A CNN Based Robust Approach for Pneumonia Detection

Pneumonia is a severe respiratory illness the prevalence of which is still very high and continues to cause high morbidity and mortality globally. Determination of the disease at its initial stage and its correct identification are so vital in the handling of the sickness but the approaches used traditionally may take a lot of time and may involve cumulative errors. To overcome this, the paper develops a convolutional neural network (CNN) based model to perform an automated diagnosis of pneumonia from chest X-ray images. Using a dataset the model is trained and validated for 20 epochs and it is found that the accuracy comes out to be 96%. The proposed method relies on the feature learning, and hierarchical learning of CNNs to quickly identify the nature of medical images in a non-invasive way. Herein, the studies show that CNNs have a promising practice for supporting existing diagnostic approaches to reduce the work strains on healthcare workers and benefit patients. The future work will be directed to collect more data, optimize the architecture of the consequent models, and investigate how it could be useful in other clinical domains.