C-COVIDNet: A CNN Model for COVID-19 Detection Using Image Processing
Neha Rajawat, Bharat Singh Hada, Mayank Meghawat, Soniya Lalwani, Rajesh Kumar
Abstract
COVID-19 has become a global disaster that has disturbed the socioeconomic fabric of the world. Efficient and cost-effective diagnosis methods are very much required for better treatment and eliminating false cases for COVID-19. COVID-19 disease is a type of respiratory syndrome, thus lung X-ray analysis has got the attention for an effective diagnosis. Hence, the proposed study introduces an Image processing based COVID-19 detection model C-COVIDNet, which is trained on a dataset of chest X-ray images belonging to three categories: COVID-19, Pneumonia, and Normal person. Image preprocessing pipeline is used for extracting the region of interest (ROI), so that the required features may be present in the input. This lightweight convolution neural network (CNN) based approach has achieved an accuracy of 97.5% and an F1-score of 97.91%. Model input images are generated in batches using a custom data generator. The performance of C-COVIDNet has outperformed the state-of-the-art. The promising results will surely help in accelerating the development of deep learning-based COVID-19 diagnosis tools using radiography.