Fully-automated sleep staging: multicenter validation of a generalizable deep neural network for Parkinson's disease and isolated REM sleep behavior disorder

2026-02-10

AI Summary

This study addresses the challenge of automated sleep staging in neurodegenerative disorders—such as Parkinson’s disease and isolated REM sleep behavior disorder (iRBD)—where abnormal EEG patterns and fragmented sleep hinder reliable manual scoring and large-scale screening. Building upon the U-Sleep deep neural network, the authors first pretrain the model on a large, multicenter dataset of non-neurodegenerative individuals and then fine-tune it via transfer learning on cohorts with Parkinson’s disease and iRBD. A confidence-thresholding strategy is further introduced to optimize REM sleep detection. This approach represents the first validation of a general-purpose sleep staging model across multicenter neurodegenerative populations, significantly improving performance: on an independent test set, inter-scorer agreement (Cohen’s κ) increased from 0.60 to 0.64 (mean) and 0.69 (median), REM detection accuracy rose from 85% to 95.5%, and ≥5 minutes of valid REM sleep were retained in 95% of participants.

Technology Category

Cognitive Modeling & Cognitive Systems: Neural Spike CodingIntelligent Robots: Multimodal Perception & Sensor FusionMachine Learning: Neuro-Symbolic Learning

Application Category

Web Mining and Content Analysis: Large pretrained models with web dataGraph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphsSearch and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Isolated REM sleep behavior disorder (iRBD) is a key prodromal marker of Parkinson's disease (PD), and video-polysomnography (vPSG) remains the diagnostic gold standard. However, manual sleep staging is particularly challenging in neurodegenerative diseases due to EEG abnormalities and fragmented sleep, making PSG assessments a bottleneck for deploying new RBD screening technologies at scale. We adapted U-Sleep, a deep neural network, for generalizable sleep staging in PD and iRBD. A pretrained U-Sleep model, based on a large, multisite non-neurodegenerative dataset (PUB; 19,236 PSGs across 12 sites), was fine-tuned on research datasets from two centers (Lundbeck Foundation Parkinson's Disease Research Center (PACE) and the Cologne-Bonn Cohort (CBC); 112 PD, 138 iRBD, 89 age-matched controls. The resulting model was evaluated on an independent dataset from the Danish Center for Sleep Medicine (DCSM; 81 PD, 36 iRBD, 87 sleep-clinic controls). A subset of PSGs with low agreement between the human rater and the model (Cohen's $\kappa$<0.6) was re-scored by a second blinded human rater to identify sources of disagreement. Finally, we applied confidence-based thresholds to optimize REM sleep staging. The pretrained model achieved mean $\kappa$ = 0.81 in PUB, but $\kappa$ = 0.66 when applied directly to PACE/CBC. By fine-tuning the model, we developed a generalized model with $\kappa$ = 0.74 on PACE/CBC (p<0.001 vs. the pretrained model). In DCSM, mean and median $\kappa$ increased from 0.60 to 0.64 (p<0.001) and 0.64 to 0.69 (p<0.001), respectively. In the interrater study, PSGs with low agreement between the model and the initial scorer showed similarly low agreement between human scorers. Applying a confidence threshold increased the proportion of correctly identified REM sleep epochs from 85% to 95.5%, while preserving sufficient (>5 min) REM sleep for 95% of subjects.

Problem

• sleep staging
• Parkinson's disease
• REM sleep behavior disorder
• polysomnography
• neurodegenerative diseases

Innovation

• automated sleep staging
• deep neural network
• Parkinson's disease
• REM sleep behavior disorder
• confidence-based thresholding