Predicting the severity of mood and neuropsychiatric symptoms from digital biomarkers using wearable physiological data and deep learning

Neuropsychiatric symptoms (NPS) and mood disorders are common in individuals with mild cognitive impairment (MCI) and increase the risk of progression to dementia. Wearable devices collecting physiological and behavioral data can help in remote, passive, and continuous monitoring of moods and NPS, overcoming limitations and inconveniences of current assessment methods. In this longitudinal study, we examined the predictive ability of digital biomarkers based on sensor data from a wrist-worn wearable to determine the severity of NPS and mood disorders on a daily basis in older adults with predominant MCI. In addition to conventional physiological biomarkers, such as heart rate variability and skin conductance levels, we leveraged deep-learning features derived from physiological data using a self-supervised convolutional autoencoder. Models combining common digital biomarkers and deep features predicted depression severity scores with a correlation of r = 0.73 on average, total severity of mood disorder symptoms with r = 0.67, and mild behavioral impairment scores with r = 0.69 in the study population. Our findings demonstrated the potential of physiological biomarkers collected from wearables and deep learning methods to be used for the continuous and unobtrusive assessments of mental health symptoms in older adults, including those with MCI. TRIAL REGISTRATION: This trial was registered with ClinicalTrials.gov (NCT05059353) on September 28, 2021, titled "Effectiveness and Safety of a Digitally Based Multidomain Intervention for Mild Cognitive Impairment".

Recent advances in the precision control strategy of artificial pancreas

The scientific diagnosis and treatment of patients with diabetes require frequent blood glucose testing and insulin delivery to normoglycemia. Therefore, an artificial pancreas with a continuous blood glucose (BG) monitoring function is an urgent research target in the medical industry. The problem of closed-loop algorithmic control of the BG with a time delay is a key and difficult issue that needs to be overcome in the development of an artificial pancreas. Firstly, the composition, structure, and control characteristics of the artificial pancreas are introduced. Subsequently, the research progress of artificial pancreas control algorithms is reviewed, and the characteristics, advantages, and disadvantages of proportional-integral-differential control, model predictive control, and artificial intelligence control are compared and analyzed to determine whether they are suitable for the practical application of the artificial pancreas. Additionally, key advancements in areas such as blood glucose data monitoring, adaptive models, wearable devices, and fully automated artificial pancreas systems are also reviewed. Finally, this review highlights that meal prediction, control safety, integration, streamlining the optimization of control algorithms, constant temperature preservation of insulin, and dual-hormone artificial pancreas are issues that require further attention in the future.

Artificial Intelligence and Machine Learning for Improving Glycemic Control in Diabetes: Best Practices, Pitfalls, and Opportunities

OBJECTIVE

Artificial intelligence and machine learning are transforming many fields including medicine. In diabetes, robust biosensing technologies and automated insulin delivery therapies have created a substantial opportunity to improve health. While the number of manuscripts addressing the topic of applying machine learning to diabetes has grown in recent years, there has been a lack of consistency in the methods, metrics, and data used to train and evaluate these algorithms. This manuscript provides consensus guidelines for machine learning practitioners in the field of diabetes, including best practice recommended approaches and warnings about pitfalls to avoid.

METHODS

Algorithmic approaches are reviewed and benefits of different algorithms are discussed including importance of clinical accuracy, explainability, interpretability, and personalization. We review the most common features used in machine learning applications in diabetes glucose control and provide an open-source library of functions for calculating features, as well as a framework for specifying data sets using data sheets. A review of current data sets available for training algorithms is provided as well as an online repository of data sources.

SIGNIFICANCE

These consensus guidelines are designed to improve performance and translatability of new machine learning algorithms developed in the field of diabetes for engineers and data scientists.

Multivariable Automated Insulin Delivery System for Handling Planned and Spontaneous Physical Activities

BACKGROUND

Hybrid closed-loop control of glucose levels in people with type 1 diabetes mellitus (T1D) is limited by the requirements on users to manually announce physical activity (PA) and meals to the artificial pancreas system. Multivariable automated insulin delivery (mvAID) systems that can handle unannounced PAs and meals without any manual announcements by the user can improve glycemic control by modulating insulin dosing in response to the occurrence and intensity of spontaneous physical activities.

METHODS

An mvAID system is developed to supplement the glucose measurements with additional physiological signals from a wristband device, with the signals analyzed using artificial intelligence algorithms to automatically detect the occurrence of PA and estimate its intensity. This additional information gained from the physiological signals enables more proactive insulin dosing adjustments in response to both planned exercise and spontaneous unanticipated physical activities.

