Dataset Information

Predicting Emotional States Using Behavioral Markers Derived From Passively Sensed Data: Data-Driven Machine Learning Approach.

ABSTRACT:

Background

Mental health disorders affect multiple aspects of patients' lives, including mood, cognition, and behavior. eHealth and mobile health (mHealth) technologies enable rich sets of information to be collected noninvasively, representing a promising opportunity to construct behavioral markers of mental health. Combining such data with self-reported information about psychological symptoms may provide a more comprehensive and contextualized view of a patient's mental state than questionnaire data alone. However, mobile sensed data are usually noisy and incomplete, with significant amounts of missing observations. Therefore, recognizing the clinical potential of mHealth tools depends critically on developing methods to cope with such data issues.

Objective

This study aims to present a machine learning-based approach for emotional state prediction that uses passively collected data from mobile phones and wearable devices and self-reported emotions. The proposed methods must cope with high-dimensional and heterogeneous time-series data with a large percentage of missing observations.

Methods

Passively sensed behavior and self-reported emotional state data from a cohort of 943 individuals (outpatients recruited from community clinics) were available for analysis. All patients had at least 30 days' worth of naturally occurring behavior observations, including information about physical activity, geolocation, sleep, and smartphone app use. These regularly sampled but frequently missing and heterogeneous time series were analyzed with the following probabilistic latent variable models for data averaging and feature extraction: mixture model (MM) and hidden Markov model (HMM). The extracted features were then combined with a classifier to predict emotional state. A variety of classical machine learning methods and recurrent neural networks were compared. Finally, a personalized Bayesian model was proposed to improve performance by considering the individual differences in the data and applying a different classifier bias term for each patient.

Results

Probabilistic generative models proved to be good preprocessing and feature extractor tools for data with large percentages of missing observations. Models that took into account the posterior probabilities of the MM and HMM latent states outperformed those that did not by more than 20%, suggesting that the underlying behavioral patterns identified were meaningful for individuals' overall emotional state. The best performing generalized models achieved a 0.81 area under the curve of the receiver operating characteristic and 0.71 area under the precision-recall curve when predicting self-reported emotional valence from behavior in held-out test data. Moreover, the proposed personalized models demonstrated that accounting for individual differences through a simple hierarchical model can substantially improve emotional state prediction performance without relying on previous days' data.

Conclusions

These findings demonstrate the feasibility of designing machine learning models for predicting emotional states from mobile sensing data capable of dealing with heterogeneous data with large numbers of missing observations. Such models may represent valuable tools for clinicians to monitor patients' mood states.

SUBMITTER: Sukei E

PROVIDER: S-EPMC8088855 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:INTRODUCTION:Adverse event (AE) under-reporting has been a recurrent issue raised during health authorities Good Clinical Practices (GCP) inspections and audits. Moreover, safety under-reporting poses a risk to patient safety and data integrity. The current clinical Quality Assurance (QA) practices used to detect AE under-reporting rely heavily on investigator site and study audits. Yet several sponsors and institutions have had repeated findings related to safety reporting, and this has led to delays in regulatory submissions. Recent developments in data management and IT systems allow data scientists to apply techniques such as machine learning to detect AE under-reporting in an automated fashion. OBJECTIVE:In this project, we developed a predictive model that enables Roche/Genentech Quality Program Leads oversight of AE reporting at the program, study, site, and patient level. This project was part of a broader effort at Roche/Genentech Product Development Quality to apply advanced analytics to augment and complement traditional clinical QA approaches. METHOD:We used a curated data set from 104 completed Roche/Genentech sponsored clinical studies to train a machine learning model to predict the expected number of AEs. Our final model used 54 features built on patient (e.g., demographics, vitals) and study attributes (e.g., molecule class, disease area). RESULTS:In order to evaluate model performance, we tested how well it would detect simulated test cases based on data not used for model training. For relevant simulation scenarios of 25%, 50%, and 75% under-reporting on the site level, our model scored an area under the curve (AUC) of the receiver operating characteristic (ROC) curve of 0.62, 0.79, and 0.92, respectively. CONCLUSION:The model has been deployed to evaluate safety reporting performance in a set of ongoing studies in the form of a QA/dashboard cockpit available to Roche Quality Program Leads. Applicability and production performance will be assessed over the next 12-24 months in which we will develop a validation strategy to fully integrate our model into Roche QA processes.

Project description:BackgroundEmotional state in everyday life is an essential indicator of health and well-being. However, daily assessment of emotional states largely depends on active self-reports, which are often inconvenient and prone to incomplete information. Automated detection of emotional states and transitions on a daily basis could be an effective solution to this problem. However, the relationship between emotional transitions and everyday context remains to be unexplored.ObjectiveThis study aims to explore the relationship between contextual information and emotional transitions and states to evaluate the feasibility of detecting emotional transitions and states from daily contextual information using machine learning (ML) techniques.MethodsThis study was conducted on the data of 18 individuals from a publicly available data set called ExtraSensory. Contextual and sensor data were collected using smartphone and smartwatch sensors in a free-living condition, where the number of days for each person varied from 3 to 9. Sensors included an accelerometer, a gyroscope, a compass, location services, a microphone, a phone state indicator, light, temperature, and a barometer. The users self-reported approximately 49 discrete emotions at different intervals via a smartphone app throughout the data collection period. We mapped the 49 reported discrete emotions to the 3 dimensions of the pleasure, arousal, and dominance model and considered 6 emotional states: discordant, pleased, dissuaded, aroused, submissive, and dominant. We built general and personalized models for detecting emotional transitions and states every 5 min. The transition detection problem is a binary classification problem that detects whether a person's emotional state has changed over time, whereas state detection is a multiclass classification problem. In both cases, a wide range of supervised ML algorithms were leveraged, in addition to data preprocessing, feature selection, and data imbalance handling techniques. Finally, an assessment was conducted to shed light on the association between everyday context and emotional states.ResultsThis study obtained promising results for emotional state and transition detection. The best area under the receiver operating characteristic (AUROC) curve for emotional state detection reached 60.55% in the general models and an average of 96.33% across personalized models. Despite the highly imbalanced data, the best AUROC curve for emotional transition detection reached 90.5% in the general models and an average of 88.73% across personalized models. In general, feature analyses show that spatiotemporal context, phone state, and motion-related information are the most informative factors for emotional state and transition detection. Our assessment showed that lifestyle has an impact on the predictability of emotion.ConclusionsOur results demonstrate a strong association of daily context with emotional states and transitions as well as the feasibility of detecting emotional states and transitions using data from smartphone and smartwatch sensors.

