{"paper_id":"2998efb5-dffc-488f-a9ac-99ae9ba2fe89","body_text":"Sleep Estimation from Low Frequency Smartphone Sensors via Bayesian Hidden Markov Model | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Sleep Estimation from Low Frequency Smartphone Sensors via Bayesian Hidden Markov Model Andrew Jin Soo Byun, Yufan Li, Sam Cong, Alex Dhima, Shiwei Wang, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7217304/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Sleep disturbances are recognized as transdiagnostic markers and potential mechanistic contributors to psychiatric illness, yet objective sleep monitoring remains rare in large-scale psychiatric research due to infrastructure and methodological barriers. While smartphones enable scalable, real-world behavioral sensing, most current approaches are limited by single-sensor thresholds, proprietary algorithms, or lack of validation in diverse populations. Here, we introduce a probabilistic Bayesian hidden Markov model that integrates accelerometer and screen state data to infer nightly sleep states and extract a set of behavioral sleep metrics. Model performance was evaluated using both empirically derived simulation and real-world self-reported sleep logs. Analyzing 5,888 nights from 516 participants, we identified substantial heterogeneity in individual sleep-symptom coupling, with unsupervised clustering of sensor-derived sleep and symptom dynamics revealing five distinct phenotypes that were consistent with independent clinical assessments. Our approach provides a robust framework for large-scale, non-invasive sleep monitoring, with direct applications in digital psychiatry and individualized intervention. Health sciences/Biomarkers/Diagnostic markers Health sciences/Medical research Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Full Text Additional Declarations There is NO Competing Interest. Supplementary Files sleepbyunsupplementaryab.docx Supplementary Materials sleepbyuntableab.docx Table 1 suppfig1.tif Supplementary Figure 1 Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-7217304\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":false,\"archivedVersions\":[],\"articleType\":\"Article\",\"associatedPublications\":[],\"authors\":[{\"id\":491055528,\"identity\":\"b48d46b0-7cbd-4ac8-912d-4c3a799ce8d5\",\"order_by\":0,\"name\":\"Andrew Jin Soo Byun\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/UlEQVRIiWNgGAWjYBACxmYwdQCImQ8wMDZARCWAPGK0sCUQpwUKQFp4DIjTwtzOe/jDh4o7iRtu93x+wbjDLt/g+NmHNxgqrBMbcDqML01yxplniRvunN1mwXgm2XLDmXRjC4Yz6Xi08Jgx87YdTtxwI3ebAWMbs4FkQxqbBCNQBI8W488QLTnPgFrqDST7nwG1/MOrxUAaqoX5AdBwA34JkC0NeLWYgfxiPPNGmhlDYttxoJZnzBYJx9KNcWkx7D9jDAox2b4byY8/fGyrNmDjT2O88aHGWhanFqiEI5Bmk0iACSdgUwsF8lDaHoiZP+BROApGwSgYBSMYAABE9F6KzqCX8AAAAABJRU5ErkJggg==\",\"orcid\":\"https://orcid.org/0000-0003-2062-7647\",\"institution\":\"Beth Israel Deaconess Medical Center, Harvard Medical School\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Andrew\",\"middleName\":\"Jin Soo\",\"lastName\":\"Byun\",\"suffix\":\"\"},{\"id\":491055529,\"identity\":\"59eb2ed6-4d95-49c4-bd05-5e1a029b8d75\",\"order_by\":1,\"name\":\"Yufan Li\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Harvard University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yufan\",\"middleName\":\"\",\"lastName\":\"Li\",\"suffix\":\"\"},{\"id\":491055530,\"identity\":\"54b485ba-500e-4cdd-86e4-75ce526d5d34\",\"order_by\":2,\"name\":\"Sam Cong\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Beth Israel Deaconess Medical Center, Harvard Medical School\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Sam\",\"middleName\":\"\",\"lastName\":\"Cong\",\"suffix\":\"\"},{\"id\":491055531,\"identity\":\"a060a9ea-c6d8-4ad5-8ce9-104f8a262c19\",\"order_by\":3,\"name\":\"Alex Dhima\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Beth Israel Deaconess Medical Center, Harvard Medical School\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Alex\",\"middleName\":\"\",\"lastName\":\"Dhima\",\"suffix\":\"\"},{\"id\":491055532,\"identity\":\"30836bdd-130a-4016-95e1-24f493927692\",\"order_by\":4,\"name\":\"Shiwei Wang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Beth Israel Deaconess Medical Center, Harvard Medical School\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Shiwei\",\"middleName\":\"\",\"lastName\":\"Wang\",\"suffix\":\"\"},{\"id\":491055533,\"identity\":\"05ec547d-d3a7-45cd-b1ad-bd3202e59d32\",\"order_by\":5,\"name\":\"Matthew Flathers\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Harvard Medical School, BIDMC\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Matthew\",\"middleName\":\"\",\"lastName\":\"Flathers\",\"suffix\":\"\"},{\"id\":491055534,\"identity\":\"76aff1b4-ded7-4636-b08d-20950539ca20\",\"order_by\":6,\"name\":\"John Torous\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0002-5362-7937\",\"institution\":\"Beth Israel Deaconess Medical Center, Harvard Medical School\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"John\",\"middleName\":\"\",\"lastName\":\"Torous\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-07-25 21:30:16\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-7217304/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-7217304/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":88470717,\"identity\":\"f9a74ec5-68db-4756-a01d-78ab60139f57\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 18:59:51\",\"extension\":\"png\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1371547,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eOverview of the smartphone-based sleep estimation pipeline.