Skip to content
DraconiticResearch
Search Sources API
203,952 studies31 sourcesUpdated 20 min ago
Peer-reviewedCC BY 4.0NPJ Digital Medicine

BCGNet: an AI model trained on 600 K hours of sleep data for a novel under-pillow contactless monitoring device

  • Shigeng Chen1,2
  • Xuesong Chen1
  • Weijun Huang3
  • Fei Lei4
  • Chuxuan Shan1
  • Zengrui Jin5
  • Yunhan Shi6
  • Yichen Wang1
  • Rui Zhao7
  • Xing Xu8
  • Dongsheng Lv7
  • Yanru Li6
10 more authors
  • Shirui Pan2
  • Ambrose Chiang9,10
  • M. Brandon Westover11,12
  • Shenda Hong13
  • Chao Zhang5
  • Shankai Yin3
  • Chun-feng Liu8
  • Hongliang Yi3
  • Xiangdong Tang4
  • Yue Leng14
14 affiliations
  1. 1Five Seasons Medical, Beijing, China
  2. 2School of Information and Communication Technology, Griffith University, Brisbane, QLD Australia
  3. 3Department of Otorhinolaryngology Head and Neck Surgery, Shanghai Sixth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China
  4. 4Sleep Medicine Center, West China Hospital, Sichuan University, Chengdu, China
  5. 5Department of Electronic Engineering, Tsinghua University, Beijing, China
  6. 6Department of Otorhinolaryngology Head and Neck Surgery, Beijing Tongren Hospital, Capital Medical University, Beijing, China
  7. 7Sleep Medicine Center, Inner Mongolia Mental Health Center (The Third Hospital of Inner Mongolia Autonomous Region, Brain Hospital of Inner Mongolia Autonomous Region), Hohhot, China
  8. 8Department of Neurology and Clinical Research Center of Neurological Disease, Second Affiliated Hospital of Soochow University, Suzhou, China
  9. 9Sleep Medicine Section, Louis Stokes Cleveland VA Medical Center, Cleveland, OH USA
  10. 10Division of Pulmonary, Critical Care, and Sleep Medicine, University Hospitals Cleveland Medical Center, Cleveland, OH USA
  11. 11Department of Neurology, Harvard Medical School, Boston, MA USA
  12. 12Department of Neurology, Beth Israel Deaconess Medical Center, Boston, MA USA
  13. 13National Institute of Health Data Science, Peking University, Beijing, China
  14. 14Department of Psychiatry and Behavioral Sciences, University of California, San Francisco, CA USA
Read full textPDFDOICite

Abstract

Sleep profoundly impacts health, yet current gold-standard Polysomnography (PSG) is constrained by cost, discomfort, and limited scalability for longitudinal monitoring. Ballistocardiography (BCG) offers a non-invasive and user-friendly alternative but often lacks the precision needed for reliable real-world applications. To address this gap, we propose BCGNet, a two-stage transfer learning model that is first pre-trained on 580,865 h of PSG and then fine-tuned and validated on 15,081 h of BCG (total 595,946 h of recordings). Across multiple validation cohorts, BCGNet achieves strong performance in 4-class sleep staging (F1: 0.710−0.817), Apnea-Hypopnea Index (AHI3%) estimation (Pearson’s r > 0.95), and robust quantification of sleep continuity and architecture (ICC and Pearson’s r generally >0.8). Notably, BCGNet maintains strong performance even on short daytime naps and demonstrates excellent generalizability across diverse external datasets. Deployed as a portable, contactless sleep tracking mat, BCGNet represents a major step towards scalable, user-friendly solutions for longitudinal home sleep monitoring, with important implications for population screening and personalized sleep medicine.

