Probabilistic sleep staging reveals clinically significant alterations in sleep-stage uncertainty and stage mixing in iRBD and may provide complementary electrophysiological markers of early α-synuclein-related neurodegeneration.
Key Findings
Results
Both iRBD and PD+RBD groups exhibited significantly increased sleep-stage uncertainty and mixing compared with controls.
Study included 104 patients with iRBD, 28 patients with PD and RBD (PD+RBD), and 50 controls
All participants underwent video polysomnography
Sleep-stage uncertainty and mixing were quantified using the U-Sleep 2.0 model
Conventional sleep staging assigns each epoch to a single dominant stage, potentially obscuring within-epoch probabilistic features
Results
REM sleep uncertainty was positively correlated with both RBD symptom severity and motor symptoms in the iRBD group.
REM uncertainty correlated with RBD symptom severity as measured by RBDQ-HK score (r = 0.364, p = 0.003)
REM uncertainty correlated with motor symptoms as measured by UPDRS-III score (r = 0.343, p = 0.005)
Both correlations were assessed using Pearson correlations within the iRBD group (n = 104)
Results
N2 sleep uncertainty was negatively correlated with global cognitive performance and particularly with delayed recall in the iRBD group.
N2 uncertainty was negatively correlated with global cognitive performance as measured by MoCA score (r = -0.345, p = 0.005)
N2 uncertainty showed a stronger negative correlation specifically with delayed recall (r = -0.402, p < 0.001)
Cognition was assessed using the Montreal Cognitive Assessment (MoCA)
Results
N2-REM sleep stage mixing was positively correlated with RBD symptom severity in the iRBD group.
N2-REM mixing correlated with RBD symptom severity as measured by RBDQ-HK score (r = 0.449, p < 0.001)
This was among the strongest clinical correlations observed in the study
N2-REM mixing represents the degree to which epochs contain probabilistic features of both N2 and REM sleep simultaneously
Results
N2-REM sleep stage mixing was negatively correlated with bilateral putaminal dopaminergic uptake on [18F]-DOPA PET imaging in iRBD patients.
N2-REM mixing was negatively correlated with left putaminal dopaminergic uptake (r = -0.492, p = 0.019)
N2-REM mixing was negatively correlated with right putaminal dopaminergic uptake (r = -0.510, p = 0.019)
Nigrostriatal dopaminergic function was assessed with [18F]-DOPA PET in a subgroup of 40 patients with iRBD
These findings link an electrophysiological sleep measure to an established neurobiological marker of α-synucleinopathy
Conclusions
Probabilistic sleep staging metrics may serve as complementary electrophysiological markers of early α-synuclein-related neurodegeneration.
iRBD is recognized as a prodromal stage of α-synucleinopathies including Parkinson's disease
The U-Sleep 2.0 model was used to generate probabilistic sleep stage assignments rather than single-stage epoch labels
Sleep-stage uncertainty and mixing metrics correlated with clinical severity, cognitive function, and dopaminergic imaging
The authors propose these measures could complement existing biomarkers for tracking neurodegeneration in the prodromal phase
What This Means
This research suggests that people with isolated REM sleep behavior disorder (iRBD) — a condition where people physically act out their dreams and which often precedes Parkinson's disease — show measurable disruptions in their sleep architecture that go beyond what standard sleep analysis captures. Using an advanced AI-based tool called U-Sleep 2.0, the researchers analyzed sleep recordings from 104 iRBD patients, 28 Parkinson's disease patients who also had RBD, and 50 healthy controls. Instead of simply labeling each 30-second sleep segment as one sleep stage, the tool calculated how much each segment looked like a mixture of multiple sleep stages — a concept called 'sleep-stage uncertainty and mixing.' Both the iRBD and Parkinson's groups showed significantly more of this uncertainty and mixing compared to healthy controls.
Importantly, these sleep disruption measures were linked to real-world clinical outcomes. Greater uncertainty during REM sleep was associated with worse RBD symptoms and more motor problems. Greater uncertainty during N2 (light non-REM) sleep was linked to worse cognitive performance, especially memory. A specific pattern called N2-REM mixing — where sleep epochs showed features of both light sleep and REM simultaneously — was associated with more severe RBD symptoms and, in a subgroup of 40 patients who underwent brain imaging, was linked to lower dopamine function in the putamen, a brain region critical to movement and affected early in Parkinson's disease.
This research suggests that probabilistic sleep staging, which captures the 'blurring' between sleep stages, could serve as a new type of biomarker for tracking how far along the neurodegenerative process has progressed in people with iRBD. Since iRBD can precede Parkinson's disease by many years, having sensitive early markers is particularly valuable for research into neuroprotective treatments and for monitoring disease progression in clinical trials.
Check Your Own Numbers
Upload your bloodwork. We'll cross-reference your results against this study and 4,700 others.
Liu A, Ma J, Peng Y, Zhang W, Yang Y, Zhang P, et al.. (2026). Sleep stage uncertainty and mixing in isolated REM sleep behavior disorder.. Sleep medicine. https://doi.org/10.1016/j.sleep.2026.109036