SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures

Keondo Park; Younghoon Na; Yourim Choi; Hyunwoo Ryu; Hyun-Woo Shin; Hyung-Sin Kim

cs.AIFeb 7, 2026

SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures

Authors: Keondo Park, Younghoon Na, Yourim Choi, Hyunwoo Ryu, Hyun-Woo Shin, Hyung-Sin Kim

Organizations: Graduate School of Data Science, Seoul National University, Seoul, South Korea · Department of Biomedical Sciences, Seoul National University College of Medicine, Seoul, Republic of Korea · Obstructive Upper Airway Research (OUaR) Laboratory, Department of Pharmacology, Seoul National University College of Medicine, Seoul, Republic of Korea · Department of Otorhinolaryngology-Head and Neck Surgery, Seoul National University Hospital, Seoul, Republic of Korea

Abstract

While the shift toward unified foundation models has revolutionized many deep learning domains, sleep medicine remains largely restricted to task-specific models that focus on localized micro-structure features. These approaches often neglect the rich, multi-modal context of Polysomnography (PSG) and fail to capture the global macro-structure of a full night's sleep. To address this, we introduce SleepMaMi , a Sleep Foundation Model engineered to master both hour-long sleep architectures and fine-grained signal morphologies. Our framework utilizes a hierarchical dual-encoder design: a Macro-Encoder to model full-night temporal dependencies and a Micro-Encoder to capture short-term characteristics from biosignals. Macro-Encoder is trained via Demographic-Guided Contrastive Learning, which aligns overnight sleep patterns with objective subject metadata, such as age, sex and BMI to refine global representations. Micro-Encoder is optimized via a hybrid Masked Autoencoder (MAE) and multi-modal contrastive objective. Pre-trained on a massive corpus of $>$ 20,000 PSG recordings (158K hours),SleepMaMi outperforms or matches state-of-the-art existing foundation models across a diverse suite of downstream tasks, demonstrating superior generalizability and label-efficient adaptation for clinical sleep analysis.

Figures & tables

Figure 1: Overview of SleepMaMi . Full-night PSG recordings are processed through a hierarchical dual-encoder architecture. The Micro-Encoder captures short-term physiological patterns such as K-complexes, sleep spindles, and respiratory events, while the Macro-Encoder models global sleep architecture including sleep cycles and stage distributions across the entire recording. This design supports diverse downstream tasks ranging from fine-grained event segmentation to subject-level clinical outcomes.

Figure 2: Micro-Encoder design and pretraining method. The Micro-Encoder adopts a private–shared encoder architecture, incorporating patch merging to improve computational efficiency in the shared encoder. The model is trained with a hybrid objective that combines masked autoencoding (reconstruction) and multi-modal contrastive learning to capture sleep micro-structure. For clarity and space constraints, only three representative modalities are shown.

Figure 3: Macro-Encoder design and pretraining method. We utilize bi-directional Mamba layers for efficient long-sequence modeling. Demographic-Guided Contrastive Learning aligns the sleep macro-structure between subjects with objective metadata.

Panel of Figure 4 — Figure 4: Sleep macro-structure variations across demographic groups . Sleep stage distributions over full-night recordings by sex, age (Younger: $<$ 60 yrs; Older: $\geq$ 60 yrs) and BMI. N3 proportion in the early sleep period or REM sleep proportion in later stage varies significantly across groups. These demographic dependent patterns motivate our Demographic-Guided Contrastive Learning objective.

Dataset	Category	Models	Accuracy	Macro-F1	Kappa
SHHS1	Time series	MOMENT-Base	79.4	65.6	70.0
	Foundation model	UniTS	64.2	59.2	53.3
	EEG FM	LaBraM	70.2	54.1	55.7
	Sleep Foundation model	SleepFM-Disease	69.7	56.3	55.9
	Sleep Foundation model	SleepMaMi (Ours)	81.9	74.1	70.0
KISS	Time series	MOMENT-Base	69.8	66.3	59.7

Table 1: Sleep stage classification results. The results demonstrate the effectiveness of SleepMaMi compared to time series and sleep foundation models. The evaluation is done on the test split of SHHS1, KISS datasets, and held-out datasets(CFS, SOF)

Dataset	Category	Models	Accuracy	F1
SHHS1	Time series	MOMENT-Base	73.4	33.4
	Foundation model	UniTS	88.2	48.8
	EEG FM	LaBraM	50.2	20.4
	Sleep	SleepFM-Disease	77.5	39.4
	Foundation model	SleepMaMi (Ours)	88.4	48.5
KISS	Time series	MOMENT-Base	76.0	58.5

Table 2: SDB segmentation results. The results demonstrate SleepMaMi ’s superior capability to capture fine-grained details, compared to time series and sleep foundation models. The evaluation is done on the test split of SHHS1 and KISS datasets. F1 is emphasized as a more robust metric under severe class imbalance in SDB labels.

