Comparative analysis of deep learning methods for accurate pneumothorax diagnosis

抄録

Pneumothorax is a medical condition characterized by the accumulation of air or gas in the pleural space, causing the lung to collapse. It can occur spontaneously, due to trauma, or because of underlying lung diseases such as chronic obstructive pulmonary disease. Symptoms include sudden chest pain and difficulty breathing, and in severe cases, it can lead to respiratory failure. Early detection and treatment are critical to prevent complications. In this study, we conduct a comparative analysis on different deep learning architectures for pneumothorax diagnosis from chest radiographs. We used CANDID-PTX dataset, which consists of 19,237 cases. These models’ performance is assessed based on metrics like accuracy, precision and recall, demonstrating its potential to aid radiologists in accurate and efficient pneumothorax detection, thereby improving clinical outcomes.