Related

  1. Self supervised multi scale learning for WiFi CSI contactless human activity recognition with potential applications in health monitoring
    Hui Xu, Deng-ju Yao, Wei-dong Liu +3
    Scientific ReportsPeer-reviewedCC BY 4.0NatureCrossref
  2. Physics-Constrained Deep Learning Model for Contactless Blood Pressure Monitoring from Triaxial Bodyseismography
    Yuanyuan Zhang, Yida Zhang, Jiahui Li +6
    arXivPreprintarXiv licenceFull textBiological & medical physicsarXiv
  3. IOT-enabled wearable device and cloud framework for chronic disease identification and health monitoring using LST-BCLNM
    J. Poongodi, K. Kavitha, Lakshmana Kumar R +1
    Scientific ReportsPeer-reviewedCC BY 4.0NatureCrossref
  4. Flexible Magnetoelectric Device Based on VB2/PVDF/Metglass for Real-Time Monitoring of External Short Circuits in Lithium-Ion Batteries
    Yang Zhou, Longtai Jiang, Tao Liu +3
    PolymersPeer-reviewedCC BYFull textMaterials & polymersEurope PMC
  5. Perceived Acceptability, Design Preferences, and User Requirements for a Proposed Ear-Worn Device for Heart-Rate and Ear-Temperature Monitoring with Localized Neck Cooling: A Pre-Prototype Mixed-Methods Study in People with Multiple Sclerosis
    Healthcare (Basel, Switzerland)Peer-reviewedCC BYFull textEurope PMC

Cite

BibTeXRIS
BibTeX
@article{chen2026bcgnet,
  title = {{BCGNet: an AI model trained on 600 K hours of sleep data for a novel under-pillow contactless monitoring device}},
  author = {Shigeng Chen and Xuesong Chen and Weijun Huang and Fei Lei and Chuxuan Shan and Zengrui Jin and Yunhan Shi and Yichen Wang and Rui Zhao and Xing Xu and Dongsheng Lv and Yanru Li and Shirui Pan and Ambrose Chiang and M. Brandon Westover and Shenda Hong and Chao Zhang and Shankai Yin and Chun-feng Liu and Hongliang Yi and Xiangdong Tang and Yue Leng},
  journal = {NPJ Digital Medicine},
  year = {2026},
  month = jul,
  doi = {10.1038/s41746-026-02885-y},
  url = {https://doi.org/10.1038/s41746-026-02885-y},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - BCGNet: an AI model trained on 600 K hours of sleep data for a novel under-pillow contactless monitoring device
AU  - Shigeng Chen
AU  - Xuesong Chen
AU  - Weijun Huang
AU  - Fei Lei
AU  - Chuxuan Shan
AU  - Zengrui Jin
AU  - Yunhan Shi
AU  - Yichen Wang
AU  - Rui Zhao
AU  - Xing Xu
AU  - Dongsheng Lv
AU  - Yanru Li
AU  - Shirui Pan
AU  - Ambrose Chiang
AU  - M. Brandon Westover
AU  - Shenda Hong
AU  - Chao Zhang
AU  - Shankai Yin
AU  - Chun-feng Liu
AU  - Hongliang Yi
AU  - Xiangdong Tang
AU  - Yue Leng
PY  - 2026
DA  - 2026/07/03
JO  - NPJ Digital Medicine
DO  - 10.1038/s41746-026-02885-y
UR  - https://doi.org/10.1038/s41746-026-02885-y
AB  - Sleep profoundly impacts health, yet current gold-standard Polysomnography (PSG) is constrained by cost, discomfort, and limited scalability for longitudinal monitoring. Ballistocardiography (BCG) offers a non-invasive and user-friendly alternative but often lacks the precision needed for reliable real-world applications. To address this gap, we propose BCGNet, a two-stage transfer learning model that is first pre-trained on 580,865 h of PSG and then fine-tuned and validated on 15,081 h of BCG (total 595,946 h of recordings). Across multiple validation cohorts, BCGNet achieves strong performance in 4-class sleep staging (F1: 0.710−0.817), Apnea-Hypopnea Index (AHI3%) estimation (Pearson’s r > 0.95), and robust quantification of sleep continuity and architecture (ICC and Pearson’s r generally >0.8). Notably, BCGNet maintains strong performance even on short daytime naps and demonstrates excellent generalizability across diverse external datasets. Deployed as a portable, contactless sleep tracking mat, BCGNet represents a major step towards scalable, user-friendly solutions for longitudinal home sleep monitoring, with important implications for population screening and personalized sleep medicine.
SN  - 2398-6352
ER  -