Deep Learning for Pulmonary Disease Identification

The ubiquitous use of machine learning in healthcare including, but not limited to analytics, data collection and processing, differential diagnosis and wearable health tracking has led to vast amounts of research in the domain. With the advent of deep learning, there has been an increasing interest in disease classification using various medical imaging techniques. This paper aims to use deep learning architecture using convolutional neural networks to accurately identify different pulmonary diseases such as pneumonia and COVID-19 from a healthy human being’s chest x-rays. The model, inherently based on CNN, uses multi-layer variables including convolutional, batch normalization, ReLU, max pooling, fully connected, SoftMax, and classification layers to make up the architecture. The dataset consisting of 3,000 images, varying from healthy lung scans to those of COVID-19 patients and pneumonia patients’ x-ray scans, aid in the training and testing of the model, which achieves an accuracy of 85%. The work is aimed to be of assistance to healthcare professionals and to be of advantage to them for faster and better decision making.