Project description:BackgroundDepression is a prevalent mental health challenge. Current depression assessment methods using self-reported and clinician-administered questionnaires have limitations. Instrumenting smartphones to passively and continuously collect moment-by-moment data sets to quantify human behaviors has the potential to augment current depression assessment methods for early diagnosis, scalable, and longitudinal monitoring of depression.ObjectiveThe objective of this study was to investigate the feasibility of predicting depression with human behaviors quantified from smartphone data sets, and to identify behaviors that can influence depression.MethodsSmartphone data sets and self-reported 8-item Patient Health Questionnaire (PHQ-8) depression assessments were collected from 629 participants in an exploratory longitudinal study over an average of 22.1 days (SD 17.90; range 8-86). We quantified 22 regularity, entropy, and SD behavioral markers from the smartphone data. We explored the relationship between the behavioral features and depression using correlation and bivariate linear mixed models (LMMs). We leveraged 5 supervised machine learning (ML) algorithms with hyperparameter optimization, nested cross-validation, and imbalanced data handling to predict depression. Finally, with the permutation importance method, we identified influential behavioral markers in predicting depression.ResultsOf the 629 participants from at least 56 countries, 69 (10.97%) were females, 546 (86.8%) were males, and 14 (2.2%) were nonbinary. Participants' age distribution is as follows: 73/629 (11.6%) were aged between 18 and 24, 204/629 (32.4%) were aged between 25 and 34, 156/629 (24.8%) were aged between 35 and 44, 166/629 (26.4%) were aged between 45 and 64, and 30/629 (4.8%) were aged 65 years and over. Of the 1374 PHQ-8 assessments, 1143 (83.19%) responses were nondepressed scores (PHQ-8 score <10), while 231 (16.81%) were depressed scores (PHQ-8 score ≥10), as identified based on PHQ-8 cut-off. A significant positive Pearson correlation was found between screen status-normalized entropy and depression (r=0.14, P<.001). LMM demonstrates an intraclass correlation of 0.7584 and a significant positive association between screen status-normalized entropy and depression (β=.48, P=.03). The best ML algorithms achieved the following metrics: precision, 85.55%-92.51%; recall, 92.19%-95.56%; F1, 88.73%-94.00%; area under the curve receiver operating characteristic, 94.69%-99.06%; Cohen κ, 86.61%-92.90%; and accuracy, 96.44%-98.14%. Including age group and gender as predictors improved the ML performances. Screen and internet connectivity features were the most influential in predicting depression.ConclusionsOur findings demonstrate that behavioral markers indicative of depression can be unobtrusively identified from smartphone sensors' data. Traditional assessment of depression can be augmented with behavioral markers from smartphones for depression diagnosis and monitoring.

Project description:The ability to carry out instrumental activities of daily living, such as paying bills, remembering appointments and shopping alone decreases with age, yet there are remarkable individual differences in the rate of decline among older adults. Understanding variables associated with a decline in instrumental activities of daily living is critical to providing appropriate intervention to prolong independence. Prior research suggests that cognitive measures, neuroimaging and fluid-based biomarkers predict functional decline. However, a priori selection of variables can lead to the over-valuation of certain variables and exclusion of others that may be predictive. In this study, we used machine learning techniques to select a wide range of baseline variables that best predicted functional decline in two years in individuals from the Alzheimer's Disease Neuroimaging Initiative dataset. The sample included 398 individuals characterized as cognitively normal or mild cognitive impairment. Support vector machine classification algorithms were used to identify the most predictive modality from five different data modality types (demographics, structural MRI, fluorodeoxyglucose-PET, neurocognitive and genetic/fluid-based biomarkers). In addition, variable selection identified individual variables across all modalities that best predicted functional decline in a testing sample. Of the five modalities examined, neurocognitive measures demonstrated the best accuracy in predicting functional decline (accuracy = 74.2%; area under the curve = 0.77), followed by fluorodeoxyglucose-PET (accuracy = 70.8%; area under the curve = 0.66). The individual variables with the greatest discriminatory ability for predicting functional decline included partner report of language in the Everyday Cognition questionnaire, the ADAS13, and activity of the left angular gyrus using fluorodeoxyglucose-PET. These three variables collectively explained 32% of the total variance in functional decline. Taken together, the machine learning model identified novel biomarkers that may be involved in the processing, retrieval, and conceptual integration of semantic information and which predict functional decline two years after assessment. These findings may be used to explore the clinical utility of the Everyday Cognition as a non-invasive, cost and time effective tool to predict future functional decline.