Early Prediction of COVID-19 Onset by Fuzzy-Neuro Inference

The capability to self-monitor and to be informed of health, empower people in preventing the spread of COVID-19 pandemic and improve health outcomes. Most of infected people are asymptomatic and they are able to spread the virus so detecting early the COVID-19 infection onset gives two advantages, it limits virus spreading and it limits the disease damages to the person. Vital signs measurements (heart rate, blood oxygen saturation, respiratory rate, heart rate variability, body temperature, electro dermal activity, motion activity) could be fused to infer about COVID-19 positive or COVID-19 negative person’s status at least one week in advance. Due to the fuzziness of measurements a softcomputing approach is requested, performing better than hardcomputing approach. Fuzzy-neuro inference paradigms, mainly the Evolving Connectionist Systems ECOS has the required capabilities to fit well such application. A wearable device (wrist) for vital signs data logging and measurement was been deployed and wireless connected to a host computing platform to run a fuzzy-neuro inference paradigm, the Evolving Fuzzy Neural Network (EFuNN) trained and tested on a dataset built by vital signs measurements. Test results confirmed the fast and effective learning capability of EfuNN paradigm. Better performance was been achieved applying the EFuNN’s evolving and online learning capabilities.