EfficientCovNet: A CNN-based approach to detect various pulmonary diseases including COVID-19 using modified EfficientNet

Abstract

The ongoing COVID-19 pandemic has brought about severe consequences, including numerous tragic losses of loved ones and significant disruptions to the global economy and daily routines. To curb the spread of the virus, timely and precise detection of COVID-19 cases is crucial. However, the traditional RT-PCR testing system is time-consuming, and its sensitivity can vary depending on the testing environment. In this context, X-ray images of the lungs have emerged as a potential tool for COVID-19 detection, as the virus predominantly affects the lungs. However, distinguishing COVID-19 from other significant pulmonary diseases such as Viral Pneumonia and Lung Opacity is essential to avoid misdiagnosis and ensure appropriate treatment.In this study, we propose a modified efficient Convolutional Neural Network (CNN)-based architecture named EfficientCovNet for accurate classification of different lung diseases, including COVID-19, Viral Pneumonia, and Lung Opacity, as they all share common symptoms of shortness of breath. Our proposed model achieved a remarkable test accuracy of +2% (96.12%) compared to existing CNN models, as demonstrated pictorially in the later sections. Furthermore, we utilized a comparatively large dataset of 20,000 radiograph images encompassing various disease categories, enhancing the robustness and reliability of our model. Our findings highlight the potential of CNN-based models in accurately classifying lung diseases, including COVID-19, and can aid in effective disease management and containment strategies.