\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e(a) Schematic of the sleep estimation analysis workflow. Raw screen state and accelerometer data are input into a hidden Markov model, which estimates the probability of being in latent sleep or active states at each timepoint. State probabilities are segmented into discrete sleep episodes using Otsu thresholding, which optimizes binarization. (b) Example participant’s raw sensor streams. Binary screen state (blue) and continuous accelerometer activity (orange) are plotted over time, illustrating alternating periods of rest and activity. (c) Sleep segment processing stages. Panel 1: Raw model-derived sleep probabilities. Panel 2: Probabilities after flat-top filtering, which smooths out noise from brief inactivity unlikely to reflect true sleep. Panel 3: Final sleep segments, aligned to latent sleep states and annotated with pre-sleep, sleep, and post-sleep intervals. We extract multiple sleep metrics from these regions, including activity level and variance (pre/post-sleep), sleep probability, fragmentation, and overall sleep score. (d) Comparison of sensor-derived and self-reported sleep. On nights where both sensor-derived and self-reported sleep are available, we directly compare their agreement.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"mainfig1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/34a1aa5be85e26d1dc687af9.png\"},{\"id\":88470286,\"identity\":\"7d7faafb-6bd1-42c8-a09f-1e3e8cae93cd\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 18:51:51\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":976002,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eSimulation Framework for Evaluating Sleep Model Performance\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e(a) Simulation performance testing pipeline. A multi-step simulation pipeline extracts real participant sensor parameters (means and variances of sleep and active states), defines two core simulation scenarios (linear drift with and without fragmentation) then generates ground truth sleep/wake labels. Synthetic accelerometer and screen state data are simulated across a range of sensitivity indices (d′ = 0.5 to 10). The Bayesian hidden Markov model (HMM) is run on these simulated data, and key performance metrics (accuracy, precision, recall, F1) are computed. (b) Simulation scenarios (ground truth labels). Visualization of two simulation scenarios. Left: Linear drift in sleep timing without fragmentation. Right: Linear drift with additional fragmentation, introducing periods of wakefulness within the sleep episode. (c) Sensitivity index (d′) manipulation. Distributions of simulated accelerometer values for sleep (blue) and active (orange) states are shown at four example sensitivity index values (d′ = 0.5, 2, 5, 10), with vertical lines marking their means (μ). Increasing d′ corresponds to greater separability between sleep and active states. (d) Performance metrics across sensitivity index. Bar plots of mean ± SEM (across 10 iterations for each d′) for accuracy, precision, recall, and F1 score, shown for both simulation scenarios across increasing d′ values. Model performance improves as d′ increases and is consistently higher in the no-fragmentation scenario. (e) Distribution of participant-derived sensitivity indices. Histogram of estimated d′ across 516 participants in the empirical dataset, indicating a median d′ of 2.93.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"mainfig2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/27fec1ca8406f382cf99f9f0.png\"},{\"id\":88470288,\"identity\":\"0a0c50b2-f6f1-4df6-811e-35510edf56c4\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 18:51:51\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1332759,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eComparison of model-derived sleep estimates with self-reported ecological momentary assessment (EMA) sleep data.