Physics-Informed Transfer Learning to Enhance Sleep Staging

OBJECTIVE

At-home sleep staging using wearable medical sensors poses a viable alternative to in-hospital polysomnography due to its lower cost and lower disruption to the daily lives of patients, especially in the case of long-term monitoring. Machine learning with wearables however is difficult due to the paucity of data from wearable sensors, making automation a challenge. Transfer learning from hospital polysomnograms can boost performance, but is still hindered by differences between wearable and in-hospital EEG resulting in part from differing electrode placement. We improve transfer learning performance by using electrophysiological models of a human head to generate synthetic EEG resembling EEG from a wearable sensor.

METHODS

The data generation method utilizes Low-Resolution Electromagnetic Tomography Analysis (LORETA). Real EEG from standard in- hospital recordings is first mapped to point currents within the brain using LORETA, after which the point currents are used to estimate EEG that would have been recorded using a wearable sensor at any given point on the head.

RESULTS

Augmenting source datasets with synthetic data statistically significantly boosted accuracy on a wearable sleep staging task from 80.8% to 81.3% on average, depending on the transfer learning parameters and data sources.

CONCLUSION

Machine learning performance can be improved using data synthesized using physical models.

SIGNIFICANCE

Our approach represents a new form of transfer learning and demonstrates that incorporating domain knowledge of electrophysiological modeling can improve machine learning results for sleep staging tasks. We expect this approach to be particularly useful for EEG data which is hard to collect, or which is obtained using unusual electrode configurations.

Heart Rate Variability, Deceleration Capacity of Heart Rate, and Death: A Veteran Twins Study

BACKGROUND

Autonomic function can be measured noninvasively using heart rate variability (HRV), which indexes overall sympathovagal balance. Deceleration capacity (DC) of heart rate is a more specific metric of vagal modulation. Higher values of these measures have been associated with reduced mortality risk primarily in patients with cardiovascular disease, but their significance in community samples is less clear.

METHODS AND RESULTS

This prospective twin study followed 501 members from the VET (Vietnam Era Twin) registry. At baseline, frequency domain HRV and DC were measured from 24-hour Holter ECGs. During an average 12-year follow-up, all-cause death was assessed via the National Death Index. Multivariable Cox frailty models with random effect for twin pair were used to examine the hazard ratios of death per 1-SD increase in log-transformed autonomic metrics. Both in the overall sample and comparing twins within pairs, higher values of low-frequency HRV and DC were significantly associated with lower hazards of all-cause death. In within-pair analysis, after adjusting for baseline factors, there was a 22% and 27% lower hazard of death per 1-SD increment in low-frequency HRV and DC, respectively. Higher low-frequency HRV and DC, measured during both daytime and nighttime, were associated with decreased hazard of death, but daytime measures showed numerically stronger associations. Results did not substantially vary by zygosity.

CONCLUSIONS

Autonomic inflexibility, and especially vagal withdrawal, are important mechanistic pathways of general mortality risk, independent of familial and genetic factors.

Disentangling Visual Exploration Differences in Cognitive Impairment

OBJECTIVE

Individuals with cognitive impairment (CI) exhibit different oculomotor functions and viewing behaviors. In this work we aimed to quantify the differences in these functions with CI severity, and assess general CI and specific cognitive functions related to visual exploration behaviors.

METHODS

A validated passive viewing memory test with eyetracking was administered to 348 healthy controls and CI individuals. Spatiotemporal properties of the scanpath, the semantic category of the viewed regions, and other composite features were extracted from the estimated eyegaze locations on the corresponding pictures displayed during the test. These features were then used to characterize viewing patterns, classify cognitive impairment, and estimate scores in various neuropsychological tests using machine learning.

RESULTS

Statistically significant differences in spatial, spatiotemporal, and semantic features were found between healthy controls and individuals with CI. The CI group spent more time gazing at the center of the image, looked at more regions of interest (ROI), transitioned less often between ROI yet in a more unpredictable manner, and exhibited different semantic preferences. A combination of these features achieved an area under the receiver-operator curve of 0.78 in differentiating CI individuals from controls. Statistically significant correlations were identified between actual and estimated CI scores and other neuropsychological tests.

CONCLUSION

Evaluating visual exploration behaviors provided quantitative and systematic evidence of differences in CI individuals, leading to an improved approach for passive cognitive impairment screening.

SIGNIFICANCE

The proposed passive, accessible, and scalable approach could help with earlier detection and a better understanding of cognitive impairment.

Impedance cardiogram based exploration of cardiac mechanisms in post-traumatic stress disorder during trauma recall

Post-traumatic stress disorder (PTSD) is an independent risk factor for developing heart failure; however, the underlying cardiac mechanisms are still elusive. This study aims to evaluate the real-time effects of experimentally induced PTSD symptom activation on various cardiac contractility and autonomic measures. We recorded synchronized electrocardiogram and impedance cardiogram from 137 male veterans (17 PTSD, 120 non-PTSD; 48 twin pairs, 41 unpaired singles) during a laboratory-based traumatic reminder stressor. To identify the parameters describing the cardiac mechanisms by which trauma reminders can create stress on the heart, we utilized a feature selection mechanism along with a random forest classifier distinguishing PTSD and non-PTSD. We extracted 99 parameters, including 76 biosignal-based and 23 sociodemographic, medical history, and psychiatric diagnosis features. A subject/twin-wise stratified nested cross-validation procedure was used for parameter tuning and model assessment to identify the important parameters. The identified parameters included biomarkers such as pre-ejection period, acceleration index, velocity index, Heather index, and several physiology-agnostic features. These identified parameters during trauma recall suggested a combination of increased sympathetic nervous system (SNS) activity and deteriorated cardiac contractility that may increase the heart failure risk for PTSD. This indicates that the PTSD symptom activation associates with real-time reductions in several cardiac contractility measures despite SNS activation. This finding may be useful in future cardiac prevention efforts.

Sex differences in response inhibition-related neural predictors of PTSD in recent trauma-exposed civilians

BACKGROUND

Females are more likely to develop posttraumatic stress disorder (PTSD) than males. Impaired inhibition has been identified as mechanism for PTSD development, but studies on the potential sex differences of this neurobiological mechanism and how it relates to PTSD severity and progression are sparse. Here we examined sex differences in neural activation during response inhibition and PTSD following recent trauma.

METHODS

Participants (N= 205, 138 female sex assigned at birth) were recruited from emergency departments within 72 hours of a traumatic event. PTSD symptoms were assessed 2-weeks and 6-months post-trauma. A Go/NoGo task was performed 2-weeks post-trauma in a 3T MRI scanner to measure neural activity during response inhibition in the ventromedial prefrontal cortex (vmPFC), right inferior frontal gyrus (rIFG), and the bilateral hippocampus. General Linear models were used to examine the interaction effect of sex on the relationship between our regions of interest (ROIs) and the whole brain, and PTSD symptoms at 6-months, and symptom progression between 2-weeks and 6-months.

RESULTS

Lower response-inhibition-related vmPFC activation 2-weeks post-trauma predicted more PTSD symptoms at 6-months in females but not in males, while greater response-inhibition-related rIFG activation predicted lower PTSD symptom progression in males but not females. Whole brain interaction effects were observed in the medial temporal gyrus and left precentral gyrus.

CONCLUSIONS

There are sex differences in the relationship between inhibition-related brain activation and PTSD symptom severity and progression. These findings suggest that sex differences should be assessed in future PTSD studies and reveal potential targets for sex-specific interventions.

Post-traumatic stress and future substance use outcomes: leveraging antecedent factors to stratify risk

Background

Post-traumatic stress disorder (PTSD) and substance use (tobacco, alcohol, and cannabis) are highly comorbid. Many factors affect this relationship, including sociodemographic and psychosocial characteristics, other prior traumas, and physical health. However, few prior studies have investigated this prospectively, examining new substance use and the extent to which a wide range of factors may modify the relationship to PTSD.

Methods

The Advancing Understanding of RecOvery afteR traumA (AURORA) study is a prospective cohort of adults presenting at emergency departments (N = 2,943). Participants self-reported PTSD symptoms and the frequency and quantity of tobacco, alcohol, and cannabis use at six total timepoints. We assessed the associations of PTSD and future substance use, lagged by one timepoint, using the Poisson generalized estimating equations. We also stratified by incident and prevalent substance use and generated causal forests to identify the most important effect modifiers of this relationship out of 128 potential variables.

Results

At baseline, 37.3% (N = 1,099) of participants reported likely PTSD. PTSD was associated with tobacco frequency (incidence rate ratio (IRR): 1.003, 95% CI: 1.00, 1.01, p = 0.02) and quantity (IRR: 1.01, 95% CI: 1.001, 1.01, p = 0.01), and alcohol frequency (IRR: 1.002, 95% CI: 1.00, 1.004, p = 0.03) and quantity (IRR: 1.003, 95% CI: 1.001, 1.01, p = 0.001), but not with cannabis use. There were slight differences in incident compared to prevalent tobacco frequency and quantity of use; prevalent tobacco frequency and quantity were associated with PTSD symptoms, while incident tobacco frequency and quantity were not. Using causal forests, lifetime worst use of cigarettes, overall self-rated physical health, and prior childhood trauma were major moderators of the relationship between PTSD symptoms and the three substances investigated.

Conclusion

PTSD symptoms were highly associated with tobacco and alcohol use, while the association with prospective cannabis use is not clear. Findings suggest that understanding the different risk stratification that occurs can aid in tailoring interventions to populations at greatest risk to best mitigate the comorbidity between PTSD symptoms and future substance use outcomes. We demonstrate that this is particularly salient for tobacco use and, to some extent, alcohol use, while cannabis is less likely to be impacted by PTSD symptoms across the strata.

Multimodal Mental Health Digital Biomarker Analysis From Remote Interviews Using Facial, Vocal, Linguistic, and Cardiovascular Patterns

OBJECTIVE

Psychiatric evaluation suffers from subjectivity and bias, and is hard to scale due to intensive professional training requirements. In this work, we investigated whether behavioral and physiological signals, extracted from tele-video interviews, differ in individuals with psychiatric disorders.

