Medical Radiology Image processing for Pneumonia Detection Using Convolutional Neural Network

Abstract

With the advent of the digital age, humans have created many ways of assistance in menial and time-consuming tasks. Machine learning and Neural Networks provided a way for us to convert large, assorted data into meaningful patterns and predictions. One of the most imperative applications of this technology is Medical Image processing for disease classification and segmentation. In this paper, we attempt to identify Pneumonia in Chest X-Rays, an infection that causes inflammation in the lungs. Bacteria such as Streptococcus pneumonia, among other viruses and fungi cause pneumonia. Due to this reason, the fluids may get filled in the air sacs which causes cough, fever and difficulty in breathing. Convolutional Neural Networks (CNNs) provide an efficient way to detect the disease quickly and have garnered a lot of interest since early diagnosis can prevent the disease from being fatal. Pneumonia claims millions of lives each year and computer-aided classification can help radiologists reduce redundancies and act faster. Our research is focused on using pre-trained CNN models on large datasets as feature extractors such as VGG- 16, InceptionV3, and CheXNet, and then classifying the data into Lung Opacity/Normal, after which we compare and select the best model.