Explainable stress type classification captures physiologically relevant responses in the Maastricht Acute Stress Test.

Introduction: Current stress detection methods concentrate on identification of stress and non-stress states despite the existence of various stress types. The present study performs a more specific, explainable stress classification, which could provide valuable information on the physiological stress reactions. Methods: Physiological responses were measured in the Maastricht Acute Stress Test (MAST), comprising alternating trials of cold pressor (inducing physiological stress and pain) and mental arithmetics (eliciting cognitive and social-evaluative stress). The responses in these subtasks were compared to each other and to the baseline through mixed model analysis. Subsequently, stress type detection was conducted with a comprehensive analysis of several machine learning components affecting classification. Finally, explainable artificial intelligence (XAI) methods were applied to analyze the influence of physiological features on model behavior. Results: Most of the investigated physiological reactions were specific to the stressors, and the subtasks could be distinguished from baseline with up to 86.5% balanced accuracy. The choice of the physiological signals to measure (up to 25%-point difference in balanced accuracy) and the selection of features (up to 7%-point difference) were the two key components in classification. Reflection of the XAI analysis to mixed model results and human physiology revealed that the stress detection model concentrated on physiological features relevant for the two stressors. Discussion: The findings confirm that multimodal machine learning classification can detect different types of stress reactions from baseline while focusing on physiologically sensible changes. Since the measured signals and feature selection affected classification performance the most, data analytic choices left limited input information uncompensated.

Positive affect state is a good predictor of movement and stress: combining data from ESM/EMA, mobile HRV measurements and trait questionnaires.

Personality describes the average behaviour and responses of individuals across situations; but personality traits are often poor predictors of behaviour in specific situations. This is known as the "personality paradox". We evaluated the interrelations between various trait and state variables in participants' everyday lives. As state measures, we used 1) experience sampling methodology (ESM/EMA) to measure perceived affect, stress, and presence of social company; and 2) heart rate variability and 3) real-time movement (accelerometer data) to indicate physiological stress and physical movement. These data were linked with self-report measures of personality and personality-like traits. Trait variables predicted affect states and multiple associations were found: traits neuroticism and rumination decreased positive affect state and increased negative affect state. Positive affect state, in turn, was the strongest predictor of observed movement. Positive affect was also associated with heart rate and heart rate variability (HRV). Negative affect, in turn, was not associated with neither movement, HR or HRV. The study provides evidence on the influence of personality-like traits and social context to affect states, and, in turn, their influence to movement and stress variables.

Personalized mental stress detection with self-organizing map: From laboratory to the field.

Stress has become a major health concern and there is a need to study and develop new digital means for real-time stress detection. Currently, the majority of stress detection research is using population based approaches that lack the capability to adapt to individual differences. They also use supervised learning methods, requiring extensive labeling of training data, and they are typically tested on data collected in a laboratory and thus do not generalize to field conditions. To address these issues, we present multiple personalized models based on an unsupervised algorithm, the Self-Organizing Map (SOM), and we propose an algorithmic pipeline to apply the method for both laboratory and field data. The performance is evaluated on a dataset of physiological measurements from a laboratory test and on a field dataset consisting of four weeks of physiological and smartphone usage data. In these tests, the performance on the field data was steady across the different personalization levels (accuracy around 60%) and a fully personalized model performed the best on the laboratory data, achieving accuracy of 92% which is comparable to state-of-the-art supervised classifiers. These results demonstrate the feasibility of SOM in personalized mental stress detection both in constrained and free-living environment.

Influence of Personality and Differences in Stress Processing Among Finnish Students on Interest to Use a Mobile Stress Management App: Survey Study.

