Detection of Pneumonia from Chest X-ray Using Deep Learning

Pneumonia has been around from a very long time affecting lakhs of people across the globe. Now, with the pandemic COVID-19 around and pneumonia being one of the symptoms of COVID-19, it must be detected in a very early stage so that the person can be treated without any further complications. Chest X-ray images can be used to detect pneumonia in an easier and cheaper way. Hence, radiologists can use deep learning algorithms to diagnose pneumonia effectively. It’s very crucial to consider the fact that not every pneumonia is because of COVID-19, hence we are classifying the X-ray into three categories, i.e., bacterial pneumonia, viral pneumonia, and normal using transfer learning like VGG19, Xception, Densenet121, and InceptionV3. A Deep Learning model’s accuracy is highly dependent on the amount of training set. So, we implement a generative adversarial network (GAN) for data augmentation. We observed that Xception achieved the highest accuracy of 83% with a precision of 85%, recall of 83%, and F1 score of 83% at the end of 100th epoch.