A Comparative Study of Machine Learning and Deep Learning Techniques on X-ray Images for Pneumonia

Abstract

Pneumonia is a serious disease causing a lot of fatalities both in adults and children worldwide. Detection of pneumonia is usually done by examining Chest X-rays and blood tests in presence of highly trained specialists. With the advancement in Computer-aided diagnosis a trend of improving diagnostic accuracy can be seen. In this paper we have implemented various Machine Learning and Deep Learning techniques for binary classification of X-ray images into normal or pneumonia and also performed comparative analysis of the implemented techniques by noting metrics viz. accuracy and confusion matrix for each technique. The dataset used is chest X-ray Images (Pneumonia) by Paul Mooney which is also publicly available on the Kaggle website containing a total of 5856 X-ray images (jpeg). Experimental results demonstrated that a higher accuracy was obtained for Deep Learning techniques as compared to other techniques implemented.