Predicting Depression From Smartphone Behavioral Markers Using Machine Learning Methods, Hyperparameter Optimization, and Feature Importance Analysis: Exploratory Study

JMIR mhealth and uhealth, Journal Year: 2021, Volume and Issue: 9(7), P. e26540 - e26540

Published: May 14, 2021

Depression is a prevalent mental health challenge. Current depression assessment methods using self-reported and clinician-administered questionnaires have limitations. Instrumenting smartphones to passively continuously collect moment-by-moment data sets quantify human behaviors has the potential augment current for early diagnosis, scalable, longitudinal monitoring of depression.

Language: Английский

---

Digital Biomarkers for Depression Screening With Wearable Devices: Cross-sectional Study With Machine Learning Modeling

JMIR mhealth and uhealth, Journal Year: 2021, Volume and Issue: 9(10), P. e24872 - e24872

Published: July 15, 2021

Background Depression is a prevalent mental disorder that undiagnosed and untreated in half of all cases. Wearable activity trackers collect fine-grained sensor data characterizing the behavior physiology users (ie, digital biomarkers), which could be used for timely, unobtrusive, scalable depression screening. Objective The aim this study was to examine predictive ability biomarkers, based on from consumer-grade wearables, detect risk working population. Methods This cross-sectional 290 healthy adults. Participants wore Fitbit Charge 2 devices 14 consecutive days completed health survey, including screening depressive symptoms using 9-item Patient Health Questionnaire (PHQ-9), at baseline weeks later. We extracted range known novel biomarkers physical activity, sleep patterns, circadian rhythms wearables steps, heart rate, energy expenditure, data. Associations between severity were examined with Spearman correlation multiple regression analyses adjusted potential confounders, sociodemographic characteristics, alcohol consumption, smoking, self-rated health, subjective loneliness. Supervised machine learning statistically selected predict symptom status). varying cutoff scores an acceptable PHQ-9 score define group different subsamples classification, while set remained same. For performance evaluation, we k-fold cross-validation obtained accuracy measures holdout folds. Results A total 267 participants included analysis. mean age 33 (SD 8.6, 21-64) years. Out participants, there mild female bias displayed (n=170, 63.7%). majority Chinese (n=211, 79.0%), single (n=163, 61.0%), had university degree (n=238, 89.1%). found greater robustly associated variation nighttime rate AM 4 6 AM; it also lower regularity weekday steps estimated nonparametric interdaily stability autocorrelation as well fewer steps-based daily peaks. Despite several reliable associations, our evidence showed limited whole sample However, balanced contrasted comprised depressed no or minimal symptoms), model achieved 80%, sensitivity 82%, specificity 78% detecting subjects high depression. Conclusions Digital have been discovered are behavioral physiological consumer increased assist screening, yet current shows ability. Machine models combining these discriminate individuals risk.

Language: Английский

---

Irregular sleep and event schedules are associated with poorer self-reported well-being in US college students

SLEEP, Journal Year: 2019, Volume and Issue: 43(6)

Published: Dec. 14, 2019

Abstract Study Objectives Sleep regularity, in addition to duration and timing, is predictive of daily variations well-being. One possible contributor changes these sleep dimensions are early morning scheduled events. We applied a composite metric—the Composite Phase Deviation (CPD)—to assess mistiming irregularity both event schedules examine their relationship with self-reported well-being US college students. Methods Daily well-being, actigraphy, timing first events (academic/exercise/other) were collected for approximately 30 days from 223 students (37% females) between 2013 2016. Participants rated upon awakening on five scales: Sleepy–Alert, Sad–Happy, Sluggish–Energetic, Sick–Healthy, Stressed–Calm. A longitudinal growth model time-varying covariates was used relationships variables (i.e. CPDSleep, duration, midsleep time) average Cluster analysis CPD vs. schedules. Results significant predictor (e.g. Stressed–Calm: b = −6.3, p < 0.01), whereas (Stressed–Calm, 1.0, 0.001). Although cluster revealed no systematic more mistimed/irregular not associated sleep), they interacted well-being: the poorest reported by whom mistimed irregular. Conclusions regularity may be risk factors lower Stabilizing and/or help improve Clinical Trial Registration NCT02846077.

Language: Английский

---

Improving Students' Daily Life Stress Forecasting using LSTM Neural Networks

Published: May 1, 2019

Accurately forecasting stress may enable people to make behavioral changes that could improve their future health. For example, accurate might inspire schedule get more sleep or exercise, in order reduce excessive tomorrow night. In this paper, we examine how accurately the previous N-days of multi-modal data can forecast evening's high/low binary levels using long short-term memory neural network models (LSTM), logistic regression (LR), and support vector machines (SVM). Using a total 2,276 days, with 1,231 overlapping 8-day sequences from 142 participants (including physiological signals, mobile phone usage, location, surveys), find LSTM significantly outperforms LR SVM best results reaching 83.6% 7 days prior data. time-series improves even when considering only subsets set, e.g., physiology particular, model reaches 81.4% accuracy objective passive data, i.e., not including subjective reports daily survey.

