Enhanced CNN Models for Precise Classification of Pneumonia and COVID-19

Pneumonia is a contagious respiratory disease that poses a significant risk, particularly for children below the age of five, as it is the primary factor contributing to death in this demographic. Pneumonia is commonly detected with the regular utilization of chest X-ray (CXR) imaging. Machine learning methods have shown efficacy in automating the classification of images in the imaging in medicine area, offering substantial computer-assisted diagnostic assistance. Convolutional Neural Networks (CNN) are renowned for their capacity to attain exceptional classification accuracy and expedite the process of diagnosis. The study compared two robust CNN models, namely Xception and VGG-19 with a new CNN model. It has been developed to improve specificity and classification accuracy in diagnosing respiratory ailments using CXR images. The model is evaluated using a curated COVID19 dataset from Mendeley, which includes a large number of carefully chosen photographs. The model yields promising results, aiding in the rapid identification of individuals afflicted with the virus, and enhancing the effectiveness of CNN-based designs in diagnosing respiratory ailments. The outcome reveals that our recommended CNN model works better than previously trained models. In particular, our recommended model exceeds the pre-trained models VGG-19 and Xception with an incredible accuracy of $98.01 \%$. Our recommended paradigm permits early diagnosis, which in turn promotes timely interventions.