Logistic Regression Analysis Of HRCT Characteristics Of Benign And Malignant Pulmonary Ground-Glass Opacity

Background Lung cancer was one of the most common types of solid malignant carcinoma and exhibited high mortality.Lung cancer-related deaths accounted for 30%of the total deaths caused by all kinds of cancer.If the early stage of cancer can be clearly diagnosed,it can greatly reduce the mortality of lung cancer and improve the prognosis of patients.According to relevant reports,the lung cancer death risk of the high-risk group could be reduced by 20% with low-dose spiral computed tomography(CT)examination annually.Nowadays,with the continuous improvement of CT technology,such as the application of high resolution thin-slice CT(HRCT)and the improvement of various aspects of CT machine performance,more and more lung ground-glass shadows(GGO)have been detected,which has attracted more and more attention.Objective 1.To evaluate the diagnostic value of high resolution CT imaging signs in differentiating benign from malignant GGO;2.Logistic regression analysis was conducted to analyze the risk factors related to malignant GGO;3.To improve the accuracy of imaging diagnosis and reduce the excessive worry and unnecessary follow-up times caused by the detection of occasional pulmonary nodules,so as to provide correct guidance for radiologists and clinicians to better manage these patients.Methods A total of 96 cases of GGO that had been confirmed by operative pathology or biopsypathology were selected from January,2016 to December,2017 in the Second Affiliated Hospital of Anhui Medical University.According to the results of operationand pathology,the patients were divided into benign GGO group(44 cases)and malignant GGO group(52 cases).A retrospective analysis of theclinical and imaging data,including the size of the lesion,the shape,the edge,the interface,the internal density of the lesion,the internal structure and the adjacent structure;Firstly,univariate analysis was conducted to screen out the image features that were significant for the identification of benign and malignant GGO.The univariate analysis results were used as independent variables and pathological results(benign/malignant)as dependent variables to establish a multivariate logistic regression model.ROC curve was drawn and youden index was calculated to evaluate the prediction effect of the model.Results The Chi-square test showed that lobulation,spiculation,interface sign(ill-def ined,well-definedand smooth,and well-defined but coarse),bubbly lucencies,density signs were statistically significantfor differentiating benign and malignant GGO(P<0.05).The sensitivity of each sign in the diagnosis of malignant GGO was 71.15%,59.62%,15.38%,32.69%,51.92%,36.54%,40.38%,respectively;The spec ificity was 86.36%,79.55%,36.36%,75.00%,88.64%,93.18% and 29.55%,resp ectively;The multivariate analysis showed that lobulation,well-defined but coarse i nterface and spiculation were associated with malignant GGO(P<0.05).The regre ssion equation Logistic(P)=-2.494+1.987×lobulation +(2.664×well-defined an d smooth +0.759×well-defined but coarse)+1.444×spiculation.The above logistic regression equation was fitted by Hosmer and Lemeshow Test,χ2=1.704,γ=7,P=0.974>0.05.It shows that the fitting of the equation has statistical significan ce and can correctly classify 85.40% of the research objects.That is to say,the s ensitivity P of the model represents the probability of malignant diagnosis.If P is more than 0.5,the prediction will be malignant.If P is less than 0.5,the pre diction will be benign.The accuracy of the prediction will be as high as 85.40%.The specificity,sensitivity,positive and negative predictive values are 81.80%,88.50%,85.19% and 85.71%,respectively.ROC curve was drawn based on t he predicted values of GGO probability of good and malignant in the model.The area under the curve was 0.907(P<0.001),the 95% confidence interval CI wa s(0.845,0.968),the Youden index was 0.703,the corresponding sensitivity was 88.50%,and the specificity was 81.80%.Conclusion 1.Lobulation,spiculation,interface sign(ill-defined,well-defined and sm ooth,and well-defined but coarse),bubbly lucencies,density signs were statistically significant for differentiating benign and malignant GGO;2.GGO with the sign of lobulation,well-defined but coarse interface and spiculation is the main risk factor for the development of malignant GGO;3.Multivariate Logistic regression analysis model based on HRCT image features has better prediction effect.