An Efficient Deep Learning Approach for Pneumonia Detection

Abstract

Pneumonia is a form of acute respiratory tract infection(ARTI) that affects the lungs which is caused by bacteria, viruses or fungi. Pneumonia is the leading disease that occurs more in children of age below 5 years. Every year almost 7,00,000 children are victimised for this disease. Hence, the accurate diagnosis of such a disease is of high importance. So, the expert radiologists role is crucial to identify the disease through chest x-ray images. But, in certain situations the doctors fail or there are no expert radiologists available in developing countries. There is a requirement of a software based support system to detect Pneumonia using Chest X-ray images to provide early diagnosis for the infected person. So, the aim of this project is to develop a software system to detect the disease Pneumonia using Chest x-ray images. This is achieved by using multiple convolutional neural network layers where the chest x-ray images are tested, trained and validated. The inception v3 model is a CNN which is 48 layers deep and is used to extract the high level features from the images. The test result obtained showed that the software is classifying the infected and non infected images. As a result, this project has reached the accuracy of 85% in detecting the disease from chest x-ray images.