Classification of Chest Diseases Using Convolutional Neural Network

Abstract

Chest radiography (Chest X-ray) is the most common image taken in medical field for diagnosis of any condition that might be affecting chest or nearby area. Due to a shortage of proficient radiologists, application of this technique has been limited. To overcome this problem, we have come with a solution where we are designing a computer-aided diagnosis for chest X-ray using convolutional neural networks (CNN). CNN is a supervised deep learning that has achieved a great recognition in medical field for automatic and adaptive learning through its various layers. It has proven to be much faster and proficient than human radiologists in diagnosing medical conditions. To make a precise resultant classifying neural network, a large amount of labeled dataset is required alternatively a pretrained CNN using a large labeled dataset can also be used with some sufficient fine-tuning. In this study, by using 108,948 frontal view X-ray images of 32,717 unique patients each diagnosed with any one of the 14 lung diseases, we will build deep learning neural network using CNN from scratch which will analyze the chest X-ray and will diagnose the disease and classify within the 15 classes, i.e., 14 pre-known lung diseases and a healthy pair of lungs.