RESULTS

In silico studies of the mvAID illustrate the safety and efficacy of the system. The mvAID is translated to pilot clinical studies to assess its performance, and the clinical experiments demonstrate an increased time in range and reduced risk of hypoglycemia following unannounced PA and meals.

CONCLUSIONS

The mvAID systems can increase the safety and efficacy of insulin delivery in the presence of unannounced physical activities and meals, leading to improved lives and less burden on people with T1D.

A physiological signal database of children with different special needs for stress recognition

This study presents a new dataset AKTIVES for evaluating the methods for stress detection and game reaction using physiological signals. We collected data from 25 children with obstetric brachial plexus injury, dyslexia, and intellectual disabilities, and typically developed children during game therapy. A wristband was used to record physiological data (blood volume pulse (BVP), electrodermal activity (EDA), and skin temperature (ST)). Furthermore, the facial expressions of children were recorded. Three experts watched the children's videos, and physiological data is labeled "Stress/No Stress" and "Reaction/No Reaction", according to the videos. The technical validation supported high-quality signals and showed consistency between the experts.

Generalizable machine learning for stress monitoring from wearable devices: A systematic literature review

INTRODUCTION

Wearable sensors have shown promise as a non-intrusive method for collecting biomarkers that may correlate with levels of elevated stress. Stressors cause a variety of biological responses, and these physiological reactions can be measured using biomarkers including Heart Rate Variability (HRV), Electrodermal Activity (EDA) and Heart Rate (HR) that represent the stress response from the Hypothalamic-Pituitary-Adrenal (HPA) axis, the Autonomic Nervous System (ANS), and the immune system. While Cortisol response magnitude remains the gold standard indicator for stress assessment [1], recent advances in wearable technologies have resulted in the availability of a number of consumer devices capable of recording HRV, EDA and HR sensor biomarkers, amongst other signals. At the same time, researchers have been applying machine learning techniques to the recorded biomarkers in order to build models that may be able to predict elevated levels of stress.

OBJECTIVE

The aim of this review is to provide an overview of machine learning techniques utilized in prior research with a specific focus on model generalization when using these public datasets as training data. We also shed light on the challenges and opportunities that machine learning-enabled stress monitoring and detection face.

METHODS

This study reviewed published works contributing and/or using public datasets designed for detecting stress and their associated machine learning methods. The electronic databases of Google Scholar, Crossref, DOAJ and PubMed were searched for relevant articles and a total of 33 articles were identified and included in the final analysis. The reviewed works were synthesized into three categories of publicly available stress datasets, machine learning techniques applied using those, and future research directions. For the machine learning studies reviewed, we provide an analysis of their approach to results validation and model generalization. The quality assessment of the included studies was conducted in accordance with the IJMEDI checklist [2].

RESULTS

A number of public datasets were identified that are labeled for stress detection. These datasets were most commonly produced from sensor biomarker data recorded using the Empatica E4 device, a well-studied, medical-grade wrist-worn wearable that provides sensor biomarkers most notable to correlate with elevated levels of stress. Most of the reviewed datasets contain less than twenty-four hours of data, and the varied experimental conditions and labeling methodologies potentially limit their ability to generalize for unseen data. In addition, we discuss that previous works show shortcomings in areas such as their labeling protocols, lack of statistical power, validity of stress biomarkers, and model generalization ability.

CONCLUSION

Health tracking and monitoring using wearable devices is growing in popularity, while the generalization of existing machine learning models still requires further study, and research in this area will continue to provide improvements as newer and more substantial datasets become available.

Multi-Task Classification of Physical Activity and Acute Psychological Stress for Advanced Diabetes Treatment

