Deep Learning Model for Edge Devices for COVID-19 Detection from CXR Images

The novel coronavirus 2019 (COVID-19) emerged in December 2019, and subsequently underwent rapid global spread, culminating in a pandemic declaration. The emergence of COVID-19 has affected public health significantly and sparked major socio-economic crises. Worldwide, more than 6 million deaths linked to COVID-19 and more than 758 million cases have been reported. Prompt and precise identification of individuals as either healthy or infected is a crucial strategy in managing and preventing the spread of coronavirus outbreaks. Recent findings suggest that patients with COVID-19 infection can be identified by chest radiography anomalies. Chest X-ray (CXR) pictures are more easily obtainable when compared to chest computed tomography (CT) images, especially in less economically developed regions where CT machines are prohibitively costly. In this research, we introduce an exceptionally lightweight convolutional neural network (CNN) tailored for the automated assessment of chest X-ray images, aiming to differentiate between COVID-19, non-COVID-19, and normal conditions. Our model has total of 59,467(0.059 million) parameters, our model's total size is 19.67 MB. Our model surpasses other COVID-19 detection models due to its substantially reduced parameter count and comparable accuracy, making it well-suited for use on machines with limited computational capabilities.