METHODS

Temporal variations in facial expression, vocal expression, linguistic expression, and cardiovascular modulation were extracted from simultaneously recorded audio and video of remote interviews. Averages, standard deviations, and Markovian process-derived statistics of these features were computed from 73 subjects. Four binary classification tasks were defined: detecting 1) any clinically-diagnosed psychiatric disorder, 2) major depressive disorder, 3) self-rated depression, and 4) self-rated anxiety. Each modality was evaluated individually and in combination.

RESULTS

Statistically significant feature differences were found between psychiatric and control subjects. Correlations were found between features and self-rated depression and anxiety scores. Heart rate dynamics provided the best unimodal performance with areas under the receiver-operator curve (AUROCs) of 0.68-0.75 (depending on the classification task). Combining multiple modalities provided AUROCs of 0.72-0.82.

CONCLUSION

Multimodal features extracted from remote interviews revealed informative characteristics of clinically diagnosed and self-rated mental health status.

SIGNIFICANCE

The proposed multimodal approach has the potential to facilitate scalable, remote, and low-cost assessment for low-burden automated mental health services.

Using a standalone ear-EEG device for focal-onset seizure detection

BACKGROUND

Seizure detection is challenging outside the clinical environment due to the lack of comfortable, reliable, and practical long-term neurophysiological monitoring devices. We developed a novel, discreet, unobstructive in-ear sensing system that enables long-term electroencephalography (EEG) recording. This is the first study we are aware of that systematically compares the seizure detection utility of in-ear EEG with that of simultaneously recorded intracranial EEG. In addition, we present a similar comparison between simultaneously recorded in-ear EEG and scalp EEG.

METHODS

In this foundational research, we conducted a clinical feasibility study and validated the ability of the ear-EEG system to capture focal-onset seizures against 1255 hrs of simultaneous ear-EEG data along with scalp or intracranial EEG in 20 patients with refractory focal epilepsy (11 with scalp EEG, 8 with intracranial EEG, and 1 with both).

RESULTS

In a blinded, independent review of the ear-EEG signals, two epileptologists were able to detect 86.4% of the seizures that were subsequently identified using the clinical gold standard EEG modalities, with a false detection rate of 0.1 per day, well below what has been reported for ambulatory monitoring. The few seizures not detected on the ear-EEG signals emanated from deep within the mesial temporal lobe or extra-temporally and remained very focal, without significant propagation. Following multiple sessions of recording for a median continuous wear time of 13 hrs, patients reported a high degree of tolerance for the device, with only minor adverse events reported by the scalp EEG cohort.

CONCLUSIONS

These preliminary results demonstrate the potential of using ear-EEG to enable routine collection of complementary, prolonged, and remote neurophysiological evidence, which may permit real-time detection of paroxysmal events such as seizures and epileptiform discharges. This study suggests that the ear-EEG device may assist clinicians in making an epilepsy diagnosis, assessing treatment efficacy, and optimizing medication titration.

Predicting Adverse Behavior in Individuals with Autism Spectrum Disorder Through Off-body Sleep Analysis

Poor sleep quality in Autism Spectrum Disorder (ASD) individuals is linked to severe daytime behaviors. This study explores the relationship between a prior night's sleep structure and its predictive power for next-day behavior in ASD individuals. The motion was extracted using a low-cost near-infrared camera in a privacy-preserving way. Over two years, we recorded overnight data from 14 individuals, spanning over 2,000 nights, and tracked challenging daytime behaviors, including aggression, self-injury, and disruption. We developed an ensemble machine learning algorithm to predict next-day behavior in the morning and the afternoon. Our findings indicate that sleep quality is a more reliable predictor of morning behavior than afternoon behavior the next day. The proposed model attained an accuracy of 74% and a F1 score of 0.74 in target-sensitive tasks and 67% accuracy and 0.69 F1 score in target-insensitive tasks. For 7 of the 14, better-than-chance balanced accuracy was obtained (p-value<0.05), with 3 showing significant trends (p-value<0.1). These results suggest off-body, privacy-preserving sleep monitoring as a viable method for predicting next-day adverse behavior in ASD individuals, with the potential for behavioral intervention and enhanced care in social and learning settings.

Indoor Group Identification and Localization Using Privacy-Preserving Edge Computing Distributed Camera Network

Social interaction behaviors change as a result of both physical and psychiatric problems, and it is important to identify subtle changes in group activity engagements for monitoring the mental health of patients in clinics. This work proposes a system to identify when and where group formations occur in an approximately 1700 m2 therapeutic built environment using a distributed edge-computing camera network. The proposed method can localize group formations when provided with noisy positions and orientations of individuals, estimated from sparsely distributed multiview cameras, which run a lightweight multiperson 2-D pose detection model. Our group identification method demonstrated an F1 score of up to 90% with a mean absolute error of 1.25 m for group localization on our benchmark dataset. The dataset consisted of seven subjects walking, sitting, and conversing for 35 min in groups of various sizes ranging from 2 to 7 subjects. The proposed system is low-cost and scalable to any ordinary building to transform the indoor space into a smart environment using edge computing systems. We expect the proposed system to enhance existing therapeutic units for passively monitoring the social behaviors of patients when implementing real-time interventions.

A Data-Driven Gaussian Process Filter for Electrocardiogram Denoising

OBJECTIVE

Gaussian Processes (GP)-based filters, which have been effectively used for various applications including electrocardiogram (ECG) filtering can be computationally demanding and the choice of their hyperparameters is typically ad hoc.

METHODS

We develop a data-driven GP filter to address both issues, using the notion of the ECG phase domain -- a time-warped representation of the ECG beats onto a fixed number of samples and aligned R-peaks, which is assumed to follow a Gaussian distribution. Under this assumption, the computation of the sample mean and covariance matrix is simplified, enabling an efficient implementation of the GP filter in a data-driven manner, with no ad hoc hyperparameters. The proposed filter is evaluated and compared with a state-of-the-art wavelet-based filter, on the PhysioNet QT Database. The performance is evaluated by measuring the signal-to-noise ratio (SNR) improvement of the filter at SNR levels ranging from -5 to 30dB, in 5dB steps, using additive noise. For a clinical evaluation, the error between the estimated QT-intervals of the original and filtered signals is measured and compared with the benchmark filter.

RESULTS

It is shown that the proposed GP filter outperforms the benchmark filter for all the tested noise levels. It also outperforms the state-of-the-art filter in terms of QT-interval estimation error bias and variance.

CONCLUSION

The proposed GP filter is a versatile technique for preprocessing the ECG in clinical and research applications, is applicable to ECG of arbitrary lengths and sampling frequencies, and provides confidence intervals for its performance.

Associations of alcohol and cannabis use with change in posttraumatic stress disorder and depression symptoms over time in recently trauma-exposed individuals

BACKGROUND

Several hypotheses may explain the association between substance use, posttraumatic stress disorder (PTSD), and depression. However, few studies have utilized a large multisite dataset to understand this complex relationship. Our study assessed the relationship between alcohol and cannabis use trajectories and PTSD and depression symptoms across 3 months in recently trauma-exposed civilians.

METHODS

In total, 1618 (1037 female) participants provided self-report data on past 30-day alcohol and cannabis use and PTSD and depression symptoms during their emergency department (baseline) visit. We reassessed participant's substance use and clinical symptoms 2, 8, and 12 weeks posttrauma. Latent class mixture modeling determined alcohol and cannabis use trajectories in the sample. Changes in PTSD and depression symptoms were assessed across alcohol and cannabis use trajectories via a mixed-model repeated-measures analysis of variance.

RESULTS

Three trajectory classes (low, high, increasing use) provided the best model fit for alcohol and cannabis use. The low alcohol use class exhibited lower PTSD symptoms at baseline than the high use class; the low cannabis use class exhibited lower PTSD and depression symptoms at baseline than the high and increasing use classes; these symptoms greatly increased at week 8 and declined at week 12. Participants who already use alcohol and cannabis exhibited greater PTSD and depression symptoms at baseline that increased at week 8 with a decrease in symptoms at week 12.

CONCLUSIONS

Our findings suggest that alcohol and cannabis use trajectories are associated with the intensity of posttrauma psychopathology. These findings could potentially inform the timing of therapeutic strategies.

The International Cardiac Arrest Research Consortium Electroencephalography Database

OBJECTIVES

To develop the International Cardiac Arrest Research (I-CARE), a harmonized multicenter clinical and electroencephalography database for acute hypoxic-ischemic brain injury research involving patients with cardiac arrest.

DESIGN

Multicenter cohort, partly prospective and partly retrospective.

SETTING

Seven academic or teaching hospitals from the United States and Europe.

PATIENTS

Individuals 16 years old or older who were comatose after return of spontaneous circulation following a cardiac arrest who had continuous electroencephalography monitoring were included.

INTERVENTIONS

Not applicable.

MEASUREMENTS AND MAIN RESULTS

Clinical and electroencephalography data were harmonized and stored in a common Waveform Database-compatible format. Automated spike frequency, background continuity, and artifact detection on electroencephalography were calculated with 10-second resolution and summarized hourly. Neurologic outcome was determined at 3-6 months using the best Cerebral Performance Category (CPC) scale. This database includes clinical data and 56,676 hours (3.9 terabytes) of continuous electroencephalography data for 1,020 patients. Most patients died ( n = 603, 59%), 48 (5%) had severe neurologic disability (CPC 3 or 4), and 369 (36%) had good functional recovery (CPC 1-2). There is significant variability in mean electroencephalography recording duration depending on the neurologic outcome (range, 53-102 hr for CPC 1 and CPC 4, respectively). Epileptiform activity averaging 1 Hz or more in frequency for at least 1 hour was seen in 258 patients (25%) (19% for CPC 1-2 and 29% for CPC 3-5). Burst suppression was observed for at least 1 hour in 207 (56%) and 635 (97%) patients with CPC 1-2 and CPC 3-5, respectively.

