Novel Deep Convolutional Neural Network for Chest X-ray Classification

COVID-19, viral pneumonia, and lung opacity are common lung diseases that are difficult to distinguish between. Furthermore, due to the rapidity with which these diseases develop and their increase in severity, it is of increasing importance that quick, effective, and accurate methods are to be developed for diagnosing and differentiating these diseases at an early stage. Recently, usage of Convolutional Neural Networks have increased in prominence within the medical sphere, including in the diagnosis of diseases. This study provides an accurate and effective method of diagnosing lung diseases and distinguishing them from healthy lungs using chest X-rays. Specifically, a deep convolutional neural network is developed using data containing X-rays of viral pneumonia, healthy lung opacity, and COVID-19. It is created using the Keras/TensorFlow Python library. The standard Sequential base model composed of three 2D convolutional layers with ReLu activation functions is applied. The pipeline achieves a resoundingly high performance of 96.39% accuracy, showing promise in the differential diagnosis of various pulmonary conditions.