Automated evaluation of lung health in covid-19 patients using x-ray scans and enhanced bilateral convolutional neural networks: Potential for lung cancer detection

Abstract

Author(s): Tarang Bhatnagar, K G Patel, Prashant Rajaram Patil, Asha K, Sudhanshu Dev, Sumit Tripathi

Background: In December 2019, the rare Coronavirus Disease (COVID-19) pandemic first appeared in Wuhan, China. Life, health and the international economy have all been severely impacted. Timely diagnosis, appropriate staging, and early treatment are the goals for inpatients with COVID attacks. The arrival of the COVID-19 pandemic changed the scenario of screening and diagnosis to stop the outbreak and treat patients rapidly for rapid detection of positive cases, especially those who are suffering from several lung diseases, including lung cancer, which must be detected early.

Objective: Research is to propose Enhanced Bilateral Convolutional Neural Networks (EBCNN) as a means of automating the assessment of the pulmonary health of Covid-19 participants utilizing X-ray images.

Methods: In this study, the CXR image dataset were collected. The acquired X-ray images provide variety and depth to the categorization. The variation in mean values across the groups conducted unpaired, 2-tailed t-tests for statistical analysis. To measure the accuracy of the classifiers, the ROC curve was plotted.

Result: In this investigation, the real-time object identification system's classifier was the EBCNN model. The research shows that EBCNN outperforms standard techniques in terms of F1 score, accuracy and recall.

Conclusion: The approach can be used to help radiologists validate their first patient scanning and can also be used through the cloud to significant health risks right away. The results demonstrate that all classifiers work with the obtaining accuracy score of over 98.5%.

Share this article