CXR-based lung disease classification using convolutional neural network / Jannie Fleur V. Oraño

Thesis -- Cebu Institute of Technology University, College of Computer Studies, October 2019

Includes bibliographical references.

Lung disease like effusion, pneumothorax, atelectasis and tuberculosis are some of the most severe and prevailing health problems in people’s life. A large percentage of the human population around the world is diagnosed annually with lung disease affecting adults, teens, children, smokers, and even non-smokers making it be considered as the leading cause of death and disability worldwide. Early and accurate diagnosis of chest diseases in mandatory and needed for timely and successful treatment, prevents further complications and a higher likelihood of survival. This study demonstrated the feasibility of classifying lung diseases in chest X-rays using conventional and deep learning approaches. With 8,125 sample images, the neural network model managed to achieve 82.53% accuracy in the classification. This accuracy rate indicates the generated model can significantly automate the differential diagnosis of lung disease using chest radiograph. The developed applications, when used as a second opinion, can help assists radiologist and doctors in their diagnosis and decision making. However, this accuracy can be further improved if more and balanced training images will be utilized.