Detection of Pleural Cavity Objects in X-Ray Images Pneumothorax Lungs Using Morphology and Convolutional Neural Networks Methods

Image processing procedures are a solution in developing chest x-ray technology, starting with the image segmentation process as a preprocessing stage, namely separating image objects from the original background. Followed by the convolutional neural network (CNN) method, where the data is trained in a learning program, allowing the program to detect objects and obtain detection results that match the expected research results. Spontaneous pneumothorax (SP) is a type of collection of air in the pleural cavity that develops without trauma. Various lung diseases, pulmonary interventions, use of mechanical ventilation, or thoracic trauma can cause pneumothorax. The diagnosis of pneumothorax is made based on the history, physical examination and chest x-ray. The proposed method is able to detect objects indicated by pneumothorax using morphology involving dilation and erosion processes. The detection process involves determining the area using region features. For patient 1, 19 pneumothorax objects were found on the right side, while not a single object was found on the left side. This process is expected to increase the level of accuracy of segmentation of the pneumothorax region. Through the use of image processing and CNN methods, the extraction process can be developed to obtain chest x-ray results with clearer, sharper images and a high level of accuracy, thereby speeding up the decision- making process for patients. In this study, the CNN algorithm used was able to classify parts of the lung that were injured or infected with an accuracy rate of 96.77%.