| Objectives 1 Study on the differential diagnosis of texture analysis in benign and malignant solitary pulmonary nodules in CT plain scan.2 To analyze the enhanced CT texture features in the diagnosis of benign and malignant lung nodules.3 The relevant study of CT texture feature analysis for differential diagnosis of lung cancer pathological classification.Methods 1 Select and analyze the CT images of 176 patients with solitary pulmonary nodules(SPN)from January 2016 to December 2018(Among them,72 were benign nodules and 104 were malignant nodules).A region of interest(ROI)is selected in the SPN CT images,and specific texture features are extracted there for analysis.Texture parameters are obtained,including Energy,Entropy,Sum Entropy,Difference Entropy,Contrast,and Correlation.In the texture parameters comparison between benign and malignant pulmonary nodules,the independent sample t test was used to establish the receiver operating characteristic curve(ROC),and the area under the ROC curve(AUC)was calculated to evaluate the sensitivity and specificity of each texture parameter to the benign and malignant diagnosis for SPN.2 Furtherly analyze whether statistical difference exists in the SPN benign and malignant identification with enhanced CT image texture features,and 60 SPN cases are selected in the above cases(30 fall within benign,the other 30 fall into malignant).3 To analyze and study 104 cases of malignant nodules,to study the role of CT texture feature analysis in differential diagnosis of lung cancer pathological classification.Results 1 In the analysis of lung benign and malignant nodules in plain CT images,CT texture feature parameters: Difference Entropy,Correlation were not statistically significant(P> 0.05);Energy,Entropy,Sum Entropy,and Contrast were statistically different(P< 0.05),the AUC were 0.713,0.796,0.774,and 0.595,respectively;while the sensitivity was 0.695,0.727,0.773,and 0.667,respectively;the specificities were 0.604,0.792,0.708,and 0.441,respectively.When the three characteristic parameters(Energy,Entropy and Sum Entropy)are combined,the AUC was 0.916,the sensitivity and specificity reach 0.895 and 0.813,respectively.2 CT-enhanced scan images were used to identify benign and malignant nodules.There was no statistical difference in Energy,Entropy,Sum Entropy,Difference Entropy and Correlation between arterial and venous periods(P>0.05),while Contrast parameter was statistically significant(P<0.05).The arterial AUC was 0.711,while the respective sensitivity and specificity were 0.807,0.736;the venous AUC 0.672,while the respective sensitivity and specificity were 0.698,0.563.3 Comparing the different pathological types of lung cancer(squamous cell carcinoma,adenocarcinoma,small cell carcinoma),the Energy,Entropy,Sum Entropy,Difference Entropy and Contrast characteristic parameters were statistically different(P<0.001),and the Correlation was not statistically significant(P>0.05).When the squamous cell carcinoma and adenocarcinoma were compared,the AUC of Contrast,Entropy,Difference Entropy,Sum Entropy and Energy were 0.734,0.899,0.822,0.903,and 0.942,respectively.When compared with small cell carcinoma and non-small cell carcinoma,contrast and entropy were compared.The AUC of Entropy,Difference Entropy,Sum Entropy and Energy were 0.771,0.706,0.775 and 0.816,respectively.Conclusion CT texture feature analysis can provide texture features that can not be observed by naked eyes,quantitatively evaluate solitary pulmonary nodules,and provide more internal structure information,which is helpful to differentiate benign from malignant solitary pulmonary nodules.CT texture feature analysis can provide some help in differentiating squamous cell carcinoma from adenocarcinoma,small cell carcinoma from non-small cell lung cancer by specific texture parameters.Figure 3;Table 9;Reference 91... |
PDF Full Text Request