CNN-Based Detection of Pneumothorax and Effusion in Chest X-Rays

Abstract

This project presents a convolutional neural network (CNN)-based classification model for thoracic diseases using chest Xrays from the NIH ChestX-ray14 dataset. Following initial challenges with 20 disease classes, the study was narrowed to Pneumothorax and Effusion, as these conditions exhibit clearer radiographic patterns. The final model, implemented in MATLAB, achieved an accuracy of 75% for Pneumothorax and 65% for Effusion. These results highlight the potential of AI in radiology while emphasizing the importance of expert input, artifact-free datasets, and rigorous clinical validation.