Smart Assistant to Ease the Process of COVID-19 and Pneumonia Detection

Abstract

COVID -19 is one of the most contagious diseases in the 21st century. Therefore, there's an emerging need to contrive an accurate, gradual new method to identify this deadly virus. Apropos, we present “Smart assistance to ease the process of COVID -19/pneumonia detection” mobile application that can use to identify covid-19 contemplating patient's symptoms, health history, breathing information, chest CT scan and chest X-ray images. Stage 1 of the proposed application will prognosticate the danger level of the patient utilizing symptoms, breathing information, health history using machine learning techniques. Recognition and drawing out of patient's health background information by engaging the user to maximize the accuracy of the outcome is the main objective of this stage. Stage 2 of the application will identify COVID-19 by a chest X-ray/CT scan image, and it predicts the danger level using deep learning techniques. Classify the image to predict the danger level for COVID-19 is the main objective of this phase. Subsequently, all the predictions are sent to a physician and validate the outcome. Finally, patient will be notified about the results. This automatized application is built with the intention of reducing the cost of covid-19 identification tests like PCR tests and to give precise results as soon as possible. Our motive is to show that the proposed application could be a finer alternative for already existing COVID -19 identification tests. As a result, we achieved the best accuracy of 92%, 96% for CT scan, X-ray images classification and 94.08%, 74.19% accuracy for health history information analysis and breathing information analysis. We also achieved 94%, 71% accuracies for the COVID-19 prediction model and severity level prediction model based on symptoms.