BACKGROUND: Excessive stress has a negative impact on many aspects of life for both individuals and societies, from studying and working to health and well-being. Each individual has their unique level of stress-proneness, and positive or negative outcomes of stress may be affected by it. Technology-aided interventions have potential efficacy in the self-management of stress. However, current Web-based or mobile stress management solutions may not reach the individuals that would need them the most, that is, stress-sensitive people. OBJECTIVE: The aim of this study was to examine how personality is associated with stress among Finnish university students and their interest to use apps that help in managing stress. METHODS: We used 2 structured online questionnaires (combined, n=1001) that were advertised in the University of Helsinki's mailing lists. The first questionnaire (n=635) was used to investigate intercorrelations between the Big Five personality variables (neuroticism, extraversion, openness, agreeableness, and conscientiousness) and other stress-related background variables. The second questionnaire (n=366) was used to study intercorrelations between the above-mentioned study variables and interest in using stress management apps. RESULTS: The quantitative findings of the first questionnaire showed that higher levels of extraversion, agreeableness, and conscientiousness were associated with lower self-reported stress. Neuroticism, in turn, was found to be strongly associated with rumination, anxiety, and depression. The findings of the second questionnaire indicated that individuals characterized by the Big Five personality traits of neuroticism and agreeableness were particularly interested to use stress management apps (r=.27, P<.001 and r=.11, P=.032, respectively). Moreover, the binary logistic regression analysis revealed that when a person's neuroticism is one SD above average (ie, it is higher than among 84% of people), the person has roughly 2 times higher odds of being interested in using a stress management app. Respectively, when a person's agreeableness is one SD above average, the person has almost 1.4 times higher odds of being interested in using a stress management app. CONCLUSIONS: Our results indicated that personality traits may have an influence on the adoption interest of stress management apps. Individuals with high neuroticism are, according to our results, adaptive in the sense that they are interested in using stress management apps that may benefit them. On the contrary, low agreeableness may lead to lower interest to use the mobile stress management apps. The practical implication is that future mobile stress interventions should meaningfully be adjusted to improve user engagement and support health even among less-motivated users, for instance, to successfully engage individuals with low agreeableness.

Accelerometry-based berg balance scale score estimation.

The objective of the study was to investigate the validity of 3-D-accelerometry-based Berg balance scale (BBS) score estimation. In particular, acceleration patterns of BBS tasks and gait were the targets of analysis. Accelerations of the lower back were measured during execution of the BBS test and corridor walking for 54 subjects, consisting of neurological patients, older adults, and healthy young persons. The BBS score was estimated from one to three BBS tasks and from gait-related data, separately, through assessment of the similarity of acceleration patterns between subjects. The work also validated both approaches' ability to classify subjects into high- and low-fall-risk groups. The gait-based method yielded the best BBS score estimates and the most accurate BBS-task-based estimates were produced with the stand to sit, reaching, and picking object tasks. The proposed gait-based method can identify subjects with high or low risk of falling with an accuracy of 77.8% and 96.6%, respectively, and the BBS-task based method with corresponding accuracy of 89.5% and 62.1%.

A computationally light classification method for mobile wellness platforms.

The core of activity recognition in mobile wellness devices is a classification engine which maps observations from sensors to estimated classes. There exists a vast number of different classification algorithms that can be used for this purpose in the machine learning literature. Unfortunately, the computational and space requirements of these methods are often too high for the current mobile devices. In this paper we study a simple linear classifier and find, automatically with SFS and SFFS feature selection methods, a suitable set of features to be used with the classification method. The results show that the simple classifier performs comparable to more complex nonlinear k-Nearest Neighbor Classifier. This depicts great potential in implementing the classifier in small mobile wellness devices.

Using heart rate monitors for personal wellness--the user experience perspective.

In this paper, we present a user study of the heart rate monitors (HRM), which is a commonly used personal wellness technology. HRMs have been used over several years for adjusting the exercise level and in the advanced form, also for measuring the users' fitness level and its progress. The user study included interviews with the HRM users and a survey with over 860 active or former users. We identified four different types of HRM users based on the current usage activity and their perceived progress of usage motivation within time, and compared their perceptions of the strengths and weaknesses related to HRMs. The findings provide valuable information for understanding the end-user needs and background knowledge for developing personal wellness technologies and applications further.

Activity classification using realistic data from wearable sensors.

Automatic classification of everyday activities can be used for promotion of health-enhancing physical activities and a healthier lifestyle. In this paper, methods used for classification of everyday activities like walking, running, and cycling are described. The aim of the study was to find out how to recognize activities, which sensors are useful and what kind of signal processing and classification is required. A large and realistic data library of sensor data was collected. Sixteen test persons took part in the data collection, resulting in approximately 31 h of annotated, 35-channel data recorded in an everyday environment. The test persons carried a set of wearable sensors while performing several activities during the 2-h measurement session. Classification results of three classifiers are shown: custom decision tree, automatically generated decision tree, and artificial neural network. The classification accuracies using leave-one-subject-out cross validation range from 58 to 97% for custom decision tree classifier, from 56 to 97% for automatically generated decision tree, and from 22 to 96% for artificial neural network. Total classification accuracy is 82 % for custom decision tree classifier, 86% for automatically generated decision tree, and 82% for artificial neural network.