Analysing and Identifying COVID-19 Risk Factors Using Machine Learning Algorithm with Smartphone Application

This study is divided into risk factor analysis (RFA) and proposed system architecture (PSA). The light gradient boosting machine (LightGBM) algorithm in the RFA will work with the PSA to predict the risk factors. The results, efficacy, and performance will be validated via a ROC-AUC curve. Therefore, a system usability scale (SUS) procedure will be implemented to increase the performance. If the SUS score reaches 85–99 and 100 thresholds, it will be classified as appropriate for use and robust. The prediction score thresholds will be 0–100. If the score is below 25, it will be classified as normal, 26–50 as moderate, 51–70 risk, and 71–100 as severe. Due to a shortage of experienced staff and intelligent technology, it is becoming progressively difficult to reduce COVID-19 fatality rates. In this research, a lightweight mobile application has been suggested from which the significant patterns and factors can be recognised. Furthermore, it will assist both doctors and patients become aware of COVID-19 risk factors and take the required steps to mitigate them. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

Diagnosis of COVID-19 infected lungs from chest x-ray images using deep learning algorithms

Severe Acute Respiratory Symptom Coronavirus 2 (SARS-CoV-2) is newly discovered as a beta coronavirus. The virus-induced unexplained etiological pneumonia and is referred to as the 2019 Coronavirus Disease (COVID-19). Though the disease has appeared in a new way, there is no medication for transited patients. So, for diagnosing the COVID-19 infected lungs from X-Ray images, an automated technique has been suggested in this manuscript. The proposed system is divided into two stages: Image Acquisition and Selection of Algorithms. In the IAA, the training data's size has been increased by augmenting the image in different ways. The Algorithm Selection portion explained the Convolutional Neural Network (CNN) and VGG19. The Tuning of hyperparameters section determines the precise hyperparameter combination in order to maximize the model's performance. In this study, CNN and VGG19 are used and found accuracy scores of 97% and 67%, respectively. The comparative analysis shows that the propound method acts better than the solution that exists. Eventually, Precision, Recall, and F1 score have been extracted and interpreted the model's loss functions in the research. This research has carried out by focusing on essential aspects in terms of COVID-19. Therefore, for the diagnosis of coronavirus infection, the technique can be used effectively. © 2021 AIUB Office of Research and Publication. All rights reserved.