A lung cancer risk warning model based on tongue images

• DOI: 10.3389/fphys.2023.1154294
• Publication Date: 2023-06-01T05:04:59Z
• Authors: Yulin Shi; Dandan Guo; Yi Chun; Jiayi Liu; Lingshuang Liu; Liping Tu; Jiatuo Xu
• Keywords: lung cancer; machine learning; benign pulmonary nodule; Physiology; QP1-981; risk warning model; tongue image; 3. Good health

Objective: To investigate the tongue image features of patients with lung cancer and benign pulmonary nodules to construct a risk warning model using machine learning methods. Methods: From July 2020 March 2022, we collected 862 participants including 263 cancer, 292 nodules, 307 healthy subjects. The TFDA-1 digital diagnosis instrument was used capture images, feature extraction technology obtain index images. statistical characteristics correlations were analyzed, six algorithms build prediction models based on different data sets. Results: Patients had than cancer. Among data, random forest performed best, accuracy 0.679 ± 0.048 an AUC 0.752 0.051. for logistic regression, decision tree, SVM, forest, neural network, naïve bayes both baseline 0.760 0.021, 0.764 0.043, 0.774 0.029, 0.770 0.050, 0.762 0.059, 0.709 0.052, respectively, while corresponding AUCs 0.808 0.031, 0.033, 0.755 0.027, 0.804 0.777 0.044, 0.795 0.039, respectively. Conclusion: under guidance traditional Chinese medicine diagnostic theory useful. performance built superior that only or data. Adding objective can significantly improve efficacy models.