부산대학교 도서관

This paper presents an approach for classifying pleural effusion using convolutional neural networks. The study utilizes a dataset of chest X-rays from the Chest X-Ray 14 database, comprising both normal and pleural effusion images. To enhance the model’s robustness and generalization, the Data Augmentation technique is employed for each class. The obtained results are highly promising. In Experiment I, the model achieves remarkable training and test success rates of 98.21% and 97.70%, respectively. In Experiment II, where an extensive data augmentation technique is applied, the model yields training, validation, and test rates of 78.66%, 87%, and 91%, respectively. These outcomes indicate the potential of the proposed classification model in facilitating the automated detection of pleural effusion and other lung diseases. This research makes a significant contribution to the advancement of computer-aided medical diagnosis, particularly by leveraging convolutional neural networks for chest X-ray image analysis.