\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e(a) Bedtime comparison. Left: Distribution of model-estimated versus EMA-reported bedtimes shows close alignment with a modest offset (Model Mean: 23:05; EMA Mean: 23:48). Middle: Histogram of residuals for bedtime, with a mean residual of 42 minutes. Right: Scatterplot of model versus EMA bedtime (r = 0.68, p\\u0026lt;1e-5), showing a positive correlation. (b) Waketime comparison. Left: Distribution of model-estimated versus EMA-reported waketimes (Model Mean: 07:12; EMA Mean: 07:35). Middle: Waketime residual histogram with a mean residual of 23 minutes. Right: Scatterplot showing a positive correlation between the model and EMA wake time (r = 0.65, p \\u0026lt; 1e-5). (c) Sleep duration comparison. Left: Distribution of model-estimated and EMA-derived sleep durations shows a tighter peak in model estimates (Model Mean: 7 hr 50 min; EMA Mean: 7 hr 47 min). Middle: Histogram of duration residuals (mean residual: 3 minutes). Right: Scatterplot of model versus EMA sleep duration (r = 0.48, p \\u0026lt; 1e-5).\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"mainfig3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/97891adb7d59919694732b3a.png\"},{\"id\":88470289,\"identity\":\"e5490792-56d3-468d-bc11-9b36842845c0\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 18:51:51\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1872151,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eParticipant-Specific Noise Filtering Enhances Model–EMA Sleep Agreement, Independent of Psychiatric Symptom Severity\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e(a) Distribution of root mean square error (RMSE) values per participant for model-estimated versus EMA-reported sleep metrics. Left: Bedtime RMSE shows a positively skewed distribution centered near 1.2 hours. Middle: Waketime RMSE. Right: Duration RMSE distribution. (b) r values between model and EMA estimates as a function of progressively stricter RMSE filtering thresholds. Filtering out participants with high RMSE improves the explained variance across all sleep metrics, with r for bedtime and waketime reaching over 0.89 under the \\u0026lt;1.2hr threshold. This highlights the benefit of participant-level quality control when validating model-derived sleep estimates against self-report. (c) Scatterplots depict the relationship between per-participant RMSE (model–EMA disagreement) and average EMA-derived symptom scores across anxiety, functioning, and mood domains. Left column: Weak associations are observed between RMSE (bedtime, waketime, and duration) and average anxiety scores (r = 0.09, r = 0.07, r = 0.15 respectively). Middle column: Similarly weak correlations are found between RMSE and mood, particularly for waketime RMSE (r = 0.11, r = 0.03, r = 0.16 respectively). Right column: Minimal associations are observed between RMSE and functioning (r = 0.17, r = 0.09, r = 0.19 respectively).\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"mainfig4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/407328575e2ddb6f6e7fc586.png\"},{\"id\":88471019,\"identity\":\"26c3ddd0-8637-4ea0-a231-077a5a9780ee\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 19:07:52\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1615284,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eParticipant Heterogeneity in Sleep-Symptom Relationships Is Captured by Data-Driven Subtyping\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e(a) Participant-level associations between sleep score and clinical measures. Histograms show the distribution of within-participant regression slopes linking daily sleep score to anxiety (left), mood (middle), and functioning (right) scores. Significant associations (p \\u0026lt; 0.05) are highlighted, indicating substantial individual heterogeneity in how changes in sleep score relate to clinical symptom fluctuations. (b) Cluster number selection for participant subtyping. Three clustering validity metrics (Elbow method, Silhouette score, Davies-Bouldin index) were used to determine the optimal number of participant subtypes. All metrics indicate a 5-cluster solution (highlighted) provides the best balance of within- and between-group structure. (c) Identification and characterization of sleep-symptom subtypes. Left: UMAP projection of all participants, colored by cluster assignment. Right: Each subtype's z-scored sleep and clinical feature profiles (mean ± SEM), highlighting distinct patterns across clusters: (1) Healthy early chronotype with low pre/post-sleep activity (N = 107), (2) Severe symptom insomnia (N = 91), (3) Late chronotype, low symptom (N = 106), (4) Healthy early chronotype, high pre/post-sleep activity (N = 124), and (5) Moderate symptoms with fragmented sleep (N = 81).\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"mainfig5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/89013fe38df46a9ecb1943fd.png\"},{\"id\":88471698,\"identity\":\"7f1613a4-294e-4343-a00f-f9710dc29096\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 19:16:05\",\"extension\":\"pdf\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":3298338,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"sleepbyunsubmissionabjtFinal.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1_covered_0d2c1b1a-d3d9-4713-9baf-4058d5300213.pdf\"},{\"id\":88470719,\"identity\":\"6a693d49-bc5a-45b0-a180-1779345c7029\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 18:59:51\",\"extension\":\"docx\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":5163504,\"visible\":true,\"origin\":\"\",\"legend\":\"Supplementary Materials\",\"description\":\"\",\"filename\":\"sleepbyunsupplementaryab.