CONCLUSIONS

The I-CARE consortium electroencephalography database provides a comprehensive real-world clinical and electroencephalography dataset for neurophysiology research of comatose patients after cardiac arrest. This dataset covers the spectrum of abnormal electroencephalography patterns after cardiac arrest, including epileptiform patterns and those in the ictal-interictal continuum.

Internal capsule microstructure mediates the relationship between childhood maltreatment and PTSD following adulthood trauma exposure

Childhood trauma is a known risk factor for trauma and stress-related disorders in adulthood. However, limited research has investigated the impact of childhood trauma on brain structure linked to later posttraumatic dysfunction. We investigated the effect of childhood trauma on white matter microstructure after recent trauma and its relationship with future posttraumatic dysfunction among trauma-exposed adult participants (n = 202) recruited from emergency departments as part of the AURORA Study. Participants completed self-report scales assessing prior childhood maltreatment within 2-weeks in addition to assessments of PTSD, depression, anxiety, and dissociation symptoms within 6-months of their traumatic event. Fractional anisotropy (FA) obtained from diffusion tensor imaging (DTI) collected at 2-weeks and 6-months was used to index white matter microstructure. Childhood maltreatment load predicted 6-month PTSD symptoms (b = 1.75, SE = 0.78, 95% CI = [0.20, 3.29]) and inversely varied with FA in the bilateral internal capsule (IC) at 2-weeks (p = 0.0294, FDR corrected) and 6-months (p = 0.0238, FDR corrected). We observed a significant indirect effect of childhood maltreatment load on 6-month PTSD symptoms through 2-week IC microstructure (b = 0.37, Boot SE = 0.18, 95% CI = [0.05, 0.76]) that fully mediated the effect of childhood maltreatment load on PCL-5 scores (b = 1.37, SE = 0.79, 95% CI = [-0.18, 2.93]). IC microstructure did not mediate relationships between childhood maltreatment and depressive, anxiety, or dissociative symptomatology. Our findings suggest a unique role for IC microstructure as a stable neural pathway between childhood trauma and future PTSD symptoms following recent trauma. Notably, our work did not support roles of white matter tracts previously found to vary with PTSD symptoms and childhood trauma exposure, including the cingulum bundle, uncinate fasciculus, and corpus callosum. Given the IC contains sensory fibers linked to perception and motor control, childhood maltreatment might impact the neural circuits that relay and process threat-related inputs and responses to trauma.

Graph Trilateration for Indoor Localization in Sparsely Distributed Edge Computing Devices in Complex Environments Using Bluetooth Technology

Spatial navigation patterns in indoor space usage can reveal important cues about the cognitive health of participants. In this work, we present a low-cost, scalable, open-source edge computing system using Bluetooth low energy (BLE) beacons for tracking indoor movements in a large, 1700 m2 facility used to carry out therapeutic activities for participants with mild cognitive impairment (MCI). The facility is instrumented with 39 edge computing systems, along with an on-premise fog server. The participants carry a BLE beacon, in which BLE signals are received and analyzed by the edge computing systems. Edge computing systems are sparsely distributed in the wide, complex indoor space, challenging the standard trilateration technique for localizing subjects, which assumes a dense installation of BLE beacons. We propose a graph trilateration approach that considers the temporal density of hits from the BLE beacon to surrounding edge devices to handle the inconsistent coverage of edge devices. This proposed method helps us tackle the varying signal strength, which leads to intermittent detection of beacons. The proposed method can pinpoint the positions of multiple participants with an average error of 4.4 m and over 85% accuracy in region-level localization across the entire study area. Our experimental results, evaluated in a clinical environment, suggest that an ordinary medical facility can be transformed into a smart space that enables automatic assessment of individuals' movements, which may reflect health status or response to treatment.

The 2023 wearable photoplethysmography roadmap

Photoplethysmography is a key sensing technology which is used in wearable devices such as smartwatches and fitness trackers. Currently, photoplethysmography sensors are used to monitor physiological parameters including heart rate and heart rhythm, and to track activities like sleep and exercise. Yet, wearable photoplethysmography has potential to provide much more information on health and wellbeing, which could inform clinical decision making. This Roadmap outlines directions for research and development to realise the full potential of wearable photoplethysmography. Experts discuss key topics within the areas of sensor design, signal processing, clinical applications, and research directions. Their perspectives provide valuable guidance to researchers developing wearable photoplethysmography technology.

Association between microbiome and the development of adverse posttraumatic neuropsychiatric sequelae after traumatic stress exposure

Patients exposed to trauma often experience high rates of adverse post-traumatic neuropsychiatric sequelae (APNS). The biological mechanisms promoting APNS are currently unknown, but the microbiota-gut-brain axis offers an avenue to understanding mechanisms as well as possibilities for intervention. Microbiome composition after trauma exposure has been poorly examined regarding neuropsychiatric outcomes. We aimed to determine whether the gut microbiomes of trauma-exposed emergency department patients who develop APNS have dysfunctional gut microbiome profiles and discover potential associated mechanisms. We performed metagenomic analysis on stool samples (n = 51) from a subset of adults enrolled in the Advancing Understanding of RecOvery afteR traumA (AURORA) study. Two-, eight- and twelve-week post-trauma outcomes for post-traumatic stress disorder (PTSD) (PTSD checklist for DSM-5), normalized depression scores (PROMIS Depression Short Form 8b) and somatic symptom counts were collected. Generalized linear models were created for each outcome using microbial abundances and relevant demographics. Mixed-effect random forest machine learning models were used to identify associations between APNS outcomes and microbial features and encoded metabolic pathways from stool metagenomics. Microbial species, including Flavonifractor plautii, Ruminococcus gnavus and, Bifidobacterium species, which are prevalent commensal gut microbes, were found to be important in predicting worse APNS outcomes from microbial abundance data. Notably, through APNS outcome modeling using microbial metabolic pathways, worse APNS outcomes were highly predicted by decreased L-arginine related pathway genes and increased citrulline and ornithine pathways. Common commensal microbial species are enriched in individuals who develop APNS. More notably, we identified a biological mechanism through which the gut microbiome reduces global arginine bioavailability, a metabolic change that has also been demonstrated in the plasma of patients with PTSD.

Fairness as an afterthought: An American perspective on fairness in model developer-clinician user collaborations

Numerous ethics guidelines have been handed down over the last few years on the ethical applications of machine learning models. Virtually every one of them mentions the importance of "fairness" in the development and use of these models. Unfortunately, though, these ethics documents omit providing a consensually adopted definition or characterization of fairness. As one group of authors observed, these documents treat fairness as an "afterthought" whose importance is undeniable but whose essence seems strikingly elusive. In this essay, which offers a distinctly American treatment of "fairness," we comment on a number of fairness formulations and on qualitative or statistical methods that have been encouraged to achieve fairness. We argue that none of them, at least from an American moral perspective, provides a one-size-fits-all definition of or methodology for securing fairness that could inform or standardize fairness over the universe of use cases witnessing machine learning applications. Instead, we argue that because fairness comprehensions and applications reflect a vast range of use contexts, model developers and clinician users will need to engage in thoughtful collaborations that examine how fairness should be conceived and operationalized in the use case at issue. Part II of this paper illustrates key moments in these collaborations, especially when inter and intra disagreement occurs among model developer and clinician user groups over whether a model is fair or unfair. We conclude by noting that these collaborations will likely occur over the lifetime of a model if its claim to fairness is to advance beyond "afterthought" status.

Using HIPAA (Health Insurance Portability and Accountability Act)-Compliant Transcription Services for Virtual Psychiatric Interviews: Pilot Comparison Study

BACKGROUND

Automatic speech recognition (ASR) technology is increasingly being used for transcription in clinical contexts. Although there are numerous transcription services using ASR, few studies have compared the word error rate (WER) between different transcription services among different diagnostic groups in a mental health setting. There has also been little research into the types of words ASR transcriptions mistakenly generate or omit.

OBJECTIVE

This study compared the WER of 3 ASR transcription services (Amazon Transcribe [Amazon.com, Inc], Zoom-Otter AI [Zoom Video Communications, Inc], and Whisper [OpenAI Inc]) in interviews across 2 different clinical categories (controls and participants experiencing a variety of mental health conditions). These ASR transcription services were also compared with a commercial human transcription service, Rev (Rev.Com, Inc). Words that were either included or excluded by the error in the transcripts were systematically analyzed by their Linguistic Inquiry and Word Count categories.

METHODS

Participants completed a 1-time research psychiatric interview, which was recorded on a secure server. Transcriptions created by the research team were used as the gold standard from which WER was calculated. The interviewees were categorized into either the control group (n=18) or the mental health condition group (n=47) using the Mini-International Neuropsychiatric Interview. The total sample included 65 participants. Brunner-Munzel tests were used for comparing independent sets, such as the diagnostic groupings, and Wilcoxon signed rank tests were used for correlated samples when comparing the total sample between different transcription services.

RESULTS

There were significant differences between each ASR transcription service's WER (P<.001). Amazon Transcribe's output exhibited significantly lower WERs compared with the Zoom-Otter AI's and Whisper's ASR. ASR performances did not significantly differ across the 2 different clinical categories within each service (P>.05). A comparison between the human transcription service output from Rev and the best-performing ASR (Amazon Transcribe) demonstrated a significant difference (P<.001), with Rev having a slightly lower median WER (7.6%, IQR 5.4%-11.35 vs 8.9%, IQR 6.9%-11.6%). Heat maps and spider plots were used to visualize the most common errors in Linguistic Inquiry and Word Count categories, which were found to be within 3 overarching categories: Conversation, Cognition, and Function.

CONCLUSIONS

Overall, consistent with previous literature, our results suggest that the WER between manual and automated transcription services may be narrowing as ASR services advance. These advances, coupled with decreased cost and time in receiving transcriptions, may make ASR transcriptions a more viable option within health care settings. However, more research is required to determine if errors in specific types of words impact the analysis and usability of these transcriptions, particularly for specific applications and in a variety of populations in terms of clinical diagnosis, literacy level, accent, and cultural origin.

