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

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Park, Keondo, Na, Younghoon, Choi, Yourim, Ryu, Hyunwoo, Shin, Hyun-Woo, Kim, Hyung-Sin
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915784216805376
author Park, Keondo
Na, Younghoon
Choi, Yourim
Ryu, Hyunwoo
Shin, Hyun-Woo
Kim, Hyung-Sin
author_facet Park, Keondo
Na, Younghoon
Choi, Yourim
Ryu, Hyunwoo
Shin, Hyun-Woo
Kim, Hyung-Sin
contents 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 existing foundation models across a diverse suite of downstream tasks, demonstrating superior generalizability and label-efficient adaptation for clinical sleep analysis.
format Preprint
id arxiv_https___arxiv_org_abs_2602_07628
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures
Park, Keondo
Na, Younghoon
Choi, Yourim
Ryu, Hyunwoo
Shin, Hyun-Woo
Kim, Hyung-Sin
Artificial Intelligence
Machine Learning
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 existing foundation models across a diverse suite of downstream tasks, demonstrating superior generalizability and label-efficient adaptation for clinical sleep analysis.
title SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2602.07628