Automatic Benign and Malignant Classification of Pulmonary Nodules in Thoracic Computed Tomography Images based on CNN Algorithm

Abstract

The majority of the time, cancer symptoms only appear in the final stages of the disease, but with the use of advanced technology in which computer-aided systems, we can identify cancer at an early stage, then we may be able to treat it. These automated detection systems use a variety of machine learning techniques to identify lung cancer in its early stages. This type of automatic diagnosis of lung cancer is detected in computed tomography image modalities using convolution neural network. Compared to MRI and X-ray, CT pictures have the advantage of having less noise interference. Median filtering is applied to de-noising the CT scans to enhance the image quality. These pre-processed images are then fed through the Re-Lu layer CNN architecture. This architecture's several layers handle the feature extraction and categorization work. Different low-level and distinct high-level characteristics are extracted during the feature extraction phase. The classification layer is in charge of determining whether the presented image contains a benign or malignant, and it can able to classify the severity of malignance based on the infected area. To carry out the proposed model, CT image datasets are taken from Imaging data base Resource Initiative (IDRI) and Lung image Database consortium (LIDC) datasets. According to the findings, the suggested proposed system works with optimal results. The comparison demonstrates that the suggested proposed system outperforms the current cutting-edge technologies. The suggested proposed system will be beneficial for health care systems and medical diagnosis research. The proposed methodology is verified on MATLAB software.