Childhood mental health problems are impairing, predictive of health problems later in life, and becoming increasingly prevalent. A critical first step toward addressing this growing crisis is facilitating more widespread screening; however, gold-standard assessments remain subjective, time-consuming, largely unable to capture subthreshold conditions and comorbidity, and limited by access to the clinical experts needed to provide and interpret the results. In response, researchers have developed digital phenotype screening tools that capture and utilize objective physiological and behavioral measures to augment traditional mental health screening. However, the efficacy of these tools has traditionally been evaluated on their ability to predict mental health diagnoses. In contrast, in this work and in line with the Research Domain Criteria (RDoC) framework, we explore the utility of these objective measures for quantifying trans diagnostic severity of impairment across a range of widely used clinical measurements. Using canonical correlation analysis, we find that linear combinations of movement and audio features extracted from smartphone sensor data collected during short (< 7 minutes) objective assessments are significantly correlated with a range of widely used clinical measurements. These findings suggest that easy-to-collect objective physiological and behavioral measures are indicative of severity across a range of psychopathologies.

As the mental health of college students continues to worsen, identifying students who need additional support and timely, targeted intervention is vital for college campuses. Leveraging Oura rings, a validated consumer wearable, we identify and examine heart rate variability (HRV) trajectories in 233 first-semester college students to characterize patterns of physiological well-being. The dominant trajectory among students was a decline in HRV (43.25%), followed by students who fluctuated but ultimately improved their HRV (24.46%), and students who remained consistent around their baseline, but experienced low points at the end of October and the end of the semester (19.3%). The least common HRV trajectory was a continuous improvement across the semester (12.9%). Findings suggest strong associations between HRV trajectories and experiential backgrounds, including first-generation status and total reported adverse life experiences, as well as weekly self-reported stress levels, highlighting the potential of wearables for providing real-time insight that could enable targeted intervention for student well-being.

The transition to college represents a great change for adolescents, with new challenges living independently and increased academic pressure and subsequent increases in anxiety, depression, and stress. Wearable sensors offer an opportunity to passively and continuously collect physiological quantities that underly these mental health changes. In this study we analyze smart-ring (Oura) heart rate variability (HRV) data from 108 students continuously for a period of 6 months during their first year of college. Students self-reported their mental health diagnoses upon college entry and completed surveys (DASS-21) about their total anxiety, depression and stress symptoms every two weeks. Linear mixed models reveal that increases in mental health symptoms across weeks was associated with decreases in HRV (p = 0.01), consistent with prior research. Interestingly, there was a significant interaction driven mainly by females such that students with an anxiety disorder showed the opposite relationship between mental health symptom changes and HRV (p < 0.05). For females with an anxiety disorder, increases in mental health symptoms were associated with increases in HRV, possibly indicating different coping strategies. In conclusion, HRV measured via smart-rings is a promising digital biomarker for tracking fluctuations in mental health symptoms.

Childhood mental health problems are common, impairing, and can become chronic if left untreated. Children are not reliable reporters of their emotional and behavioral health, and caregivers often unintentionally under- or over-report child symptoms, making assessment challenging. Objective physiological and behavioral measures of emotional and behavioral health are emerging. However, these methods typically require specialized equipment and expertise in data and sensor engineering to administer and analyze. To address this challenge, we have developed the ChAMP (Childhood Assessment and Management of digital Phenotypes) System, which includes a mobile application for collecting movement and audio data during a battery of mood induction tasks and an open-source platform for extracting digital biomarkers. As proof of principle, we present ChAMP System data from 101 children 4-8 years old, with and without diagnosed mental health disorders. Machine learning models trained on these data detect the presence of specific disorders with 70-73% balanced accuracy, with similar results to clinical thresholds on established parent-report measures (63-82% balanced accuracy). Features favored in model architectures are described using Shapley Additive Explanations (SHAP). Canonical Correlation Analysis reveals moderate to strong associations between predictors of each disorder and associated symptom severity (r = .51-.83). The open-source ChAMP System provides clinically-relevant digital biomarkers that may later complement parent-report measures of emotional and behavioral health for detecting kids with underlying mental health conditions and lowers the barrier to entry for researchers interested in exploring digital phenotyping of childhood mental health.

