Spontaneous Pneumothorax Detection in Chest X‑rays using Convolutional Neural Networks

This paper evaluates the application of Convolutional Neural Networks (CNNs) for detecting spontaneous pneumothorax in chest X-rays. For this study, we randomly selected 2000 chest X-ray images with pneumothorax from the publicly available National Institutes of Health (NIH) database, which were subsequently allocated to training and testing datasets. To enhance the model's generalization, the images were preprocessed by normalizing, resizing, and augmenting the dataset. The proposed CNN model includes multiple convolutional layers that extract low-level features, followed by max-pooling layers for progressive dimensionality reduction. To combine the extracted features and make predictions, fully connected layers are employed, and a hyperbolic tangent (tanh) activation function is used in the output layer for binary classification. The Adam optimizer is utilized to train the model, and its performance is assessed using standard performance metrics, including accuracy, precision, sensitivity, specificity, and specificity. The results show that the developed CNN-based method outperforms conventional approaches based on machine learning, such as Decision Tree, Random Forest, and SVM.