AnySleep: a channel-agnostic deep learning system for high-resolution sleep staging in multi-center cohorts

2025-12-16

AI Summary

Traditional sleep staging relies on labor-intensive manual annotation and faces significant challenges due to high heterogeneity across multi-center polysomnography (PSG) data—spanning electrode configurations, subject demographics, and temporal resolutions—hindering biomarker discovery and cross-site generalization. To address this, we propose AnySleep, the first channel-agnostic deep learning framework designed specifically for heterogeneous multi-center PSG data. It supports arbitrary EEG/EOG channel combinations and configurable temporal resolutions—including sub-30-second granularity. Our key contributions are: (i) the first channel-agnostic modeling paradigm with dynamic sub-30-second staging capability; (ii) an end-to-end temporal modeling architecture integrating multi-center self-supervised pretraining and cross-dataset adversarial alignment; and (iii) state-of-the-art performance across 21 datasets (19,000+ nights, ~200,000 hours), outperforming mainstream baselines in standard 30-second staging and significantly improving detection of micro-arousals, age/gender-related patterns, and sleep apnea severity prediction at sub-30-second resolution.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Hardware-aware ML
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Sleep is essential for good health throughout our lives, yet studying its dynamics requires manual sleep staging, a labor-intensive step in sleep research and clinical care. Across centers, polysomnography (PSG) recordings are traditionally scored in 30-s epochs for pragmatic, not physiological, reasons and can vary considerably in electrode count, montage, and subject characteristics. These constraints present challenges in conducting harmonized multi-center sleep studies and discovering novel, robust biomarkers on shorter timescales. Here, we present AnySleep, a deep neural network model that uses any electroencephalography (EEG) or electrooculography (EOG) data to score sleep at adjustable temporal resolutions. We trained and validated the model on over 19,000 overnight recordings from 21 datasets collected across multiple clinics, spanning nearly 200,000 hours of EEG and EOG data, to promote robust generalization across sites. The model attains state-of-the-art performance and surpasses or equals established baselines at 30-s epochs. Performance improves as more channels are provided, yet remains strong when EOG is absent or when only EOG or single EEG derivations (frontal, central, or occipital) are available. On sub-30-s timescales, the model captures short wake intrusions consistent with arousals and improves prediction of physiological characteristics (age, sex) and pathophysiological conditions (sleep apnea), relative to standard 30-s scoring. We make the model publicly available to facilitate large-scale studies with heterogeneous electrode setups and to accelerate the discovery of novel biomarkers in sleep.

Problem

• Automates manual sleep staging using EEG/EOG data
• Handles variable electrode setups across multi-center studies
• Enables high-resolution sleep analysis for biomarker discovery

Innovation

• Channel-agnostic deep learning for sleep staging
• Adjustable temporal resolution scoring system
• Trained on multi-center data for generalization