Long-duration surface electromyography (sEMG) recordings are not considered in recent consensus on the appropriate methods for sEMG normalization. Here we find that sEMG data normalized by the gold standard, maximum voluntary contraction, fails to appropriately represent the amplitude recorded from walking bouts over an 18-hour period, suggesting that normalization reference values may not remain valid over long periods. To address this limitation, we present a new adaptive method for sEMG normalization that leverages data collected during typical daily activities. We explore several candidate daily activities for performing this normalization, and assess their ability to resolve expected sEMG amplitude changes with stride time and activity intensity. Normalization to walking, and particularly to a self-selected comfortable speed, yields the best results.

Panic attacks are an impairing mental health problem that affects 11% of adults every year [1]. Those who suffer from panic attacks often do not seek psychological treatment, citing the inability to receive care during their attacks as a contributing factor. A digital medicine solution which provides an accessible, real-time mobile health (mHealth) biofeedback intervention for panic attacks may address this problem. Critical to this approach are methods for capturing physiological arousal during an attack. Herein, we validate an algorithm for capturing physiological arousal using smartphone video of the fingertip. Results demonstrate that the algorithm is able to estimate heart rates that are highly correlated with ECG-derived values (r > 0.99), effectively reject low-quality data often captured outside of controlled laboratory environments (AUC > 0.90), and resolve the physiological arousal experienced during a panic attack. Moreover, patient reported measures indicate that this measurement modality is feasible during panic attacks, and the act of taking the measurement may stop the attack. These results point toward the need for future development and clinical evaluation of this mHealth intervention for preventing panic attacks.

Childhood internalizing disorders, like anxiety and depression, are common, impairing, and difficult to detect. Universal childhood mental health screening has been recommended, but new technologies are needed to provide objective detection. Instrumented mood induction tasks, designed to press children for specific behavioral responses, have emerged as means for detecting childhood internalizing psychopathology. In our previous work, we leveraged machine learning to identify digital phenotypes of childhood internalizing psychopathology from movement and voice data collected during negative valence tasks (pressing for anxiety and fear). In this work, we develop a digital phenotype for childhood internalizing disorders based on wearable inertial sensor data recorded from a Positive Valence task during which a child plays with bubbles. We find that a phenotype derived from features that capture reward responsiveness is able to accurately detect children with underlying internalizing psychopathology (AUC = 0.81). In so doing, we explore the impact of a variety of feature sets computed from wearable sensors deployed to two body locations on phenotype performance across two phases of the task. We further consider this novel digital phenotype in the context of our previous Negative Valence digital phenotypes and find that each task brings unique information to the problem of detecting childhood internalizing psychopathology, capturing different problems and disorder subtypes. Collectively, these results provide preliminary evidence for a mood induction task battery to develop a novel diagnostic for childhood internalizing disorders.

Falls are a significant problem for persons with multiple sclerosis (PwMS). Yet fall prevention interventions are not often prescribed until after a fall has been reported to a healthcare provider. While still nascent, objective fall risk assessments could help in prescribing preventative interventions. To this end, retrospective fall status classification commonly serves as an intermediate step in developing prospective fall risk assessments. Previous research has identified measures of gait biomechanics that differ between PwMS who have fallen and those who have not, but these biomechanical indices have not yet been leveraged to detect PwMS who have fallen. Moreover, they require the use of laboratory-based measurement technologies, which prevent clinical deployment. Here we demonstrate that a bidirectional long short-term (BiLSTM) memory deep neural network was able to identify PwMS who have recently fallen with good performance (AUC of 0.88) based on accelerometer data recorded from two wearable sensors during a one-minute walking task. These results provide substantial improvements over machine learning models trained on spatiotemporal gait parameters (21% improvement in AUC), statistical features from the wearable sensor data (16%), and patient-reported (19%) and neurologist-administered (24%) measures in this sample. The success and simplicity (two wearable sensors, only one-minute of walking) of this approach indicates the promise of inexpensive wearable sensors for capturing fall risk in PwMS.