Classifying Tremor Dominant and Postural Instability and Gait Difficulty Subtypes of Parkinson's Disease from Full-Body Kinematics

Characterizing motor subtypes of Parkinson's disease (PD) is an important aspect of clinical care that is useful for prognosis and medical management. Although all PD cases involve the loss of dopaminergic neurons in the brain, individual cases may present with different combinations of motor signs, which may indicate differences in underlying pathology and potential response to treatment. However, the conventional method for distinguishing PD motor subtypes involves resource-intensive physical examination by a movement disorders specialist. Moreover, the standardized rating scales for PD rely on subjective observation, which requires specialized training and unavoidable inter-rater variability. In this work, we propose a system that uses machine learning models to automatically and objectively identify some PD motor subtypes, specifically Tremor-Dominant (TD) and Postural Instability and Gait Difficulty (PIGD), from 3D kinematic data recorded during walking tasks for patients with PD (MDS-UPDRS-III Score, 34.7 ± 10.5, average disease duration 7.5 ± 4.5 years). This study demonstrates a machine learning model utilizing kinematic data that identifies PD motor subtypes with a 79.6% F1 score (N = 55 patients with parkinsonism). This significantly outperformed a comparison model using classification based on gait features (19.8% F1 score). Variants of our model trained to individual patients achieved a 95.4% F1 score. This analysis revealed that both temporal, spectral, and statistical features from lower body movements are helpful in distinguishing motor subtypes. Automatically assessing PD motor subtypes simply from walking may reduce the time and resources required from specialists, thereby improving patient care for PD treatments. Furthermore, this system can provide objective assessments to track the changes in PD motor subtypes over time to implement and modify appropriate treatment plans for individual patients as needed.

Prior Sexual Trauma Exposure Impacts Posttraumatic Dysfunction and Neural Circuitry Following a Recent Traumatic Event in the AURORA Study

Background

Prior sexual trauma (ST) is associated with greater risk for posttraumatic stress disorder after a subsequent traumatic event; however, the underlying neurobiological mechanisms remain opaque. We investigated longitudinal posttraumatic dysfunction and amygdala functional dynamics following admission to an emergency department for new primarily nonsexual trauma in participants with and without previous ST.

Methods

Participants (N = 2178) were recruited following acute trauma exposure (primarily motor vehicle collision). A subset (n = 242) completed magnetic resonance imaging that included a fearful faces task and a resting-state scan 2 weeks after the trauma. We investigated associations between prior ST and several dimensions of posttraumatic symptoms over 6 months. We further assessed amygdala activation and connectivity differences between groups with or without prior ST.

Results

Prior ST was associated with greater posttraumatic depression (F1,1120 = 28.35, p = 1.22 × 10-7, ηp2 = 0.06), anxiety (F1,1113 = 17.43, p = 3.21 × 10-5, ηp2 = 0.05), and posttraumatic stress disorder (F1,1027 = 11.34, p = 7.85 × 10-4, ηp2 = 0.04) severity and more maladaptive beliefs about pain (F1,1113 = 8.51, p = .004, ηp2 = 0.02) but was not related to amygdala reactivity to fearful versus neutral faces (all ps > .05). A secondary analysis revealed an interaction between ST and lifetime trauma load on the left amygdala to visual cortex connectivity (peak Z value: -4.41, corrected p < .02).

Conclusions

Findings suggest that prior ST is associated with heightened posttraumatic dysfunction following a new trauma exposure but not increased amygdala activity. In addition, ST may interact with lifetime trauma load to alter neural circuitry in visual processing regions following acute trauma exposure. Further research should probe the relationship between trauma type and visual circuitry in the acute aftermath of trauma.

Multimodal mental health assessment with remote interviews using facial, vocal, linguistic, and cardiovascular patterns

Objective

The current clinical practice of psychiatric evaluation suffers from subjectivity and bias, and requires highly skilled professionals that are often unavailable or unaffordable. Objective digital biomarkers have shown the potential to address these issues. In this work, we investigated whether behavioral and physiological signals, extracted from remote interviews, provided complimentary information for assessing psychiatric disorders.

Methods

Time series of multimodal features were derived from four conceptual modes: facial expression, vocal expression, linguistic expression, and cardiovascular modulation. The features were extracted from simultaneously recorded audio and video of remote interviews using task-specific and foundation models. Averages, standard deviations, and hidden Markov model-derived statistics of these features were computed from 73 subjects. Four binary classification tasks were defined: detecting 1) any clinically-diagnosed psychiatric disorder, 2) major depressive disorder, 3) self-rated depression, and 4) self-rated anxiety. Each modality was evaluated individually and in combination.

Results

Statistically significant feature differences were found between controls and subjects with mental health conditions. Correlations were found between features and self-rated depression and anxiety scores. Visual heart rate dynamics achieved the best unimodal performance with areas under the receiver-operator curve (AUROCs) of 0.68-0.75 (depending on the classification task). Combining multiple modalities achieved AUROCs of 0.72-0.82. Features from task-specific models outperformed features from foundation models.

Conclusion

Multimodal features extracted from remote interviews revealed informative characteristics of clinically diagnosed and self-rated mental health status.

Significance

The proposed multimodal approach has the potential to facilitate objective, remote, and low-cost assessment for low-burden automated mental health services.

Neighborhood Disadvantage and Neural Correlates of Threat and Reward Processing in Survivors of Recent Trauma

Importance

Differences in neighborhood socioeconomic characteristics are important considerations in understanding differences in risk vs resilience in mental health. Neighborhood disadvantage is associated with alterations in the function and structure of threat neurocircuitry.

Objective

To investigate associations of neighborhood disadvantage with white and gray matter and neural reactivity to positive and negative stimuli in the context of trauma exposure.

Design, Setting, and Participants

In this cross-sectional study, survivors of trauma who completed sociodemographic and posttraumatic symptom assessments and neuroimaging were recruited as part of the Advancing Understanding of Recovery After Trauma (AURORA) study between September 2017 and June 2021. Data analysis was performed from October 25, 2022, to February 15, 2023.

Exposure

Neighborhood disadvantage was measured with the Area Deprivation Index (ADI) for each participant home address.

Main Outcomes and Measures

Participants completed separate threat and reward tasks during functional magnetic resonance imaging. Diffusion-weighted and high-resolution structural images were also collected. Linear models assessed the association of ADI with reactivity, microstructure, and macrostructure of a priori regions of interest after adjusting for income, lifetime trauma, sex at birth, and age. A moderated-mediation model tested whether ADI was associated with neural activity via microstructural changes and if this was modulated by PTSD symptoms.

Results

A total of 280 participants (183 females [65.4%]; mean [SD] age, 35.39 [13.29] years) completed the threat task and 244 participants (156 females [63.9%]; mean [SD] age, 35.10 [13.26] years) completed the reward task. Higher ADI (per 1-unit increase) was associated with greater insula (t274 = 3.20; β = 0.20; corrected P = .008) and anterior cingulate cortex (ACC; t274 = 2.56; β = 0.16; corrected P = .04) threat-related activity after considering covariates, but ADI was not associated with reward reactivity. Greater disadvantage was also associated with altered microstructure of the cingulum bundle (t274 = 3.48; β = 0.21; corrected P = .001) and gray matter morphology of the ACC (cortical thickness: t273 = -2.29; β = -0.13; corrected P = .02; surface area: t273 = 2.53; β = 0.13; corrected P = .02). The moderated-mediation model revealed that ADI was associated with ACC threat reactivity via cingulum microstructural changes (index of moderated mediation = -0.02). However, this mediation was only present in individuals with greater PTSD symptom severity (at the mean: β = -0.17; standard error = 0.06, t= -2.28; P = .007; at 1 SD above the mean: β = -0.28; standard error = 0.08; t = -3.35; P < .001).

Conclusions and Relevance

In this study, neighborhood disadvantage was associated with neurobiology that supports threat processing, revealing associations of neighborhood disadvantage with neural susceptibility for PTSD and suggesting how altered structure-function associations may complicate symptoms. Future work should investigate specific components of neighborhood disadvantage that may be associated with these outcomes.

Heart murmur detection from phonocardiogram recordings: The George B. Moody PhysioNet Challenge 2022

Cardiac auscultation is an accessible diagnostic screening tool that can help to identify patients with heart murmurs, who may need follow-up diagnostic screening and treatment for abnormal cardiac function. However, experts are needed to interpret the heart sounds, limiting the accessibility of cardiac auscultation in resource-constrained environments. Therefore, the George B. Moody PhysioNet Challenge 2022 invited teams to develop algorithmic approaches for detecting heart murmurs and abnormal cardiac function from phonocardiogram (PCG) recordings of heart sounds. For the Challenge, we sourced 5272 PCG recordings from 1452 primarily pediatric patients in rural Brazil, and we invited teams to implement diagnostic screening algorithms for detecting heart murmurs and abnormal cardiac function from the recordings. We required the participants to submit the complete training and inference code for their algorithms, improving the transparency, reproducibility, and utility of their work. We also devised an evaluation metric that considered the costs of screening, diagnosis, misdiagnosis, and treatment, allowing us to investigate the benefits of algorithmic diagnostic screening and facilitate the development of more clinically relevant algorithms. We received 779 algorithms from 87 teams during the Challenge, resulting in 53 working codebases for detecting heart murmurs and abnormal cardiac function from PCG recordings. These algorithms represent a diversity of approaches from both academia and industry, including methods that use more traditional machine learning techniques with engineered clinical and statistical features as well as methods that rely primarily on deep learning models to discover informative features. The use of heart sound recordings for identifying heart murmurs and abnormal cardiac function allowed us to explore the potential of algorithmic approaches for providing more accessible diagnostic screening in resource-constrained environments. The submission of working, open-source algorithms and the use of novel evaluation metrics supported the reproducibility, generalizability, and clinical relevance of the research from the Challenge.

