Correlation analysis of tongue image features between patients with benign lung nodules and lung cancer

Abstract

Abstract Lung nodules are high-risk factors for lung cancer, which often present as lung nodules in the early stages of lung cancer and have no obvious clinical symptoms. Identifying the benign or malignant nature of lung nodules can not only avoid over-diagnosis and over-treatment of benign pulmonary nodules but also save medical resources and reduce patients' psychological anxiety. Early detection, diagnosis, and treatment of lung cancer are crucial for reducing the mortality rate of lung cancer patients. In this study, we attempted to analyze the correlation of the tongue image features of benign lung nodules and lung cancer, using a bivariate correlation analysis that used Spearman correlation for variables that did not have a normal distribution and Pearson correlation for those that did, and a canonical correlation analysis for multivariate correlation between the two sets, mining the differences of the two groups' tongue image features in order to serve as a reference for classification models based on machine learning techniques. The results showed that there was a different association of diagnostic features on tongue images between patients with benign lung nodules and lung cancer. In the lung cancer group, the features of the tongue body and tongue coating had a stronger canonical association. In the group of benign lung nodules, the correlation coefficients for the first and second pairs of canonical variables were 0.918 and 0.817 (P < 0.01), respectively, whereas in the group of lung cancer, the correlation coefficients for the first and second pairs of canonical variables were 0.940 and 0.822 (P < 0.01). The objective data of tongue image provide a new methodological reference for the risk warning of lung cancer, it can be helpful in the risk warning of lung cancer to some extent.