A Study On Differentiation Of Benign And Malignant Pulmonary Nodules Based On Radiomics Combine With Mayo Clinc Model

Objective:Prediction model is an important auxiliary diagnostic tool in the early diagnosis of lung cancer.Mayo Clinic Model(Mayo Clinic Model)is the earliest lung cancer risk prediction model applied in clinic.Its reliability has been verified by real world studies in recent years,and it has been written into many lung cancer diagnosis and treatment guidelines.Extensive external verification has been conducted to compare the accuracy of Mayo model with that of physicians in predicting benign and malignant pulmonary nodules.It is found that compared with clinicians’ diagnosis of pulmonary nodules,Mayo model can improve the accuracy of predicting benign and malignant pulmonary nodules.However,the Mayo model also has its shortcomings.When the Mayo model was established,12% of pulmonary nodules had no clear pathological diagnosis(according to the 2-year follow-up,if no significant changes in pulmonary nodules were determined to be benign),so the prediction efficiency of the model would be affected to some extent.In order to improve the accuracy of the prediction model,this study built a clinical-imaging logistic regression model to predict benign and malignant pulmonary nodules based on the combined classical model of imaging omics(Mayo model),and analyzed the prediction efficiency of imaging omics model,Mayo model and combined model,so as to select the prediction model with the best predictive efficiency,so as to guide the clinical practice and promote the progress of precision medicine.Methods:This study was a retrospective cohort study.Chest CT images and clinical data of 102 patients with lung nodules with definite pathological diagnosis were collected from our hospital(Affiliated Hospital of Qingdao University),including: Patients were divided into benign group(42 cases)and malignant group(60 cases)by age,smoking history,history of external thoracic malignancy,nodule diameter,burr sign,and nodule location.Imaging Biomarker Explorer(Imaging Biomarker Explorer,MDACC,USA,version v1.0β)was applied to delineate the region of interest(ROI)and extract the image omics data.Manually delineate the ROI layer by layer along the nodule boundary on the chest CT(lung window image)of the patient,and finally generate the three-dimensional Volume of Interest(VOI),and extract its image omics features.The Pearson correlation coefficient(PCCs)and corrplot toolkit were used to screen the image omics features.Independent sample t test or MannWhitney U test were used to screen the image omics data after dimensionality reduction.To verify whether there is a significant difference between benign and malignant pulmonary nodules in the imaging omics data.The image omics data after statistical verification can be applied to Lasso regression model(Least absolute shrinkage and selection operator)and glmnet toolkit to construct the image omics model.The image omics model,Mayo model and joint model were established respectively.The Area Under Curve(AUC)was used to analyze the prediction efficiency of the image omics model,Mayo model and joint model,and the prediction model with the best prediction efficiency was selected.IBEX software was used to extract 30 image omics feature parameters.After removing repeated image omics features,7imageological features were selected after dimensionality reduction,including X5 percentile,Long Run Low Gray Levela,Voxel Size,Sphericity,Mass,Kurtosis and Texture Strength)was used to build the image omics model.Results: The AUC of imaging omics model,Mayo model and combined model in predicting benign and malignant pulmonary nodular were 0.619,0.788 and 0.830,respectively.Conclusions:1.The Sphericity(sphericity)and Mass(compact)characteristics in the image omics data extracted and processed by IBEX software have statistical significance in the difference between benign and malignant pulmonary nodules.2.The AUCs of the three models were 0.788(95%CI: 0.696-0.880)(P<0.001),0.619(95%CI: 0.501-0.737)(P=0.619),and 0.830(95%CI: 0.96-0.880)for the combined model.0.749-0.912)(P<0.001),the results showed that the prediction model based on the combination of imaging omics and classical model could better predict the benign and malignant pulmonary nodules.Among the three prediction models,the combination model had the best prediction efficiency,which could provide reference for the clinical diagnosis of benign and malignant pulmonary nodules.