The International Cardiac Arrest Research (I-CARE) Consortium Electroencephalography Database

Objective

To develop a harmonized multicenter clinical and electroencephalography (EEG) database for acute hypoxic-ischemic brain injury research involving patients with cardiac arrest.

Design

Multicenter cohort, partly prospective and partly retrospective.

Setting

Seven academic or teaching hospitals from the U.S. and Europe.

Patients

Individuals aged 16 or older who were comatose after return of spontaneous circulation following a cardiac arrest who had continuous EEG monitoring were included.

Interventions

not applicable.

Measurements and Main Results

Clinical and EEG data were harmonized and stored in a common Waveform Database (WFDB)-compatible format. Automated spike frequency, background continuity, and artifact detection on EEG were calculated with 10 second resolution and summarized hourly. Neurological outcome was determined at 3-6 months using the best Cerebral Performance Category (CPC) scale. This database includes clinical and 56,676 hours (3.9 TB) of continuous EEG data for 1,020 patients. Most patients died (N=603, 59%), 48 (5%) had severe neurological disability (CPC 3 or 4), and 369 (36%) had good functional recovery (CPC 1-2). There is significant variability in mean EEG recording duration depending on the neurological outcome (range 53-102h for CPC 1 and CPC 4, respectively). Epileptiform activity averaging 1 Hz or more in frequency for at least one hour was seen in 258 (25%) patients (19% for CPC 1-2 and 29% for CPC 3-5). Burst suppression was observed for at least one hour in 207 (56%) and 635 (97%) patients with CPC 1-2 and CPC 3-5, respectively.

Conclusions

The International Cardiac Arrest Research (I-CARE) consortium database provides a comprehensive real-world clinical and EEG dataset for neurophysiology research of comatose patients after cardiac arrest. This dataset covers the spectrum of abnormal EEG patterns after cardiac arrest, including epileptiform patterns and those in the ictal-interictal continuum.

Beyond Heart Murmur Detection: Automatic Murmur Grading From Phonocardiogram

OBJECTIVE

Murmurs are abnormal heart sounds, identified by experts through cardiac auscultation. The murmur grade, a quantitative measure of the murmur intensity, is strongly correlated with the patient's clinical condition. This work aims to estimate each patient's murmur grade (i.e., absent, soft, loud) from multiple auscultation location phonocardiograms (PCGs) of a large population of pediatric patients from a low-resource rural area.

METHODS

The Mel spectrogram representation of each PCG recording is given to an ensemble of 15 convolutional residual neural networks with channel-wise attention mechanisms to classify each PCG recording. The final murmur grade for each patient is derived based on the proposed decision rule and considering all estimated labels for available recordings. The proposed method is cross-validated on a dataset consisting of 3456 PCG recordings from 1007 patients using a stratified ten-fold cross-validation. Additionally, the method was tested on a hidden test set comprised of 1538 PCG recordings from 442 patients.

RESULTS

The overall cross-validation performances for patient-level murmur gradings are 86.3% and 81.6% in terms of the unweighted average of sensitivities and F1-scores, respectively. The sensitivities (and F1-scores) for absent, soft, and loud murmurs are 90.7% (93.6%), 75.8% (66.8%), and 92.3% (84.2%), respectively. On the test set, the algorithm achieves an unweighted average of sensitivities of 80.4% and an F1-score of 75.8%.

CONCLUSIONS

This study provides a potential approach for algorithmic pre-screening in low-resource settings with relatively high expert screening costs.

SIGNIFICANCE

The proposed method represents a significant step beyond detection of murmurs, providing characterization of intensity, which may provide an enhanced classification of clinical outcomes.

Combining crowd-sourcing, census data, and public review forums for real-time, high-resolution food desert estimation

BACKGROUND

It has been hypothesized that low access to healthy and nutritious food increases health disparities. Low-accessibility areas, called food deserts, are particularly commonplace in lower-income neighborhoods. The metrics for measuring the food environment's health, called food desert indices, are primarily based on decadal census data, limiting their frequency and geographical resolution to that of the census. We aimed to create a food desert index with finer geographic resolution than census data and better responsiveness to environmental changes.

MATERIALS AND METHODS

We augmented decadal census data with real-time data from platforms such as Yelp and Google Maps and crowd-sourced answers to questionnaires by the Amazon Mechanical Turks to create a real-time, context-aware, and geographically refined food desert index. Finally, we used this refined index in a concept application that suggests alternative routes with similar ETAs between a source and destination in the Atlanta metropolitan area as an intervention to expose a traveler to better food environments.

RESULTS

We made 139,000 pull requests to Yelp, analyzing 15,000 unique food retailers in the metro Atlanta area. In addition, we performed 248,000 walking and driving route analyses on these retailers using Google Maps' API. As a result, we discovered that the metro Atlanta food environment creates a strong bias towards eating out rather than preparing a meal at home when access to vehicles is limited. Contrary to the food desert index that we started with, which changed values only at neighborhood boundaries, the food desert index that we built on top of it captured the changing exposure of a subject as they walked or drove through the city. This model was also sensitive to the changes in the environment that occurred after the census data was collected.

CONCLUSIONS

Research on the environmental components of health disparities is flourishing. New machine learning models have the potential to augment various information sources and create fine-tuned models of the environment. This opens the way to better understanding the environment and its effects on health and suggesting better interventions.

Structural inequities contribute to racial/ethnic differences in neurophysiological tone, but not threat reactivity, after trauma exposure

Considerable racial/ethnic disparities persist in exposure to life stressors and socioeconomic resources that can directly affect threat neurocircuitry, particularly the amygdala, that partially mediates susceptibility to adverse posttraumatic outcomes. Limited work to date, however, has investigated potential racial/ethnic variability in amygdala reactivity or connectivity that may in turn be related to outcomes such as post-traumatic stress disorder (PTSD). Participants from the AURORA study (n = 283), a multisite longitudinal study of trauma outcomes, completed functional magnetic resonance imaging and psychophysiology within approximately two-weeks of trauma exposure. Seed-based amygdala connectivity and amygdala reactivity during passive viewing of fearful and neutral faces were assessed during fMRI. Physiological activity was assessed during Pavlovian threat conditioning. Participants also reported the severity of posttraumatic symptoms 3 and 6 months after trauma. Black individuals showed lower baseline skin conductance levels and startle compared to White individuals, but no differences were observed in physiological reactions to threat. Further, Hispanic and Black participants showed greater amygdala connectivity to regions including the dorsolateral prefrontal cortex (PFC), dorsal anterior cingulate cortex, insula, and cerebellum compared to White participants. No differences were observed in amygdala reactivity to threat. Amygdala connectivity was associated with 3-month PTSD symptoms, but the associations differed by racial/ethnic group and were partly driven by group differences in structural inequities. The present findings suggest variability in tonic neurophysiological arousal in the early aftermath of trauma between racial/ethnic groups, driven by structural inequality, impacts neural processes that mediate susceptibility to later PTSD symptoms.

ProductGraphSleepNet: Sleep staging using product spatio-temporal graph learning with attentive temporal aggregation

The classification of sleep stages plays a crucial role in understanding and diagnosing sleep pathophysiology. Sleep stage scoring relies heavily on visual inspection by an expert, which is a time-consuming and subjective procedure. Recently, deep learning neural network approaches have been leveraged to develop a generalized automated sleep staging and account for shifts in distributions that may be caused by inherent inter/intra-subject variability, heterogeneity across datasets, and different recording environments. However, these networks (mostly) ignore the connections among brain regions and disregard modeling the connections between temporally adjacent sleep epochs. To address these issues, this work proposes an adaptive product graph learning-based graph convolutional network, named ProductGraphSleepNet, for learning joint spatio-temporal graphs along with a bidirectional gated recurrent unit and a modified graph attention network to capture the attentive dynamics of sleep stage transitions. Evaluation on two public databases: the Montreal Archive of Sleep Studies (MASS) SS3; and the SleepEDF, which contain full night polysomnography recordings of 62 and 20 healthy subjects, respectively, demonstrates performance comparable to the state-of-the-art (Accuracy: 0.867;0.838, F1-score: 0.818;0.774 and Kappa: 0.802;0.775, on each database respectively). More importantly, the proposed network makes it possible for clinicians to comprehend and interpret the learned spatial and temporal connectivity graphs for sleep stages.

Disentangling visual exploration differences in cognitive impairment

Objective

Compared to individuals without cognitive impairment (CI), those with CI exhibit differences in both basic oculomotor functions and complex viewing behaviors. However, the characteristics of the differences and how those differences relate to various cognitive functions have not been widely explored. In this work we aimed to quantify those differences and assess general cognitive impairment and specific cognitive functions.

Methods

A validated passive viewing memory test with eyetracking was administered to 348 healthy controls and CI individuals. Spatial, temporal, semantic, and other composite features were extracted from the estimated eye-gaze locations on the corresponding pictures displayed during the test. These features were then used to characterize viewing patterns, classify cognitive impairment, and estimate scores in various neuropsychological tests using machine learning.

Results

Statistically significant differences in spatial, spatiotemporal, and semantic features were found between healthy controls and individuals with CI. CI group spent more time gazing at the center of the image, looked at more regions of interest (ROI), transitioned less often between ROI yet in a more unpredictable manner, and had different semantic preferences. A combination of these features achieved an area under the receiver-operator curve of 0.78 in differentiating CI individuals from controls. Statistically significant correlations were identified between actual and estimated MoCA scores and other neuropsychological tests.

Conclusion

Evaluating visual exploration behaviors provided quantitative and systematic evidence of differences in CI individuals, leading to an improved approach for passive cognitive impairment screening.

Significance

The proposed passive, accessible, and scalable approach could help with earlier detection and a better understanding of cognitive impairment.