This paper presents an automated, wearable sensor-based remote gait analysis method and demonstrates its utility by evaluating gait in patients recovering from anterior cruciate ligament (ACL) reconstruction. Patients wore a single wearable sensor over the rectus femoris of each leg to collect over 15 hours of 3-axis accelerometer and surface electromyography data during daily life. A support vector machine classifier was used to identify four second windows of walking activity. Across all subjects 10,451 strides were extracted from these windows and categorized as occurring during either fast or slow walking. Muscle activation and 3-axis thigh acceleration time-series for each stride were resampled as a percentage of stride cycle and ensemble curves from both legs were compared using correlation as an index of gait pattern symmetry. Our results suggest the proposed method successfully identifies time-series gait asymmetries between affected and unaffected legs when comparing patients early post-surgery ( weeks) and later ( weeks) in recovery for each walking speed. These results point toward future use of this approach as a digital biomarker for rehabilitation progress in this population.

Panic attacks are an impairing mental health problem that affects 11% of adults every year. Those who suffer from panic attacks often do not seek psychological treatment, citing the inability to receive care during their attacks as a contributing factor. Herein, we introduce a mobile health (mHealth) biofeedback system that enables treatment of panic attacks wherever and whenever they occur and describe the results of an initial feasibility study. We find that only three of nine chronic panic attack sufferers experienced a panic attack during the study, potentially suggesting a preventative placebo effect common to similar pharmacological interventions. Of the four panic attacks observed, subjects noted that the act of using their phone to record their physiology during the attack helped to stop the attack. While preliminary, these results point toward the need for future development of this mHealth system and a future clinical study to assess its efficacy for preventing panic attacks.

Childhood anxiety and depression often go undiagnosed. If left untreated these conditions, collectively known as internalizing disorders, are associated with long-term negative outcomes including substance abuse and increased risk for suicide. This paper presents a new approach for identifying young children with internalizing disorders using a 3-min speech task. We show that machine learning analysis of audio data from the task can be used to identify children with an internalizing disorder with 80% accuracy (54% sensitivity, 93% specificity). The speech features most discriminative of internalizing disorder are analyzed in detail, showing that affected children exhibit especially low-pitch voices, with repeatable speech inflections and content, and high-pitched response to surprising stimuli relative to controls. This new tool is shown to outperform clinical thresholds on parent-reported child symptoms, which identify children with an internalizing disorder with lower accuracy (67-77% versus 80%), and similar specificity (85-100% versus 93%), and sensitivity (0-58% versus 54%) in this sample. These results point toward the future use of this approach for screening children for internalizing disorders so that interventions can be deployed when they have the highest chance for long-term success.

We examine leg motion and quadriceps muscle activation patterns during free-living activity in a cross-sectional sample of patients recovering from anterior cruciate ligament (ACL) reconstruction using wearable sensors. Analysis suggests that this technology is feasible, and captures expected differences in daily muscle activity between the affected and contralateral limbs within and across patients and, during gait, differences in muscle activation patterns and kinematics across patients associated with recovery progress. Results point toward the need for future studies to explore these indicators of rehabilitation progress longitudinally in patients recovering from ACL reconstruction.

Temporal phases of threat response, including potential threat (anxiety), acute threat (startle, fear), and post-threat response modulation, have been identified as the underlying markers of anxiety disorders. Objective measures of response during these phases may help identify children at risk for anxiety; however, the complexity of current assessment techniques prevent their adoption in many research and clinical contexts. We propose an alternative technology, an inertial measurement unit (IMU), that enables noninvasive measurement of the movements associated with threat response, and test its ability to detect threat response phases in young children at a heightened risk for developing anxiety. We quantified the motion of 18 children (3-7 years old) during an anxiety-/fear-provoking behavioral task using an IMU. Specifically, measurements from a single IMU secured to the child's waist were used to extract root-mean-square acceleration and angular velocity in the horizontal and vertical directions, and tilt and yaw range of motion during each threat response phase. IMU measurements detected expected differences in child motion by threat phase. Additionally, potential threat motion was positively correlated to familial anxiety risk, startle range of motion was positively correlated with child internalizing symptoms, and response modulation motion was negatively correlated to familial anxiety risk. Results suggest differential theory-driven threat response phases and support previous literature connecting maternal child risk to anxiety with behavioral measures using more feasible objective methods. This is the first study demonstrating the utility of an IMU for characterizing the motion of young children to mark the phases of threat response modulation. The technique provides a novel and objective measure of threat response for mental health researchers.