Wearable sensor data can be integrated and interpreted to improve the treatment of chronic conditions, such as diabetes, by enabling adjustments in treatment decisions based on physical activity and psychological stress assessments. The challenges in using biological analytes to frequently detect physical activity (PA) and acute psychological stress (APS) in daily life necessitate the use of data from noninvasive sensors in wearable devices, such as wristbands. We developed a recurrent multi-task deep neural network (NN) with long-short-term-memory architecture to integrate data from multiple sensors (blood volume pulse, skin temperature, galvanic skin response, three-axis accelerometers) and simultaneously detect and classify the type of PA, namely, sedentary state, treadmill run, stationary bike, and APS, such as non-stress, emotional anxiety stress, mental stress, and estimate the energy expenditure (EE). The objective was to assess the feasibility of using the multi-task recurrent NN (RNN) rather than independent RNNs for detection and classification of AP and APS. The multi-task RNN achieves comparable performance to independent RNNs, with the multi-task RNN having F1 scores of 98.00% for PA and 98.97% for APS, and a root mean square error (RMSE) of 0.728 calhr.kg for EE estimation for testing data. The independent RNNs have F1 scores of 99.64% for PA and 98.83% for APS, and an RMSE of 0.666 calhr.kg for EE estimation. The results indicate that a multi-task RNN can effectively interpret the signals from wearable sensors. Additionally, we developed individual and multi-task extreme gradient boosting (XGBoost) for separate and simultaneous classification of PA types and APS types. Multi-task XGBoost achieved F1 scores of 99.89% and 98.31% for the classification of PA types and APS types, respectively, while the independent XGBoost achieved F1 scores of 99.68% and 96.77%, respectively. The results indicate that both multi-task RNN and XGBoost can be used for the detection and classification of PA and APS without loss of performance with respect to individual separate classification systems. People with diabetes can achieve better outcomes and quality of life by including physical activity and psychological stress assessments in treatment decision-making.

Internal model control based module for the elimination of meal and exercise announcements in hybrid artificial pancreas systems

BACKGROUND AND OBJECTIVES

Hybrid artificial pancreas systems outperform current insulin pump therapies in blood glucose regulation in type 1 diabetes. However, subjects still have to inform the system about meals intake and exercise to achieve reasonable control. These patient announcements may result in overburden and compromise controller performance if not provided timely and accurately. Here, a hybrid artificial pancreas is extended with an add-on module that releases subjects from meals and exercise announcements.

METHODS

The add-on module consists of an internal-model controller that generates a "virtual" control action to compensate for disturbances. This "virtual" action is converted into insulin delivery, rescue carbohydrates suggestions, or insulin-on-board limitations, depending on a switching logic based on glucose measurements and predictions. The controller parameters are tuned by optimization and then related to standard parameters from the open-loop therapy. This module is implemented in a hybrid artificial pancreas system proposed by our research group for validation. This hybrid system extended with the add-on module is compared with the hybrid controller with carbohydrate counting errors (hybrid) and the hybrid controller with an alternative unannounced meal compensation module based on a meal detection algorithm (meal detector). The validation used the educational version of the UVa/Padova simulator to simulate the three controllers under two scenarios: one with only meals and another with meals and exercise. The exercise was modeled as a temporal increase of the insulin sensitivity resulting in the glucose drop usually related to an aerobic exercise.

RESULTS

For the scenario with only meals, the three controllers achieved similar time in range (proposed: 85.1 [77.9,88.1]%, hybrid: 84.0 [75.9,86.4]%, meal detector: 81.9 [79.3,83.8]%, median [interquartile range]) with low time in moderate hypoglycemia. Under the scenario with meals and exercise, the proposed module reduces 4.61% the time in hypoglycemia achieved with the other controllers, suggesting an acceptable amount of rescues (27.2 [23.7, 31.0] g).

CONCLUSIONS

The proposed add-on module achieved promising results: it outperformed the meal-detector-based controller, even achieving a postprandial performance as good as the hybrid controller (with carbohydrate counting errors). Also, the rescue suggestion feature of the module mitigated exercise-induced hypoglycemia with admissible rescue amounts.

Activity detection and classification from wristband accelerometer data collected on people with type 1 diabetes in free-living conditions

This paper introduces methods to estimate aspects of physical activity and sedentary behavior from three-axis accelerometer data collected with a wrist-worn device at a sampling rate of 32 [Hz] on adults with type 1 diabetes (T1D) in free-living conditions. In particular, we present two methods able to detect and grade activity based on its intensity and individual fitness as sedentary, mild, moderate or vigorous, and a method that performs activity classification in a supervised learning framework to predict specific user behaviors. Population results for activity level grading show multi-class average accuracy of 99.99%, precision of 98.0 ± 2.2%, recall of 97.9 ± 3.5% and F1 score of 0.9 ± 0.0. As for the specific behavior prediction, our best performing classifier, gave population multi-class average accuracy of 92.43 ± 10.32%, precision of 92.94 ± 9.80%, recall of 92.20 ± 10.16% and F1 score of 92.56 ± 9.94%. Our investigation showed that physical activity and sedentary behavior can be detected, graded and classified with good accuracy and precision from three-axial accelerometer data collected in free-living conditions on people with T1D. This is particularly significant in the context of automated glucose control systems for diabetes, in that the methods we propose have the potential to inform changes in treatment parameters in response to the intensity of physical activity, allowing patients to meet their glycemic targets.