Search

Background: We investigated machine learning based identification of the pre-symptomatic coronavirus disease 2019 (COVID-19) and detection of infection-related changes in physiology using a wearable device (the Ava bracelet). Methods: Participants from an ongoing cohort study (GAPP) of the general population in Liechtenstein were included in the current sub-study (COVI-GAPP). Nightly they wore the fertility bracelet that measured every ten seconds skin temperature, heart rate, respiratory rate, skin perfusion, and heart rate variability. Participants reported daily symptoms in a complementary app. Laboratory reverse transcription polymerase chain reaction (RT-PCR) and/or COVID-19 serology samples were collected from all participants. Long short-term memory (LSTM) based recurrent neural networks (RNN) were chosen for the binary classification of an individual as healthy or infected on a given day in a derivation and validation procedure. Findings: A total of 1ž5 million hours of physiological data were recorded from 1163 participants (mean age 44 +/- 5ž5 years). COVID-19 was confirmed in 127 participants. Of these, 66 (52%) had worn their device from baseline to symptom onset and were included in the analysis and RNN. Multi-level modelling revealed significantly different values in pre- versus post-symptomatic respiratory rate, temperature, heart rate, heart rate variability ratio, and skin perfusion. The developed RNN algorithm had a recall of 0ž73 in the training set and 0ž68 in the testing set (overall recall of 0ž71) when detecting COVID-19 up to two days prior to symptom onset. Interpretation: Our proposed RNN algorithm identified 71% of COVID-19 positive participants two days prior to symptom onset. Wearable sensor technology can therefore enable COVID-19’ detection during the pre-symptomatic period. Funding: IMI grant agreement number 101005177, the Princely House of the Principality of Liechtenstein, the government of the Principality of Liechtenstein, and the Hanela Foundation in Switzerland. Declaration of Interest: Lorenz Risch, and Martin Risch are key shareholders of the Dr Risch Medical Laboratory. David Conen has received consulting fees from Roche Diagnostics, outside of the current work. The other authors have no financial or personal conflicts of interest to declare. Ethical Approval: The local ethics committee approved the study protocol, and written informed consent was obtained from each participant (BASEC 2020-00786).

Specialized brain structures encode spatial locations and movements, yet there is growing evidence that this information is also represented in the rodent medial prefrontal cortex (mPFC). Disambiguating such information from the encoding of other types of task-relevant information has proven challenging. To determine the extent to which movement and location information is relevant to mPFC neurons, tetrodes were used to record neuronal activity while limb positions, poses (i.e., recurring constellations of limb positions), velocity, and spatial locations were simultaneously recorded with two cameras every 200 ms as rats freely roamed in an experimental enclosure. Regression analyses using generalized linear models revealed that more than half of the individual mPFC neurons were significantly responsive to at least one of the factors, and many were responsive to more than one. On the other hand, each factor accounted for only a very small portion of the total spike count variance of any given neuron (