Fine tuned personalized machine learning models to detect insomnia risk based on data from a smart bed platform

Aug 5, 2026

Personalized fine-tuned machine learning is applied to detect insomnia risk using smart-bed data, combining questionnaire responses with longitudinal objective sleep measurements. Participants from an IRB-approved study completed four questionnaires, including the Insomnia Severity Index (ISI), separated by six-week intervals. For 1,489 participants with at least three questionnaires, objective platform metrics covering sleep/wake and cardio-respiratory signals were analyzed with an incremental passive-aggressive learning approach. Generic models showed limited discrimination, while personalized tuning using only five sessions achieved AUC values above 0.8 across ISI thresholds.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Insomnia prevalence and clinical impact

Underrecognition and related sleep-technology progress

Methods

Study design and participant questionnaires

Objective smart-bed metrics

Incremental passive-aggressive learning and ISI thresholds

Results

Generic vs personalized model performance

Effect of number of tuning sessions

Discussion

Implications for personalized detection and sleep medicine

TYPE Original Research PUBLISHED 14 February 2024 DOI 10. 3389/fneur.2024.1303978 OPEN ACCESS EDITED BY Miguel Meira E. Cruz, Centro Cardiovascular da Faculdade de Medicina da Universidade de Lisboa, Portugal REVIEWED BY Jung Bin Kim, Korea University Anam Hospital, Republic of Korea Daniel Combs, University of Arizona, United States *CORRESPONDENCE Gary Garcia-Molina [gary.garciamolina@sleepnumber.com](gary.garciamolina@sleepnumber.com); [gmgarcia@wisc.edu](gmgarcia@wisc.edu) RECEIVED 28 September 2023 ACCEPTED 24 January 2024 PUBLISHED 14 February 2024 CITATION Winger T, Chellamuthu V, Guzenko D, Aloia M, Barr S, DeFranco S, Gorski B, Mushtaq F and Garcia-Molina G (2024) Fine tuned personalized machine learning models to detect insomnia risk based on data from a smart bed platform. Front. Neurol. 15:1303978 . doi: 10.3389/fneur.2024.1303978 COPYRIGHT © 2024 Winger, Chellamuthu, Guzenko, Aloia, Barr, DeFranco, Gorski, Mushtaq and Garcia-Molina. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Fine tuned personalized machine learning models to detect insomnia risk based on data from a smart bed platform Trevor Winger1,2 , Vidhya Chellamuthu1 , Dmytro Guzenko3 , Mark Aloia4,5 , Shawn Barr1 , Susan DeFranco4 , Brandon Gorski4 , Faisal Mushtaq1 and Gary Garcia-Molina1,6* 1 Sleep Number Labs, Sleep Number, San Jose, CA, United States, 2 Department of Computer Science and Engineering, University of Minnesota, Minneapolis, MN, United States, 3 GlobalLogic, Kyiv, Ukraine, 4 Sleep Number Corporation, Minneapolis, MN, United States, 5 National Jewish Health, Denver, CO, United States, 6 Department of Psychiatry, University of Wisconsin-Madison, Madison, WI, United States Introduction: Insomnia causes serious adverse health e􀀀ects and is estimated toa􀀀ect 10–30% of the worldwide population. This study leverages personalized ﬁne-tuned machine learning algorithms to detect insomnia risk based on questionnaire and longitudinal objective sleep data collected by a smart bed platform. Methods: Users of the Sleep Number smart bed were invited to participate inan IRB approved study which required them to respond to four questionnaires (which included the Insomnia Severity Index; ISI) administered 6 weeks apart from each other in the period from November 2021 to March 2022 . For 1,489 participants who completed at least 3 questionnaires, objective data (which includes sleep/wake and cardio-respiratory metrics) collected by the platform were queried for analysis. An incremental, passive-aggressive machine learning model was used to detect insomnia risk which was deﬁned by the ISI exceeding a given threshold. Three ISI thresholds (8, 10, and 15) were considered. The incremental model is advantageous because it allows personalized ﬁne-tuning by adding individual training data to a generic model. Results: The generic model, without personalizing, resulted in an area under the receiving-operating curve (AUC) of about 0 . 5 for each ISI threshold. The personalized ﬁne-tuning with the data of just ﬁve sleep sessions from the individual for whom the model is being personalized resulted in AUCs exceeding 0.8 for all ISI thresholds. Interestingly, no further AUC enhancements resulted by adding personalized data exceeding ten sessions. Discussion: These are encouraging results motivating further investigation into the application of personalized ﬁne tuning machine learning to detect insomnia risk based on longitudinal sleep data and the extension of this paradigm to sleep medicine. KEYWORDS insomnia

FAQs

How does personalization affect model performance?

Personalized fine-tuning using data from five sleep sessions for an individual increased discrimination substantially, with AUC values exceeding 0.8 across ISI thresholds.

Aug 5, 2026

Upload by Emma Wilson