Hanieh Alimiri Dehbaghi - Department of Medical Physics, University of medical science, Kermanshah, Iran
Karim Khoshgard - Department of Medical Physics, University of medical science, Kermanshah, Iran
Hamid Sharini - Department of Biomedical Engineering, University of medical science, Kermanshah, Iran
Samira Jafari Khairabadi - Student Research Committee, University of medical science, Kermanshah, Iran

Imaging in the emergency department is a time-dependent process. The development of artificial intelligence models can improve the performance of any diagnostic system by minimizing diagnostic errors and increasing the speed of image interpretation by radiologists. The purpose of this study is to create an intelligent model for pneumothorax detection in CT scan images using radiomics features and implementing machine learning algorithms. The data used in this study was extracted from the files of ۱۷۵ patients suspected of pneumothorax. The collected images were pre-processed in the Matlab software. Then the machine learning algorithms including Gradient Tree Boosting (GBM), eXtreme Gradient Boosting (XGBOOST) and Light GBM (LGBM) were used to classify the images. Various evaluation criteria such as precision, accuracy, specificity, sensitivity, F۱ score, the area under the ROC curve and misclassification were calculated to evaluate the performance of these models. According to the calculated evaluation criteria, for the Light GBM model, the accuracy, precision, specificity and F۱ scores were ۰.۹۸۹۷۹, ۰.۹۹۵۵۹, ۰.۹۸۴۳۵ and ۰.۹۹۴۳۰, respectively. These findings indicate the better performance of the Light GBM model compared to other models. The Light GBM model with a sensitivity value of ۰.۹۹۷۶۳ had the best performance among these algorithms. The obtained results showed that the machine learning algorithms used in this research can accurately identify healthy and pneumothorax images and thus facilitate and accelerate the process of diagnosis and treatment of this disease.

“Pneumothorax”, “Radiomics”, “CT scan”, ”Machine Learning”, “Artifitial Intelligence”