Hierarchical Attentive Network for Gestational Age Estimation in Low-Resource Settings

Assessing fetal development is essential to the provision of healthcare for both mothers and fetuses. In low- and middle-income countries, conditions that increase the risk of fetal growth restriction (FGR) are often more prevalent. In these regions, barriers to accessing healthcare and social services exacerbate fetal maternal health problems. One of these barriers is the lack of affordable diagnostic technologies. To address this issue, this work introduces an end-to-end algorithm applied to a low-cost, hand-held Doppler ultrasound device for estimating gestational age (GA), and by inference, FGR. The Doppler ultrasound signals used in this study were collected from 226 pregnancies (45 low birth weight at delivery) between 5 and 9 months GA by lay midwives in highland Guatemala. We designed a hierarchical deep sequence learning model with an attention mechanism to learn the normative dynamics of fetal cardiac activity in different stages of development. This resulted in a state-of-the-art GA estimation performance, with an average error of 0.79 months. This is close to the theoretical minimum for the given quantization level of one month. The model was then tested on Doppler recordings of the fetuses with low birth weight and the estimated GA was shown to be lower than the GA calculated from last menstruation. Thus, this could be interpreted as a potential sign of developmental retardation (or FGR) associated with low birth weight, and referral and intervention may be necessary.

Integrating multimodal information in machine learning for classifying acute myocardial infarction

OBJECTIVE

Prompt identification and recognization of myocardial ischemia/infarction (MI) is the most important goal in the management of acute coronary syndrome (ACS). The 12-lead electrocardiogram (ECG) is widely used as the initial screening tool for patients with chest pain but its diagnostic accuracy remains limited. There is early evidence that machine learning (ML) algorithms applied to ECG waveforms can improve performance. Most studies are designed to classify MI from healthy controls and thus are limited due to the lack of consideration of ECG abnormalities from other cardiac conditions, leading to false positives. Moreover, clinical information beyond ECG has not yet been well leveraged in existing ML models.

APPROACH

The present study considered downstream clinical implementation scenarios in the initial model design by dichotomizing study recordings from a public large-scale ECG dataset into a MI class and a non-MI class with the inclusion of MI-confounding conditions. Two experiments were conducted to systematically investigate the impact of two important factors entrained in the modeling process, including the duration of ECG, and the value of multimodal information for model training. A novel multimodal deep learning architecture was proposed to learn joint features from both ECG and patient demographics.

MAIN RESULTS

The multimodal model achieved better performance than the ECG-only model, with a mean area under the receiver operating characteristic curve (AUROC) of 92.1% and a mean accuracy of 87.4%, which is on par with existing studies despite the increased task difficulty due to the new class definition. By investigation of model explainability, it revealed the contribution of patient information in model performance and clinical concordance of the model's attention with existing clinical insights.

SIGNIFICANCE

The findings in this study help guide the development of ML solutions for prompt MI detection and move the models one step closer to real-world clinical applications.

Association between PTSD and Impedance Cardiogram-based contractility metrics during trauma recall: A controlled twin study

Post-traumatic stress disorder (PTSD) is an independent risk factor for incident heart failure, but the underlying cardiac mechanisms remained elusive. Impedance cardiography (ICG), especially when measured during stress, can help understand the underlying psychophysiological pathways linking PTSD with heart failure. We investigated the association between PTSD and ICG-based contractility metrics (pre-ejection period (PEP) and Heather index (HI)) using a controlled twin study design with a laboratory-based traumatic reminder stressor. PTSD status was assessed using structured clinical interviews. We acquired synchronized electrocardiograms and ICG data while playing personalized-trauma scripts. Using linear mixed-effects models, we examined twins as individuals and within PTSD-discordant pairs. We studied 137 male veterans (48 pairs, 41 unpaired singles) from Vietnam War Era with a mean (standard deviation) age of 68.5(2.5) years. HI during trauma stress was lower in the PTSD vs. non-PTSD individuals (7.2 vs. 9.3 [ohm/s2 ], p = .003). PEP reactivity (trauma minus neutral) was also more negative in PTSD vs. non-PTSD individuals (-7.4 vs. -2.0 [ms], p = .009). The HI and PEP associations with PTSD persisted for adjusted models during trauma and reactivity, respectively. For within-pair analysis of eight PTSD-discordant twin pairs (out of 48 pairs), PTSD was associated with lower HI in neutral, trauma, and reactivity, whereas no association was found between PTSD and PEP. PTSD was associated with reduced HI and PEP, especially with trauma recall stress. This combination of increased sympathetic activation and decreased cardiac contractility combined may be concerning for increased heart failure risk after recurrent trauma re-experiencing in PTSD.

Derivation and Validation of a Brief Emergency Department-Based Prediction Tool for Posttraumatic Stress After Motor Vehicle Collision

STUDY OBJECTIVE

To derive and initially validate a brief bedside clinical decision support tool that identifies emergency department (ED) patients at high risk of substantial, persistent posttraumatic stress symptoms after a motor vehicle collision.

METHODS

Derivation (n=1,282, 19 ED sites) and validation (n=282, 11 separate ED sites) data were obtained from adults prospectively enrolled in the Advancing Understanding of RecOvery afteR traumA study who were discharged from the ED after motor vehicle collision-related trauma. The primary outcome was substantial posttraumatic stress symptoms at 3 months (Posttraumatic Stress Disorder Checklist for Diagnostic and Statistical Manual of Mental Disorders-5 ≥38). Logistic regression derivation models were evaluated for discriminative ability using the area under the curve and the accuracy of predicted risk probabilities (Brier score). Candidate posttraumatic stress predictors assessed in these models (n=265) spanned a range of sociodemographic, baseline health, peritraumatic, and mechanistic domains. The final model selection was based on performance and ease of administration.

RESULTS

Significant 3-month posttraumatic stress symptoms were common in the derivation (27%) and validation (26%) cohort. The area under the curve and Brier score of the final 8-question tool were 0.82 and 0.14 in the derivation cohort and 0.76 and 0.17 in the validation cohort.

CONCLUSION

This simple 8-question tool demonstrates promise to risk-stratify individuals with substantial posttraumatic stress symptoms who are discharged to home after a motor vehicle collision. Both external validation of this instrument, and work to further develop more accurate tools, are needed. Such tools might benefit public health by enabling the conduct of preventive intervention trials and assisting the growing number of EDs that provide services to trauma survivors aimed at promoting psychological recovery.

Rett syndrome severity estimation with the BioStamp nPoint using interactions between heart rate variability and body movement

Rett syndrome, a rare genetic neurodevelopmental disorder in humans, does not have an effective cure. However, multiple therapies and medications exist to treat symptoms and improve patients' quality of life. As research continues to discover and evaluate new medications for Rett syndrome patients, there remains a lack of objective physiological and motor activity-based (physio-motor) biomarkers that enable the measurement of the effect of these medications on the change in patients' Rett syndrome severity. In our work, using a commercially available wearable chest patch, we recorded simultaneous electrocardiogram and three-axis acceleration from 20 patients suffering from Rett syndrome along with the corresponding Clinical Global Impression-Severity score, which measures the overall disease severity on a 7-point Likert scale. We derived physio-motor features from these recordings that captured heart rate variability, activity metrics, and the interactions between heart rate and activity. Further, we developed machine learning (ML) models to classify high-severity Rett patients from low-severity Rett patients using the derived physio-motor features. For the best-trained model, we obtained a pooled area under the receiver operating curve equal to 0.92 via a leave-one-out-patient cross-validation approach. Finally, we computed the feature popularity scores for all the trained ML models and identified physio-motor biomarkers for Rett syndrome.

An Explainable Spatial-Temporal Graphical Convolutional Network to Score Freezing of Gait in Parkinsonian Patients

Freezing of gait (FOG) is a poorly understood heterogeneous gait disorder seen in patients with parkinsonism which contributes to significant morbidity and social isolation. FOG is currently measured with scales that are typically performed by movement disorders specialists (i.e., MDS-UPDRS), or through patient completed questionnaires (N-FOG-Q) both of which are inadequate in addressing the heterogeneous nature of the disorder and are unsuitable for use in clinical trials The purpose of this study was to devise a method to measure FOG objectively, hence improving our ability to identify it and accurately evaluate new therapies. A major innovation of our study is that it is the first study of its kind that uses the largest sample size (>30 h, N = 57) in order to apply explainable, multi-task deep learning models for quantifying FOG over the course of the medication cycle and at varying levels of parkinsonism severity. We trained interpretable deep learning models with multi-task learning to simultaneously score FOG (cross-validated F1 score 97.6%), identify medication state (OFF vs. ON levodopa; cross-validated F1 score 96.8%), and measure total PD severity (MDS-UPDRS-III score prediction error ≤ 2.7 points) using kinematic data of a well-characterized sample of N = 57 patients during levodopa challenge tests. The proposed model was able to explain how kinematic movements are associated with each FOG severity level that were highly consistent with the features, in which movement disorders specialists are trained to identify as characteristics of freezing. Overall, we demonstrate that deep learning models' capability to capture complex movement patterns in kinematic data can automatically and objectively score FOG with high accuracy. These models have the potential to discover novel kinematic biomarkers for FOG that can be used for hypothesis generation and potentially as clinical trial outcome measures.

Association of Autonomic Activation with traumatic reminder challenges in posttraumatic stress disorder: A co-twin control study

Post-traumatic stress disorder (PTSD) has been associated with cardiovascular disease (CVD), but the mechanisms remain unclear. Autonomic dysfunction, associated with higher CVD risk, may be triggered by acute PTSD symptoms. We hypothesized that a laboratory-based trauma reminder challenge, which induces acute PTSD symptoms, provokes autonomic dysfunction in a cohort of veteran twins. We investigated PTSD-associated real-time physiologic changes with a simulation of traumatic experiences in which the twins listened to audio recordings of a one-minute neutral script followed by a one-minute trauma script. We examined two heart rate variability metrics: deceleration capacity (DC) and logarithmic low frequency (log-LF) power from beat-to-beat intervals extracted from ambulatory electrocardiograms. We assessed longitudinal PTSD status with a structured clinical interview and the severity with the PTSD Symptoms Scale. We used linear mixed-effects models to examine twin dyads and account for cardiovascular and behavioral risk factors. We examined 238 male Veteran twins (age 68 ± 3 years old, 4% black). PTSD status and acute PTSD symptom severity were not associated with DC or log-LF measured during the neutral session, but were significantly associated with lower DC and log-LF during the traumatic script listening session. Long-standing PTSD was associated with a 0.38 (95% confidence interval, -0.83,-0.08) and 0.79 (-1.30,-0.29) standardized unit lower DC and log-LF, respectively, compared to no history of PTSD. Traumatic reminders in patients with PTSD lead to real-time autonomic dysregulation and suggest a potential causal mechanism for increased CVD risk, based on the well-known relationships between autonomic dysfunction and CVD mortality.

