Improving Emphysema Diagnosis Accuracy through Deep Convolutional Neural Networks: Analysis of Chest X-ray and Chest CT Scan Images

Abstract

Chronic obstructive pulmonary disease (COPD) is a complex respiratory condition involving various lung parenchymal anomalies. The major aim of this research is to enhance the diagnostic accuracy of Emphysema, a type of COPD, by leveraging Convolutional Neural Networks (CNNs) applied to Chest CT scan and Chest X-ray (CXR) images. The proposed CNN method consists of nine convolutional layers, each followed by batch normalization (BN), ReLU activation, and eight max-pooling layers, along with two strategically placed dropout layers to prevent overfitting. The model further includes fully connected, SoftMax, and classification layers, encompassing 15.7 million trainable parameters. The primary objective is to minimize false negatives, where Emphysema patients are misclassified as healthy while also maintaining low false positive and false negative rates, crucial for healthcare applications. On the benchmark dataset, the model achieved a validation accuracy of 97.01%. When applied to the real-time database from SRM Medical College Hospital & Research Centre, the model’s performance improved significantly, reaching a validation accuracy of 99.05%. The study demonstrates the effectiveness of deep CNNs in enhancing Emphysema diagnosis accuracy by analyzing Chest CT and X-ray images. The model’s ability to reduce false negatives is particularly impactful for early COPD detection, offering significant potential for improving patient outcomes in healthcare settings.