Habib Rostami

Pneumothrax can be a medical emergency due to lung collaps and respiratory failury. Pneumothorax is usually diagnosed on chest X-rays. Howere, treatment depends on timely radiographic examination. The purpose of this thesis is to present a model based on convolutional neural networks to detect the presence or absence of pneumothorax disease in chest X-ray images. For this purpose, the thesis used a CheXpert dataset containing 224316 chest radiographs of 65240 patients with radiological reports. First, we used a model that pre-trained on pneumonia disease data set of children to extract the feature. At this point, it attempts to select a set of key features to distinguish between normal and abnormal images. Next, a vector of the features of each image is applied to the network structure, and finally the network classifies the data into two classes, one and zero, which indicate the presence or absence of disease, respectively. The classification accuracy of the proposed method reached 97, which is more efficient than the existing methods