An explainable spatial-temporal graphical convolutional network to score freezing of gait in parkinsonian patients

Freezing of gait (FOG) is a poorly understood heterogeneous gait disorder seen in patients with parkinsonism which contributes to significant morbidity and social isolation. FOG is currently measured with scales that are typically performed by movement disorders specialists (ie. MDS-UPDRS), or through patient completed questionnaires (N-FOG-Q) both of which are inadequate in addressing the heterogeneous nature of the disorder and are unsuitable for use in clinical trials The purpose of this study was to devise a method to measure FOG objectively, hence improving our ability to identify it and accurately evaluate new therapies. We trained interpretable deep learning models with multi-task learning to simultaneously score FOG (cross-validated F1 score 97.6%), identify medication state (OFF vs. ON levodopa; cross-validated F1 score 96.8%), and measure total PD severity (MDS-UPDRS-III score prediction error ≤ 2.7 points) using kinematic data of a well-characterized sample of N=57 patients during levodopa challenge tests. The proposed model was able to identify kinematic features associated with each FOG severity level that were highly consistent with the features that movement disorders specialists are trained to identify as characteristic of freezing. In this work, we demonstrate that deep learning models' capability to capture complex movement patterns in kinematic data can automatically and objectively score FOG with high accuracy. These models have the potential to discover novel kinematic biomarkers for FOG that can be used for hypothesis generation and potentially as clinical trial outcome measures.

Use of serial smartphone-based assessments to characterize diverse neuropsychiatric symptom trajectories in a large trauma survivor cohort

The authors sought to characterize adverse posttraumatic neuropsychiatric sequelae (APNS) symptom trajectories across ten symptom domains (pain, depression, sleep, nightmares, avoidance, re-experiencing, anxiety, hyperarousal, somatic, and mental/fatigue symptoms) in a large, diverse, understudied sample of motor vehicle collision (MVC) survivors. More than two thousand MVC survivors were enrolled in the emergency department (ED) and completed a rotating battery of brief smartphone-based surveys over a 2-month period. Measurement models developed from survey item responses were used in latent growth curve/mixture modeling to characterize homogeneous symptom trajectories. Associations between individual trajectories and pre-trauma and peritraumatic characteristics and traditional outcomes were compared, along with associations within and between trajectories. APNS across all ten symptom domains were common in the first two months after trauma. Many risk factors and associations with high symptom burden trajectories were shared across domains. Both across and within traditional diagnostic boundaries, APNS trajectory intercepts, and slopes were substantially correlated. Across all domains, symptom severity in the immediate aftermath of trauma (trajectory intercepts) had the greatest influence on the outcome. An interactive data visualization tool was developed to allow readers to explore relationships of interest between individual characteristics, symptom trajectories, and traditional outcomes ( http://itr.med.unc.edu/aurora/parcoord/ ). Individuals presenting to the ED after MVC commonly experience a broad constellation of adverse posttraumatic symptoms. Many risk factors for diverse APNS are shared. Individuals diagnosed with a single traditional outcome should be screened for others. The utility of multidimensional categorizations that characterize individuals across traditional diagnostic domains should be explored.

Data-driven approach for automatic detection of aortic valve opening: B point detection from impedance cardiogram

Pre-ejection period (PEP), an indicator of sympathetic nervous system activity, is useful in psychophysiology and cardiovascular studies. Accurate PEP measurement is challenging and relies on robust identification of the timing of aortic valve opening, marked as the B point on impedance cardiogram (ICG) signals. The ICG sensitivity to noise and its waveform's morphological variability makes automated B point detection difficult, requiring inefficient and cumbersome expert visual annotation. In this article, we propose a machine learning-based automated algorithm to detect the aortic valve opening for PEP measurement, which is robust against noise and ICG morphological variations. We analyzed over 60 hr of synchronized ECG and ICG records from 189 subjects. A total of 3657 averaged beats were formed using our recently developed ICG noise removal algorithm. Features such as the averaged ICG waveform, its first and second derivatives, as well as high-level morphological and critical hemodynamic parameters were extracted and fed into the regression algorithms to estimate the B point location. The morphological features were extracted from our proposed "variable" physiologically valid search-window related to diverse B point shapes. A subject-wise nested cross-validation procedure was performed for parameter tuning and model assessment. After examining multiple regression models, Adaboost was selected, which demonstrated superior performance and higher robustness to five state-of-the-art algorithms that were evaluated in terms of low mean absolute error of 3.5 ms, low median absolute error of 0.0 ms, high correlation with experts' estimates (Pearson coefficient = 0.9), and low standard deviation of errors of 9.2 ms. For reproducibility, an open-source toolbox is provided.

Anxiety sensitivity as a transdiagnostic risk factor for trajectories of adverse posttraumatic neuropsychiatric sequelae in the AURORA study

Anxiety sensitivity, or fear of anxious arousal, is cross-sectionally associated with a wide array of adverse posttraumatic neuropsychiatric sequelae, including symptoms of posttraumatic stress disorder, depression, anxiety, sleep disturbance, pain, and somatization. The current study utilizes a large-scale, multi-site, prospective study of trauma survivors presenting to emergency departments. Hypotheses tested whether elevated anxiety sensitivity in the immediate posttrauma period is associated with more severe and persistent trajectories of common adverse posttraumatic neuropsychiatric sequelae in the eight weeks posttrauma. Participants from the AURORA study (n = 2,269 recruited from 23 emergency departments) completed self-report assessments over eight weeks posttrauma. Associations between heightened anxiety sensitivity and more severe and/or persistent trajectories of trauma-related symptoms identified by growth mixture modeling were analyzed. Anxiety sensitivity assessed two weeks posttrauma was associated with severe and/or persistent posttraumatic stress, depression, anxiety, sleep disturbance, pain, and somatic symptoms in the eight weeks posttrauma. Effect sizes were in the small to medium range in multivariate models accounting for various demographic, trauma-related, pre-trauma mental health-related, and personality-related factors. Anxiety sensitivity may be a useful transdiagnostic risk factor in the immediate posttraumatic period identifying individuals at risk for the development of adverse posttraumatic neuropsychiatric sequelae. Further, considering anxiety sensitivity is malleable via brief intervention, it could be a useful secondary prevention target. Future research should continue to evaluate associations between anxiety sensitivity and trauma-related pathology.

Predicting at-risk opioid use three months after ed visit for trauma: Results from the AURORA study

Objective

Whether short-term, low-potency opioid prescriptions for acute pain lead to future at-risk opioid use remains controversial and inadequately characterized. Our objective was to measure the association between emergency department (ED) opioid analgesic exposure after a physical, trauma-related event and subsequent opioid use. We hypothesized ED opioid analgesic exposure is associated with subsequent at-risk opioid use.

Methods

Participants were enrolled in AURORA, a prospective cohort study of adult patients in 29 U.S., urban EDs receiving care for a traumatic event. Exclusion criteria were hospital admission, persons reporting any non-medical opioid use (e.g., opioids without prescription or taking more than prescribed for euphoria) in the 30 days before enrollment, and missing or incomplete data regarding opioid exposure or pain. We used multivariable logistic regression to assess the relationship between ED opioid exposure and at-risk opioid use, defined as any self-reported non-medical opioid use after initial ED encounter or prescription opioid use at 3-months.

Results

Of 1441 subjects completing 3-month follow-up, 872 participants were included for analysis. At-risk opioid use occurred within 3 months in 33/620 (5.3%, CI: 3.7,7.4) participants without ED opioid analgesic exposure; 4/16 (25.0%, CI: 8.3, 52.6) with ED opioid prescription only; 17/146 (11.6%, CI: 7.1, 18.3) with ED opioid administration only; 12/90 (13.3%, CI: 7.4, 22.5) with both. Controlling for clinical factors, adjusted odds ratios (aORs) for at-risk opioid use after ED opioid exposure were: ED prescription only: 4.9 (95% CI 1.4, 17.4); ED administration for analgesia only: 2.0 (CI 1.0, 3.8); both: 2.8 (CI 1.2, 6.5).

Conclusions

ED opioids were associated with subsequent at-risk opioid use within three months in a geographically diverse cohort of adult trauma patients. This supports need for prospective studies focused on the long-term consequences of ED opioid analgesic exposure to estimate individual risk and guide therapeutic decision-making.

An investigation of privacy preservation in deep learning-based eye-tracking

Background

The expanding usage of complex machine learning methods such as deep learning has led to an explosion in human activity recognition, particularly applied to health. However, complex models which handle private and sometimes protected data, raise concerns about the potential leak of identifiable data. In this work, we focus on the case of a deep network model trained on images of individual faces.

Materials and methods

A previously published deep learning model, trained to estimate the gaze from full-face image sequences was stress tested for personal information leakage by a white box inference attack. Full-face video recordings taken from 493 individuals undergoing an eye-tracking- based evaluation of neurological function were used. Outputs, gradients, intermediate layer outputs, loss, and labels were used as inputs for a deep network with an added support vector machine emission layer to recognize membership in the training data.

Results

The inference attack method and associated mathematical analysis indicate that there is a low likelihood of unintended memorization of facial features in the deep learning model.

Conclusions

In this study, it is showed that the named model preserves the integrity of training data with reasonable confidence. The same process can be implemented in similar conditions for different models.

