Analysis of COVID Patients Employing Approximate Entropy and Deep Learning for Classification and Early Diagnosis

Abstract

Due to its rapid propagation and enormous number of infected people, COVID-19 is the greatest pandemic in the past 100 years, with millions of deaths. The need for accessible, quick, and non-invasive diagnostic techniques persists despite a decline in cases recently. Because of this, in the current work we develop a densely connected neural network that uses heart rate data to identify between patients with COVID and healthy individuals. The Stanford University database was used, which underwent a feature extraction and the usage of approximation entropy. With an accuracy of 93% and an AUC of 0.956, the results demonstrated to be more than good at categorization, supporting the usefulness of this approach for the accurate identification of COVID cases.