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/f27600401b34412cc094e068.docx\"},{\"id\":88470292,\"identity\":\"b7c9bf04-c1ae-42ab-99ac-5893e4bdff1c\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 18:51:52\",\"extension\":\"docx\",\"order_by\":2,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":4935470,\"visible\":true,\"origin\":\"\",\"legend\":\"Table 1\",\"description\":\"\",\"filename\":\"sleepbyuntableab.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/ad091492eefd2f16347037d9.docx\"},{\"id\":88470293,\"identity\":\"8664b081-c9ed-4485-a023-cb664fd3c81b\",\"added_by\":\"auto\",\"created_at\":\"2025-08-06 18:51:52\",\"extension\":\"tif\",\"order_by\":3,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":16158084,\"visible\":true,\"origin\":\"\",\"legend\":\"Supplementary Figure 1\",\"description\":\"\",\"filename\":\"suppfig1.tif\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7217304/v1/eedf2c64703c5936f592c8f2.tif\"}],\"financialInterests\":\"There is \\u003cb\\u003eNO\\u003c/b\\u003e Competing Interest.\",\"formattedTitle\":\"Sleep Estimation from Low Frequency Smartphone Sensors via Bayesian Hidden Markov Model\",\"fulltext\":[],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":false,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":true,\"hideJournal\":false,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":false,\"isAuthorSuppliedPdf\":true,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":true,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"nature-portfolio\",\"isNatureJournal\":true,\"hasQc\":false,\"allowDirectSubmit\":false,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"\",\"title\":\"Nature Portfolio\",\"twitterHandle\":\"\",\"acdcEnabled\":false,\"dfaEnabled\":false,\"editorialSystem\":\"ejp\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":false},\"keywords\":\"\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-7217304/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-7217304/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"Sleep disturbances are recognized as transdiagnostic markers and potential mechanistic contributors to psychiatric illness, yet objective sleep monitoring remains rare in large-scale psychiatric research due to infrastructure and methodological barriers. While smartphones enable scalable, real-world behavioral sensing, most current approaches are limited by single-sensor thresholds, proprietary algorithms, or lack of validation in diverse populations. Here, we introduce a probabilistic Bayesian hidden Markov model that integrates accelerometer and screen state data to infer nightly sleep states and extract a set of behavioral sleep metrics. Model performance was evaluated using both empirically derived simulation and real-world self-reported sleep logs. Analyzing 5,888 nights from 516 participants, we identified substantial heterogeneity in individual sleep-symptom coupling, with unsupervised clustering of sensor-derived sleep and symptom dynamics revealing five distinct phenotypes that were consistent with independent clinical assessments. Our approach provides a robust framework for large-scale, non-invasive sleep monitoring, with direct applications in digital psychiatry and individualized intervention.\",\"manuscriptTitle\":\"Sleep Estimation from Low Frequency Smartphone Sensors via Bayesian Hidden Markov Model\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-08-06 18:51:47\",\"doi\":\"10.21203/rs.3.rs-7217304/v1\",\"editorialEvents\":[],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"nature-mental-health\",\"isNatureJournal\":true,\"hasQc\":false,\"allowDirectSubmit\":false,\"externalIdentity\":\"natmentalhealth\",\"sideBox\":\"Learn more about [Nature Mental Health](https://www.nature.com/natmentalhealth/)\",\"snPcode\":\"44220\",\"submissionUrl\":\"https://mts-natmentalhealth.nature.com/cgi-bin/main.plex\",\"title\":\"Nature Mental Health\",\"twitterHandle\":\"\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"ejp\",\"reportingPortfolio\":\"Nature Research\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":false}}],\"origin\":\"\",\"ownerIdentity\":\"016fa5e1-9655-4f4d-972d-bf218b902bdb\",\"owner\":[],\"postedDate\":\"August 6th, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"under-review\",\"subjectAreas\":[{\"id\":52144940,\"name\":\"Health sciences/Biomarkers/Diagnostic markers\"},{\"id\":52144941,\"name\":\"Health sciences/Medical research\"}],\"tags\":[],\"updatedAt\":\"2026-03-18T10:04:17+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2025-08-06 18:51:47\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-7217304\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-7217304\",\"identity\":\"rs-7217304\",\"version\":[\"v1\"]},\"buildId\":\"8U1c8b4HqxoKbykW_rLl7\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}