Language: Английский

---

Extraction and Interpretation of Deep Autoencoder-based Temporal Features from Wearables for Forecasting Personalized Mood, Health, and Stress

Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies, Journal Year: 2020, Volume and Issue: 4(2), P. 1 - 26

Published: June 15, 2020

Continuous wearable sensor data in high resolution contain physiological and behavioral information that can be utilized to predict human health wellbeing, establishing the foundation for developing early warning systems eventually improve wellbeing. We propose a deep neural network framework, Locally Connected Long Short-Term Memory Denoising AutoEncoder (LC-LSTM-DAE), automatically extract features from passively collected raw perform personalized prediction of self-reported mood, health, stress scores with precision. enabled learning by finetuning general representation model participant-specific data. The framework was evaluated using wellbeing labels college students (total 6391 days N=239). Sensor include skin temperature, conductance, acceleration; scored 0 - 100. Compared performance based on hand-crafted features, proposed achieved higher precision smaller number features. also provide statistical interpretation visual explanation learned models. Our results show possibility predicting accurately an interpretable ultimately real-time monitoring intervention benefit various populations.

Language: Английский

---

Exploring the digital footprint of depression: a PRISMA systematic literature review of the empirical evidence

BMC Psychiatry, Journal Year: 2022, Volume and Issue: 22(1)

Published: June 22, 2022

This PRISMA systematic literature review examined the use of digital data collection methods (including ecological momentary assessment [EMA], experience sampling method [ESM], biomarkers, passive sensing, mobile ambulatory assessment, and time-series analysis), emphasizing on phenotyping (DP) to study depression. DP is defined as profile health information objectively.Four distinct yet interrelated goals underpin this study: (a) identify empirical research examining depression; (b) describe different technology employed; (c) integrate evidence regarding efficacy in examination, diagnosis, monitoring depression (d) clarify definitions mental records terminology.Overall, 118 studies were assessed eligible. Considering terms employed, "EMA", "ESM", "DP" most predominant. A variety sources reported, including voice, language, keyboard typing kinematics, phone calls texts, geocoded activity, actigraphy sensor-related recordings (i.e., steps, sleep, circadian rhythm), self-reported apps' information. Reviewed employed subjectively objectively recorded combination with interviews psychometric scales.Findings suggest links between a person's Future recommendations include deriving consensus definition expanding consider broader contextual developmental circumstances relation their data/records.

Language: Английский

---

A scoping review on monitoring mental health using smart wearable devices

Mathematical Biosciences & Engineering, Journal Year: 2022, Volume and Issue: 19(8), P. 7899 - 7919

Published: Jan. 1, 2022

<abstract> <p>With the continuous development of times, social competition is becoming increasingly fierce, people are facing enormous pressure and mental health problems have become common. Long-term persistent can lead to severe disorders even death in individuals. The real-time accurate prediction individual has an effective method prevent occurrence disorders. In recent years, smart wearable devices been widely used for monitoring played important role. This paper provides a comprehensive review application fields, mechanisms, common signals, techniques results detection problems, aiming achieve more efficient health, early identification, prevention intervention provide reference improving level health.</p> </abstract>

Language: Английский

---

Machine learning for passive mental health symptom prediction: Generalization across different longitudinal mobile sensing studies

PLoS ONE, Journal Year: 2022, Volume and Issue: 17(4), P. e0266516 - e0266516

Published: April 27, 2022

Mobile sensing data processed using machine learning models can passively and remotely assess mental health symptoms from the context of patients’ lives. Prior work has trained single longitudinal studies, collected demographically homogeneous populations, over short time periods, a collection platform or mobile application. The generalizability model performance across studies not been assessed. This study presents first analysis to understand if combined predict generalize current publicly available data. We CrossCheck (individuals living with schizophrenia) StudentLife (university students) studies. In addition assessing generalizability, we explored personalizing align data, oversampling less-represented severe symptoms, improved performance. Leave-one-subject-out cross-validation (LOSO-CV) results were reported. Two (sleep quality stress) had similar question-response structures used as outcomes explore cross-dataset prediction. Models more likely be predictive (significant improvement predicting training mean) than single-study Expected distance between validation feature distributions decreased versus Personalization aligned each LOSO-CV participant but only stress. Oversampling significantly symptom classification sensitivity positive value, specificity. Taken together, these show that on may heterogeneous datasets. encourage researchers disseminate de-identified further standardize types enable better assessment generalizability.

Language: Английский

---

Digital Phenotyping of Mental Health using multimodal sensing of multiple situations of interest: A Systematic Literature Review

Journal of Biomedical Informatics, Journal Year: 2022, Volume and Issue: 138, P. 104278 - 104278

Published: Dec. 29, 2022

Language: Английский

---

Measuring emotions in education using wearable devices: A systematic review

Computers & Education, Journal Year: 2023, Volume and Issue: 200, P. 104797 - 104797

Published: April 8, 2023

Language: Английский

---