Design a Computer Aided Diagnosis System for Automatic Detection of Pulmonary Nodules in Lung CT Scan Images

Pulmonary nodules are the first stage of lung cancer and with early detection of them can be used to treat the disease more effectively. Detecting nodules from a CT scan image by a physician is challenging. After pre-processing and segmentation of lung images into areas with dimensions of 64 × 64 pixels andlabeling them, the classification of lung images into two categories of nodules and non-nodules was implemented with the help of training 3 networks DenseNet, InceptionV3 and Xception, and then the average prediction of them were evaluated to classify the images.The CT images were selected from the LIDC IDRI database, and nodule regions in these images were identified using the pylidc program. The aim of this study was to design a computer-aided diagnosis (CAD) system to automate the classification of lung areas into nodules and non-nodules, which can help physiciansand radiologists in more accurate and early diagnosis of lung cancer. In the proposed system, with the evaluations made on the performance of the three networks, the best value of accuracy and sensitivity is obtained in average predicting of the three networks. Higher sensitivity in this case indicates a smalleramount of negative error, which plays an effective role in more accurate diagnosis of cancer and treatment measures

Design a Computer Aided Diagnosis System for Automatic Detection of Pulmonary Nodules in Lung CT Scan Images Keywords:

DenseNet , InceptionV3 , Xception , classification , Computer Aided Diagnosis