Prediction Study Of Malignancy In Patients With Lung Neoplasm Based On Mathematical Medical Model

Background:With the development of imaging examination level and the popularization of people’s awareness of prevention and screening, more and more lung neoplasm could be found. Therefore, it is very important to distinguish benign neoplasm from malignancy and to take appropriate clinical intervention timely. At present, the accuracy of the assessment in tumor characteristics is greatly determined by the clinical experience of doctors and the level of diagnosis and treatment, which makes misdiagnosis and excessive intervention become inevitable. Hence the mathematical prediction model that overcomes the defects of empirical medicine emerges as the times require, and the follow-up and treatment of lung neoplasm could be much more objectively guided by quantified malignancy probability.Purpose:According to the clinical characteristics and some imaging features of collected patients with pulmonary neoplasm, we analyzed independent risk factors of lung malignant neoplasms, evaluated and compared the accuracy of two different kinds of foreign mathematical models in predicting the characteristics of lung neoplasms,and finally illustrated their value of clinical application in Chinese.Methods:The clinical features and imaging data of patients with pathological confirmed lung neoplasms were collected from May 2013 to September 2015 in Sir Run Run Shaw Hospital. Their distribution differences between patients with benign neoplasms and patients with malignant neoplasms were analyzed retrospectively. The independent risk factors of benign and malignant lung neoplasms were analyzed by statistical methods including One-Way ANOVA and chi square test, the accuracy of Mayo Clinic models were compared with that of Veterans Affairs models by calculating area under the ROC curve, in order to explore the prediction accuracy of two models for solitary pulmonary nodules(SPN), lung masses with diameter greater than 30mm and metastatic lung cancer.Results:Univariate analysis showed that statistically significant differences were found in age of diagnosis and the diameter of pulmonary neoplasms among benign lung neoplasm group, primary malignant lung neoplasm group and metastatic lung neoplasm group (P<0.05).However, no significant difference was found in years of quitting smoking among the above three groups.And the further inter-group analysis showed that statistic significant difference was only found between benign lung neoplasm group and primary malignant lung neoplasm group in age of diagnosis.The chi square test showed that not only the position and shape of neoplasms but also the diameter of neoplasms, the medical history of cancer were significant different among the above three groups (P< 0.05). In contrast, no significant differences were found in the gender and smoking history.The area under the ROC curve of Mayo Clinic model is 0.653 and that of VA model is 0.554, indicating that both of two models have a good prediction accuracy, but the Mayo Clinic model seems better in predictive value.Conclusion:According to the area under the ROC curve, both of Mayo Clinic model and VA model have a good prediction accuracy. They can be clinical used for primary screening and evaluation of lung neoplasm’s property. This may offer a great reference for clinician. However, the Mayo Clinic model has a better predictive value.The prediction efficiency of mathematical models have no statistic significant difference between SPN and lung mass.That is,the above two models could be used to predict the malignant probability not only of SPN but also of lung mass.