0.98, 0.90], LPD 0.97 [0.97, 0.97)], BS 0.94 [0.93, 0.94]). Long-term EEGs are represented as trajectories in this continuous embedding space which predict outcome (recovery or death) after cardiac arrest with an AUROC of 0.86 (N = 575 patients). For these patients, the map identifies interpretable factors governing prognosis: the distance to healthy states, the variability of state transitions, and time spent in burst suppression. These factors allow personalized monitoring and comparison of patients to themselves over time. In conclusion, UM-EEG presents a novel, comprehensive and physiologically meaningful representation to characterize and classify brain states along the health-disease continuum. It demonstrates immediate clinical utility by enhancing current diagnostic and prognostic capabilities.

Unified Convolutional Sparse Transformer for Disease Diagnosis, Monitoring, Drug Development, and Therapeutic Effect Prediction from EEG Raw Data

Real-Time Epileptic Seizure Detection Based on Deep Learning.

EEG Convolutional Sparse Transformer for Epilepsy Detection and Related Drug Classification.

Simultaneous Human Health Monitoring and Time-Frequency Sparse Representation Using EEG and ECG Signals.

An Extended Computer-Aided Diagnosis System for Multidomain EEG Classification

Automatic diagnostics of electroencephalography pathology based on multi-domain feature fusion

EpilepsyNet: Novel automated detection of epilepsy using transformer model with EEG signals from 121 patient population

Towards a Universal Map of EEG: A semantic, low- dimensional manifold for EEG Classification, Clustering and Prognostication

EENED: End-to-End Neural Epilepsy Detection based on Convolutional Transformer

Automatic Classification of EEG Signals Via Deep Learning.

EEGformer: A transformer-based brain activity classification method using EEG signal

Real-Time Non-Invasive Imaging and Detection of Spreading Depolarizations through EEG: An Ultra-Light Explainable Deep Learning Approach

DOCTer: a novel EEG-based diagnosis framework for disorders of consciousness

CwA-T: A Channelwise AutoEncoder with Transformer for EEG Abnormality Detection

Deep-EEG: An Optimized and Robust Framework and Method for EEG-Based Diagnosis of Epileptic Seizure

Transformer-based Spatial-Temporal Feature Learning for EEG Decoding

ScatterFormer: Locally-Invariant Scattering Transformer for Patient-Independent Multispectral Detection of Epileptiform Discharges

Automatic Analysis of EEGs Using Big Data and Hybrid Deep Learning Architectures

Deep C-LSTM Neural Network for Epileptic Seizure and Tumor Detection Using High-Dimension EEG Signals

A Multi-View Deep Learning Framework for EEG Seizure Detection

EEG-based Patient-Specific Seizure Prediction Based on Spatial-Temporal Hypergraph Attention Transformer