Evaluate The Ability Of The "Artificial Intelligence Diagnosis And Treatment System" To Predict The Pathology And Infiltration Of Lung Nodules

Objective: To retrospectively analyze the predictive efficacy of the "malignant probability",average density,solid proportion,volume doubling time and other indicators of lung nodules calculated by the "intelligent diagnosis and treatment system" for the benign,malignant and infiltration conditions of nodules.Methods: A total of 147 patients with pulmonary nodules who were admitted to the First Affiliated Hospital of Xiamen University from February 2019 to December 2020,with a total of 151 nodules,57 male nodules(37.7%)and 94 female nodules(62.3%),131 cases(86.8%)were classified as malignant nodules(86.8%)and 20 cases(13.2%)were benign nodules according to postoperative routine pathological results;96 cases of malignant nodules and 13 cases of benign nodules met the preoperative There were2 chest CT scans during the year,which met the requirement to calculate the nodule volume doubling time.Collect basic clinical data of patients: gender,age,smoking history,tumor family history,and record the average density of nodules,solid proportions,"malignant probability",and volume doubling time.Calculate the predictive ability of the above indicators for the benign and malignant pulmonary nodules and the degree of infiltration.Results: There was no significant difference between the malignant nodule group and the benign nodule group in terms of gender,age,tumor family history,smoking history,and nodule diameter(p>0.05).There was a statistically significant difference in "malignant probability" between the two groups Significance,and the "malignant probability" of the malignant nodule group is significantly higher than the "malignant probability" of the benign nodule group(83.5% vs.30.1%,P<0.01).Using the "malignant probability" to predict the benign and malignant pulmonary nodules The ROC curve analysis result is(AUC=0.963,the cut-off value is 65.10%);analysis of invasive adenocarcinoma group and non-invasive adenocarcinoma group,there is no statistical difference in gender,age,tumor family history and smoking history between the two groups Significance(p>0.05).The maximum diameter,malignant probability,average density,and solid proportions of the nodule in the invasive adenocarcinoma group were significantly greater than those in the non-invasive adenocarcinoma group(P<0.01).The "malignant probability" of the nodule was used respectively.The results of ROC curve for predicting the degree of pulmonary nodule infiltration with solid proportion,average density and diameter are(AUC=0.692,cut-off value 91.2%;AUC=0.847,cut-off value 26.5%;AUC=0.818,cut-off value-449HU;AUC=0.840,cut-off value of 0.95cm);Analysis of the ROC curve results of nodule doubling time after screening to predict the benign and malignant and infiltration degree of lung nodules is(AUC=0.929,cut-off value is 776.0d;AUC=0.828,the cutoff value is 524d).Conclusion: This study found that the "malignant probability" and doubling time of nodules provided by the "intelligent diagnosis and treatment system" can both accurately predict the benign and malignant pulmonary nodules,but cannot distinguish the degree of pulmonary nodule infiltration.In addition,the three indicators of average density,solid proportion and maximum diameter of lung nodules are not very accurate in predicting the degree of lung nodule infiltration;in future work,we can further try to train the "intelligent diagnosis and treatment system" for The ability to identify the degree of pulmonary nodule infiltration allows it to not only identify benign and malignant nodules,but also to help physicians judge the infiltration of pulmonary nodules.