Artificial Intelligence Combined With Imaging Features To Predict Benign And Malignant Lung Nodules And Degree Of Infiltration

Aim:To explore the clinical application value of artificial intelligence-assisted diagnosis system in the diagnosis of benign and malignant pulmonary nodules,and further explore the possibility of predicting the degree of lung cancer invasion by combining CT image features,and understand the factors related to the aggressiveness of lung cancer by building a prediction model,which will provide clinicians with Predicting the degree of lung cancer infiltration provides an applicable nomogram to guide clinicians to use the artificial intelligence-assisted diagnosis system to accurately judge the benign and malignant pulmonary nodules and the degree of infiltration.Materials and Methods:Clinical data and images of lung nodules were extracted from our electronic patient case system and the artificial intelligence-aided diagnosis system,and the exact pathological data and probability of malignancy were used to draw the receiveroperating characteristic curve(ROC)for the diagnosis of benign and malignant lung nodules by the artificial intelligence-aided diagnosis system.The diagnostic performance of the CAD system was evaluated.At the same time,the independent risk factors associated with the degree of lung cancer infiltration were screened using univariate analysis and multi-factor logistic regression analysis of clinical data and imaging data,and the prediction model was constructed and plotted in columns.The Hosmer-Lemeshow test was used to evaluate the effectiveness of the model.Results:1.A total of 199 eligible postoperative patients with pulmonary nodules were included,including 91 males and 108 females,with an average age of 56.31±11.88years;Among them,there were 49 cases of benign nodules(7 cases of pure groundglass nodules,7 cases of partial solid nodules,35 cases of solid nodules),149 cases of malignant nodules(35 cases of partial solid nodules,74 cases of pure ground-glass nodules,40 cases of solid nodules),and 1 case of precancerous lesions(pure groundglass nodules).2、Based on pathological data and malignant probability,a ROC curve was plotted with an area under the curve of 0.886.The accuracy of an artificial intelligence model for predicting the benign or malignant status of pulmonary nodules was found to be 84.9%,with a sensitivity of 97.3%,a specificity of 48%,a positive predictive value of 84.8%,and a negative predictive value of 85.7%.The optimal cutoff value was determined to be 79.5% for diagnosing malignant pulmonary nodules,with a sensitivity of 74.5% and a specificity of 90.0%.3、The independent risk factors associated with the degree of infiltration of lung cancer were identified as maximum diameter,age,lobar sign,pleural depression sign and vacuole sign by single factor analysis and multi-factor logistic regression analysis;the prediction model was constructed and plotted in columns,and the area under the curve of the ROC curve was calculated as 0.88.The Hosmer-Lemeshow test(H-L test)was used to test that the model fitted well.Conclusions:1.The critical value of the probability of malignancy for the prediction of benign and malignant lung nodules by the application of artificial intelligence-aided diagnosis system was 79.5%.2.Maximum diameter,age,lobar sign,pleural depression sign and vacuolar sign were independent risk factors for predicting the degree of lung cancer infiltration.