An AI healthcare ecosystem framework for Covid-19 detection and forecasting using CronaSona

Abstract

The primary purpose of this paper is to establish a healthcare ecosystem framework for COVID-19, CronaSona. Unlike some studies that focus solely on detection or forecasting, CronaSona aims to provide a holistic solution, for managing data and/or knowledge, incorporating detection, forecasting, expert advice, treatment recommendations, real-time tracking, and finally visualizing results. The innovation lies in creating a comprehensive healthcare ecosystem framework and an application that not only aids in COVID-19 diagnosis but also addresses broader health challenges. The main objective is to introduce a novel framework designed to simplify the development and construction of applications by standardizing essential components required for applications focused on addressing diseases. CronaSona includes two parts, which are stakeholders and shared components, and four subsystems: (1) the management information subsystem, (2) the expert subsystem, (3) the COVID-19 detection and forecasting subsystem, and (4) the mobile tracker subsystem. In the proposed framework, a CronaSona app. was built to try to put the virus under control. It is a reactive mobile application for all users, especially COVID-19 patients and doctors. It aims to provide a reliable diagnostic tool for COVID-19 using deep learning techniques, accelerating diagnosis and referral processes, and focuses on forecasting the transmission of COVID-19. It also includes a mobile tracker subsystem for monitoring potential carriers and minimizing the virus spread. It was built to compete with other applications and to help people face the COVID-19 virus. Upon receiving the proposed framework, an application was developed to validate and test the framework's functionalities. The main aim of the developed application, CronaSona app., is to develop and test a reliable diagnostic tool using deep learning techniques to avoid increasing the spread of the disease as much as possible and to accelerate the diagnosis and referral of patients by detecting COVID-19 features from their chest X-ray images. By using CronaSona, human health is saved and stress is reduced by knowing everything about the virus. It performs with the highest accuracy, F1-score, and precision, with consecutive values of 97%, 97.6%, and 96.6%.

Keywords: COVID-19; COVID-19 detection; COVID-19 ecosystem; COVID-19 forecasting; COVID-19 framework; Chest X-ray; CronaSona; Deep learning.

Fig. 1

The proposed COVID-19 ecosystem framework

Algorithm 1

Detection Module

Fig. 2

COVID-19 prediction model (CronaSona)

Algorithm 2

Forecasting Module

Fig. 3

A comparison between the used DL models vs. CronaSona in terms of accuracy

Fig. 4

The applied five DL classification model accuracy and loss for each epoch. a Inception_V3 accuracy, b Inception_V3 loss, c Inception_ResNet50_V2 accuracy, d Inception_ResNet50_V2 loss, e DenseNet accuracy, f DenseNet loss, g MobileNet_V2 accuracy, h MobileNet_V2 loss, i NasNetMobile accuracy, j NasNetMobile loss, k Xception accuracy, and l Xception loss

Fig. 5

The confusion matrix of the six models for COVID-19, normal, respectively. a Inception_V3, b ResNet50_V2, c DenseNet, d MobileNet_V2, e NasNetMobile, and f Xception

Fig. 6

The forecasting curve for confirmed cases

Fig. 7

CronaSona interface design