Fitbeat: COVID-19 estimation based on wristband heart rate using a contrastive convolutional auto-encoder

Abstract

This study proposes a contrastive convolutional auto-encoder (contrastive CAE), a combined architecture of an auto-encoder and contrastive loss, to identify individuals with suspected COVID-19 infection using heart-rate data from participants with multiple sclerosis (MS) in the ongoing RADAR-CNS mHealth research project. Heart-rate data was remotely collected using a Fitbit wristband. COVID-19 infection was either confirmed through a positive swab test, or inferred through a self-reported set of recognised symptoms of the virus. The contrastive CAE outperforms a conventional convolutional neural network (CNN), a long short-term memory (LSTM) model, and a convolutional auto-encoder without contrastive loss (CAE). On a test set of 19 participants with MS with reported symptoms of COVID-19, each one paired with a participant with MS with no COVID-19 symptoms, the contrastive CAE achieves an unweighted average recall of , a sensitivity of and a specificity of , an area under the receiver operating characteristic curve (AUC-ROC) of 0.944, indicating a maximum successful detection of symptoms in the given heart rate measurement period, whilst at the same time keeping a low false alarm rate.

Consumer-grade wearables identify changes in multiple physiological systems during COVID-19 disease progression

“… 4 , 9 , 10 , 11 In addition, recent work has explored how wearables may be used for real-time detection or retrospective analysis of COVID-19, for example, as a way to reduce the spread of infection. 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 Many of these approaches use a physiological signal from wearables, such as heart rate (HR), along with black box or machine learning methods to classify case status or disease progression. 22 , 23 , 24 …”

Biopsychosocial response to the COVID-19 lockdown in people with major depressive disorder and multiple sclerosis

“…An elevated resting mean HR between the lockdown and post-lockdown may be caused by other factors such as psychological stress perception (38) acute respiratory infections (51,82) or an increased alcohol intake (83) et al 2021). 21.8% of our participants with MS (total sample=399) reported major symptoms similar to COVID-19 symptoms (42) and found an association with HR parameters (51). Low HR variability was reported in people with MS (84), which may be caused by a significant increase in sympathetic cardiovascular tone (38) and can influence the course of the disease (39).…”

“…Briefly, for assessing the HR profile, the mean level and standard deviation of HR were computed for the whole day, during resting/sedentary periods during the day and during resting/sedentary periods at night. This device was previously proven to measure HR accurately (31,(50)(51)(52). Sedentary, light, moderate and vigorous activity were selected due to their strong association with depression (53-55) and during the pandemic ( i.e.…”

Audio Self-supervised Learning: A Survey

“…A too small τ loses the tolerance to group the similar input samples and hence may break the underlying semantic structure, by this harming the learnt features for its use in downstream tasks. The effect of the temperature parameter is similar as the margin value set in Equation ( 5), which has been investigated in detail by [80].…”