Comparison of deep transfer learning algorithms and transferability measures for wearable sleep staging

Background

Obtaining medical data using wearable sensors is a potential replacement for in-hospital monitoring, but the lack of data for such sensors poses a challenge for development. One solution is using in-hospital recordings to boost performance via transfer learning. While there are many possible transfer learning algorithms, few have been tested in the domain of EEG-based sleep staging. Furthermore, there are few ways for determining which transfer learning method will work best besides exhaustive testing. Measures of transferability do exist, but are typically used for selection of pre-trained models rather than algorithms and few have been tested on medical signals. We tested several supervised transfer learning algorithms on a sleep staging task using a single channel of EEG (AF7-Fpz) captured from an in-home commercial system.

Results

Two neural networks—one bespoke and another state-of-art open-source architecture—were pre-trained on one of six source datasets comprising 11,561 subjects undergoing clinical polysomnograms (PSGs), then re-trained on a target dataset of 75 full-night recordings from 24 subjects. Several transferability measures were then tested to determine which is most effective for assessing performance on unseen target data. Performance on the target dataset was improved using transfer learning, with re-training the head layers being the most effective in the majority of cases (up to 63.9% of cases). Transferability measures generally provided significant correlations with accuracy (up to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r = -0.53$$\end{document}r=-0.53).

Conclusion

Re-training the head layers provided the largest performance boost. Transferability measures are useful indicators of transfer learning effectiveness.

The temporal relationships between sleep disturbance and autonomic dysregulation: A co-twin control study

INTRODUCTION

Sleep disturbance is associated with autonomic dysregulation, but the temporal directionality of this relationship remains uncertain. The objective of this study was to evaluate the temporal relationships between objectively measured sleep disturbance and daytime or nighttime autonomic dysregulation in a co-twin control study.

METHODS

A total of 68 members (34 pairs) of the Vietnam Era Twin Registry were studied. Twins underwent 7-day in-home actigraphy to derive objective measures of sleep disturbance. Autonomic function indexed by heart rate variability (HRV) was obtained using 7-day ECG monitoring with a wearable patch. Multivariable vector autoregressive models with Granger causality tests were used to examine the temporal directionality of the association between daytime and nighttime HRV and sleep metrics, within twin pairs, using 7-day collected ECG data.

RESULTS

Twins were all male, mostly white (96%), with mean (SD) age of 69 (2) years. Higher daytime HRV across multiple domains was bidirectionally associated with longer total sleep time and lower wake after sleep onset; these temporal dynamics were extended to a window of 48 h. In contrast, there was no association between nighttime HRV and sleep measures in subsequent nights, or between sleep measures from previous nights and subsequent nighttime HRV.

CONCLUSIONS

Daytime, but not nighttime, autonomic function indexed by HRV has bidirectional associations with several sleep dimensions. Dysfunctions in autonomic regulation during wakefulness can lead to subsequent shorter sleep duration and worse sleep continuity, and vice versa, and their influence on each other may extend beyond 24 h.

Persistent Dissociation and Its Neural Correlates in Predicting Outcomes After Trauma Exposure

OBJECTIVE

Dissociation, a disruption or discontinuity in psychological functioning, is often linked with worse psychiatric symptoms; however, the prognostic value of dissociation after trauma is inconsistent. Determining whether trauma-related dissociation is uniquely predictive of later outcomes would enable early identification of at-risk trauma populations. The authors conducted the largest prospective longitudinal biomarker study of persistent dissociation to date to determine its predictive capacity for adverse psychiatric outcomes following acute trauma.

METHODS

All data were part of the Freeze 2 data release from the Advancing Understanding of Recovery After Trauma (AURORA) study. Study participants provided self-report data about persistent derealization (N=1,464), a severe type of dissociation, and completed a functional MRI emotion reactivity task and resting-state scan 2 weeks posttrauma (N=145). Three-month follow-up reports were collected of posttraumatic stress, depression, pain, anxiety symptoms, and functional impairment.

RESULTS

Derealization was associated with increased ventromedial prefrontal cortex (vmPFC) activation in the emotion reactivity task and decreased resting-state vmPFC connectivity with the cerebellum and orbitofrontal cortex. In separate analyses, brain-based and self-report measures of persistent derealization at 2 weeks predicted worse 3-month posttraumatic stress symptoms, distinct from the effects of childhood maltreatment history and current posttraumatic stress symptoms.

CONCLUSIONS

The findings suggest that persistent derealization is both an early psychological and biological marker of worse later psychiatric outcomes. The neural correlates of trauma-related dissociation may serve as potential targets for treatment engagement to prevent posttraumatic stress disorder. These results underscore dissociation assessment as crucial following trauma exposure to identify at-risk individuals, and they highlight an unmet clinical need for tailored early interventions.

Issues in the automated classification of multilead ecgs using heterogeneous labels and populations

Objective.The standard twelve-lead electrocardiogram (ECG) is a widely used tool for monitoring cardiac function and diagnosing cardiac disorders. The development of smaller, lower-cost, and easier-to-use ECG devices may improve access to cardiac care in lower-resource environments, but the diagnostic potential of these devices is unclear. This work explores these issues through a public competition: the 2021 PhysioNet Challenge. In addition, we explore the potential for performance boosting through a meta-learning approach.Approach.We sourced 131,149 twelve-lead ECG recordings from ten international sources. We posted 88,253 annotated recordings as public training data and withheld the remaining recordings as hidden validation and test data. We challenged teams to submit containerized, open-source algorithms for diagnosing cardiac abnormalities using various ECG lead combinations, including the code for training their algorithms. We designed and scored the algorithms using an evaluation metric that captures the risks of different misdiagnoses for 30 conditions. After the Challenge, we implemented a semi-consensus voting model on all working algorithms.Main results.A total of 68 teams submitted 1,056 algorithms during the Challenge, providing a variety of automated approaches from both academia and industry. The performance differences across the different lead combinations were smaller than the performance differences across the different test databases, showing that generalizability posed a larger challenge to the algorithms than the choice of ECG leads. A voting model improved performance by 3.5%.Significance.The use of different ECG lead combinations allowed us to assess the diagnostic potential of reduced-lead ECG recordings, and the use of different data sources allowed us to assess the generalizability of the algorithms to diverse institutions and populations. The submission of working, open-source code for both training and testing and the use of a novel evaluation metric improved the reproducibility, generalizability, and applicability of the research conducted during the Challenge.

Hippocampal Threat Reactivity Interacts with Physiological Arousal to Predict PTSD Symptoms

Hippo campal impairments are reliably associated with post-traumatic stress disorder (PTSD); however, little research has characterized how increased threat sensitivity may interact with arousal responses to alter hippocampal reactivity, and further how these interactions relate to the sequelae of trauma-related symptoms. In a sample of individuals recently exposed to trauma (N = 116, 76 female), we found that PTSD symptoms at 2 weeks were associated with decreased hippocampal responses to threat as assessed with fMRI. Further, the relationship between hippocampal threat sensitivity and PTSD symptomology only emerged in individuals who showed transient, high threat-related arousal, as assayed by an independently collected measure of fear potentiated startle. Collectively, our finding suggests that development of PTSD is associated with threat-related decreases in hippocampal function because of increases in fear-potentiated arousal.SIGNIFICANCE STATEMENT Alterations in hippocampal function linked to threat-related arousal are reliably associated with post-traumatic stress disorder (PTSD); however, how these alterations relate to the sequelae of trauma-related symptoms is unknown. Prior models based on nontrauma samples suggest that arousal may impact hippocampal neurophysiology leading to maladaptive behavior. Here we show that decreased hippocampal threat sensitivity interacts with fear-potentiated startle to predict PTSD symptoms. Specifically, individuals with high fear-potentiated startle and low, transient hippocampal threat sensitivity showed the greatest PTSD symptomology. These findings bridge literatures of threat-related arousal and hippocampal function to better understand PTSD risk.

Passively Captured Interpersonal Social Interactions and Motion From Smartphones for Predicting Decompensation in Heart Failure: Observational Cohort Study

Background

Heart failure (HF) is a major cause of frequent hospitalization and death. Early detection of HF symptoms using smartphone-based monitoring may reduce adverse events in a low-cost, scalable way.

Objective

We examined the relationship of HF decompensation events with smartphone-based features derived from passively and actively acquired data.

Methods

This was a prospective cohort study in which we monitored HF participants’ social and movement activities using a smartphone app and followed them for clinical events via phone and chart review and classified the encounters as compensated or decompensated by reviewing the provider notes in detail. We extracted motion, location, and social interaction passive features and self-reported quality of life weekly (active) with the short Kansas City Cardiomyopathy Questionnaire (KCCQ-12) survey. We developed and validated an algorithm for classifying decompensated versus compensated clinical encounters (hospitalizations or clinic visits). We evaluated models based on single modality as well as early and late fusion approaches combining patient-reported outcomes and passive smartphone data. We used Shapley additive explanation values to quantify the contribution and impact of each feature to the model.

Results

We evaluated 28 participants with a mean age of 67 years (SD 8), among whom 11% (3/28) were female and 46% (13/28) were Black. We identified 62 compensated and 48 decompensated clinical events from 24 and 22 participants, respectively. The highest area under the precision-recall curve (AUCPr) for classifying decompensation was with a late fusion approach combining KCCQ-12, motion, and social contact features using leave-one-subject-out cross-validation for a 2-day prediction window. It had an AUCPr of 0.80, with an area under the receiver operator curve (AUC) of 0.83, a positive predictive value (PPV) of 0.73, a sensitivity of 0.77, and a specificity of 0.88 for a 2-day prediction window. Similarly, the 4-day window model had an AUC of 0.82, an AUCPr of 0.69, a PPV of 0.62, a sensitivity of 0.68, and a specificity of 0.87. Passive social data provided some of the most informative features, with fewer calls of longer duration associating with a higher probability of future HF decompensation.

Conclusions

Smartphone-based data that includes both passive monitoring and actively collected surveys may provide important behavioral and functional health information on HF status in advance of clinical visits. This proof-of-concept study, although small, offers important insight into the social and behavioral determinants of health and the feasibility of using smartphone-based monitoring in this population. Our strong results are comparable to those of more active and expensive monitoring approaches, and underscore the need for larger studies to understand the clinical significance of this monitoring method.