Early Detection of Pneumothorax Using EfficientNet B0 Transfer Learning Model

Pneumothorax is a condition which requires immediate care. It is a commonly encountered state that experienced radiologists and clinicians in diverse clinical specialties may evaluate and diagnose using chest radiographs. It has been observed to have an impact on mortality rates. In the biomedical industry, the early detection of pneumothorax is required for better treatment of patients. This research paper proposes a DL neural network model to identify region of pneumothorax in chest radiographs. In this research work EfficientNet B0, a transfer learning model, has been implemented for the early analysis of pneumothorax in chest radiographs. The dataset consists of 1403 X-ray images divided into train, test, and validation, with 1200 training images, 102 test images, and 101 validation images. The batch size of the model was set to 32, and it has been trained 25 epochs altogether. A learning rate of 0.001 was used. The model has been evaluated regarding performance parameters such as precision, recall, accuracy, and F1-score. The model achieved an accuracy rate of 83%. This automatic method for early detection of pneumothorax can assist the radiologist in better treatment planning.