The Study Of Gray-level Co-occurrence Matrix In Computed Tomography Images Of Solitary Pulmonary Nodules

Lung cancer is the most popular reason of cancer death with a 5 years survival rate of 15%.Due to there is no typical symptom in the early stage of lung cancer,most of the patients belong to the middle and late stages of the disease when they have selfconscious symptoms.With the development of computed tomography(CT)technology,such as the improvement of CT device resolution and innovation of reconstruction algorithm,Low-dose CT(LDCT)is widely used in early screening of lung cancer.Solitary pulmonary nodules(SPNs)refers to the pulmonary nodules of round or oval less than 3cm in image.As a common feature of both benign and malignant lunglesions,the detection rate of SPNs in the early lung cancer screening has increased in recent years,especially the ground glass nodules(GGNs).However,because the nodules are small or thin,it is often difficult to distinguish between benign and malignant nodules from morphologic features on the image,so it is easy to cause misdiagnosis and missed diagnosis.Meanwhile,as a new discipline,which relies on deep excavation of image information and extract a large number of quantitative characteristics of medical image data,and then carries out quantitative diagnosis for disease diagnosis and prediction,radiomics has a breakthrough in the diagnosis and identification of tumor compared with the traditional image morphological analysis.As the most widely used texture analysis method in biological imaging,the gray level co-occurrence matrix(GLCM)reflects the comprehensive information of the image in direction,interval,amplitude and speed by calculating the correlation between two points in a certain distance or a certain direction in the image.Based on the current situation and technological development,this study aims to investigate the application of GLCM texture features in the identification between benign and malignant SPNs from CT images.The details are as follows:1.Through the extraction of the texture features of the SPNs,five GLCM texture features such as the energy,contrast,correlation,inverse difference moment(idm)and entropy were selected for analysis of 64 cases(32 cases were benign and 32 were malignant).The optimal threshold of the texture features with diagnostic efficacy was analyzed by the receiver operating characteristic curve(ROC).2.To further analyze whether the GLCM texture features of the enhanced CT and plain CT images in the above cases were statistically different in benign and malignant SPNs.3.The CT images of 60 solitary pulmonary nodules were retrospectively analyzed(30low-dose CT images and 30 conventionaldose CT images)to investigate if there is a statistical differences in GLCM texture features between low-dose CT andconventionaldose CT.