Models	Disease Outcomes
Models	Angina	CVD death	CHF	CHD death	MI	Stroke
C-Index
SleepFM-Disease	0.632	0.791	0.764	0.781	0.636	0.729
SleepMaMi (Ours)	0.778	0.788	0.793	0.776	0.662	0.718

Table 3: Disease prediction. C-Index is reported on 6 selected diseases from SHHS1 dataset. SleepMaMi shows comparable performance to the state-of-the-art PSG-based disease prediction model.

Panel of Figure 5 — Figure 5: Few shot evaluation. The performance is measured with varying number of fine-tuning samples in Macro-F1 score. SleepMaMi shows label-efficient adaptation to unseen dataset thanks to well-generalized embeddings derived from large-scale pretraining data.

Models	Classification (Acc. $\uparrow$ )		Regression (MAE $\downarrow$ )		Disease
Models	Sleep stage	Sex	Age	AHI	Prediction
Micro-Encoder only	79.8	80.1	9.45	9.69	0.729
SleepMaMi	81.9	88.2	6.73	8.40	0.753

Table 4: Performance comparison between Micro-only and SleepMaMi . The integration of the Macro-Encoder consistently improves performance across all tasks, validating the effectiveness of DGCL in incorporating sleep macro-structure.

Panel of Figure 6 — Figure 6: Macro-embeddings visualization. Per-subject latent embeddings from Macro-Encoder is visualized using U-MAP. Each point represents a single subject’s embeddings where color and symbol represents age and sex, respectively. More visualizations are available in Appendix. F.4

Self-Supervised Time Series Representation Learning

Self-supervised time series representation learning uses unlabeled sequences to learn features that transfer to forecasting, classification, and anomaly detection.

See all 143 papers

Omni-Sleep: A Sleep Foundation Model via Hierarchical Contrastive Learning of CNS-ANS Dynamics
Jul 4, 2026Zhoujie Hou, Song Wang, Kexin Lou +3Sleep Stage Classification Latent Dynamics Modeling
Department of Biomedical Engineering, Southern University of Science and Technology, Shenzhen, China · Omni-Intelligence, Shenzhen, China · Shenzhen Loop Area Institute, Shenzhen, China
OSF: On Pre-training and Scaling of Sleep Foundation Models
Feb 27, 2026Zitao Shuai, Zongzhe Xu, David Yang +2Self-Supervised Pre-Training Scaling Laws
University of California, Los Angeles · Emory University
Comparative Analysis of State-of-the-Art Foundation Models for Sleep Analysis Under Channel Reduction
Aug 14, 2026Hassan Mehdi, Riku Klen, Ayse Kosal Bulbul +5Sleep Stage Classification EEG Foundation Models
Institute of Biomedicine, University of Turku, Finland · Turku PET Centre, University of Turku and Turku University Hospital, Finland · University College Cork, Ireland +3

Jul 4, 2026J/K move · Enter open · S save

Omni-Sleep: A Sleep Foundation Model via Hierarchical Contrastive Learning of CNS-ANS Dynamics

Zhoujie Hou, Song Wang, Kexin Lou, Mo Wang+2

Department of Biomedical Engineering, Southern University of Science and Technology, Shenzhen, China · Omni-Intelligence, Shenzhen, China · Shenzhen Loop Area Institute, Shenzhen, China

Sleep Stage Classification Latent Dynamics Modeling Hierarchical Representation Learning Multimodal Contrastive Learning Time Series Representation Learning

Open paper AskPDF arXiv

Sleep physiology arises from the coordinated dynamics of the central nervous system (CNS) and autonomic nervous system (ANS), as reflected by multimodal polysomnography signals including EEG, EOG, EMG, ECG, and respiration. However, existing sleep foundation models often fuse heterogeneous biosignals in a topology-agnostic manner, overlooking their physiological organization. We introduce Omni-Sleep, a sleep foundation model that uses the CNS/ANS partition as a physiological prior for topology-constrained representation learning. Omni-Sleep learns structured representations through three objectives: intra-system consistency, which captures shared subsystem-level factors within neural and cardio-respiratory signals; inter-system synchronization, which aligns subsystem trajectories to model brain--body dynamics; and latent-space masked temporal modeling, which captures long-horizon sleep dynamics. Pre-trained on over 100,000 hours of multi-center multimodal PSG data, Omni-Sleep is evaluated on sleep staging and multi-disease classification. Across datasets and modality-ablation settings, Omni-Sleep outperforms strong foundation-model baselines, showing improved label efficiency, cross-dataset generalization, and robustness to missing modalities. These results highlight the value of physiological hierarchy for generalizable sleep representation learning. Code is available at https://github.com/AutoBrain-sleep/OmniSleep.

Abstract

Self-Supervised Time Series Representation Learning

Explore similar work

Omni-Sleep: A Sleep Foundation Model via Hierarchical Contrastive Learning of CNS-ANS Dynamics

OSF: On Pre-training and Scaling of Sleep Foundation Models

Comparative Analysis of State-of-the-Art Foundation Models for Sleep Analysis Under Channel Reduction