Histogram Features Of Magnetic Resonance Imaging In The Differential Diagnosis Of Lung Cancer,Infective Lesions Value And Lung Cancer Subtypes

Objective:1.To evaluate the value of whole-lesion parametric histogram analysis of monoexponential,biexponential and diffusion kurtosis models in the differential diagnosis of lung cancer and focal infectious lesions.2.To explore the differential diagnostic value of whole-lesion parameter histogram analysis in lung squamous cell carcinoma and adenocarcinoma.Methods:This prospective study included 59 patients with solitary pulmonary lesions.T₁-weighted imaging,T₂-weighted imaging,diffusion-weighted imaging（DWI）,intravoxel incoherent motion imag ing（IVIM）and diffusion kurtosis imaging（DKI）were obtained.Medical Imaging Interaction Toolkit（MITK）software was used for image post-processing,and the lesions were delineated layer by layer on ADC images to generate three-dimensional volume of interest（VOI）.The software automatically copied the delineated VOI region to the pseudo-color image of each parameter.The parameters obtained are as follows:（1）DWI derived parameters:Apparent diffusion coefficient（ADC）;（2）IVIM derived parameters:True diffusion coefficient（D）,Pseudo-diffusion coefficient（D*）and Perfusion fraction（f）;（3）DKI derived parameters:Apparent diffusional kurtosis(K_app)and Kurtosis-corrected diffusion coefficient(D_app).The extracted histogram features are the mean,median,standard deviation（SD）,skewness and kurtosis of each parameter.Student’s t-test or Mann-Whitney U test was used to compare the between-group differences in histogram data of lung cancer and infectiou s lesions.Receiver operating characteristics curve（ROC）was used to compare the diagnostic power of parameter histogram data.Univariate and multivariate Logistic regression models were used to study the independent predictors of lung cancer.Subgroup analysis of lung cancer（squamous cell carcinoma and adenocarcinoma of the lung）was further performed using the statistical methods des cribed above.Results:1.Lung cancer group and infectious focus group1.1 Analysis of differences between groups:Of the 59 patients,39 had lung cancer and 20 had infectious lesions.Compared with the infective lesion group,the ADC（mean,median）and D（mean,median）of the lung cancer group were significantly decreased,while the ADC skewness,K_app（mean,median,SD,kurtosis）and D_appskewness were significantly increased（P<0.05）.ADC（SD,kurtosis）,D（SD,skewness,kurtosis）,K_appskewness,D_app（mean,median,SD,kurtosis）,D*and f（mean,median,SD,skewness,kurtosis）were not significantly different between the two groups（P>0.05）.1.2 ROC curve analysis:D median was the most effective in differentiating lung cancer from infectious lesions,with an area under curve（AUC）of 0.777.When the best cut-off value wa s 1.091×10^-3mm²/s,its sensitivity,specificity,positive predictive value and negative predictive value were 69.23%,85.00%,90.00%and 58.62%,respectively.The AUCs of the remaining parameters of interest were:ADC mean 0.673（95%CI:0.536-0.810）,median ADC 0.681（95%CI:0.544-0.818）,ADC skewness 0.678（95%CI:0.517-0.840）,Dmean 0.685（95%CI:0.543-0.826）,K_appmean 0.673（95%CI:0.532-0.814）,K_appmedian 0.686（95%CI:0.546-0.826）,K_appSD 0.715（95%CI:0.586-0.845）,K_appkurtosis 0.711（95%CI:0.577-0.846）and D_appskewness 0.721（95%CI:0.571-0.870）.1.3 Regression analysis:Univariate regression analysis showed that ADC mean（P=0.03）,ADC median（P=0.024）,ADC skewness（P=0.012）,D median（P=0.019）,K_appmean（P=0.036）,K_appmedian（P=0.023）,K_appSD（P=0.015）,K_appkurtosis（P=0.029）,and D_appskewness（P=0.019）were predictors of lung cancer（P<0.05）.After adjusting for confounding factors,multivariate regression analysis showed that ADC skewness（OR=7.061,P=0.019）and D median（OR=0.044,P=0.031）were independent predictors of lung cancer.1.4 Joint model:Independently in the multi-factor Logistic regression analysis,related to the ADC of skewness and D median of risk factors and protective factors,respectively will protect factor D median to joint the bottom and the ADC of skewness analysis,the results showed that the combined model of AUC value of 0.809（95%CI:0.692 0.926）,and take the best cutoff value>0.694,the sensitivity was71.79%,85%.2.Subgroup analysis of lung squamous cell carcinoma and adenocarcinoma2.1 Analysis of differences between groupsThere were 12 cases of squamous cell carcinoma and 27 cases of adenocarcinoma in lung cancer group.Compared with adenocarcinoma group,D kurtosis,K_appkurtosis and D_appSD were significantly increased,while K_appSD and K_appskewness were significantly decreased in squamous cell carcinoma group（P<0.05）.ADC（mean,median,SD,skewness,kurtosis）,D（mean,median,SD,skewness）,K_app（mean,median）,D_app（mean,median,skewness,kurtosis）,D*,and f（mean,median,SD,skewness,kurtosis）were not significantly different between the two groups（P>0.05）.2.2 ROC curve analysis:D kurtosis was the most effective in differentiating squamous cell carcinoma from adenocarcinoma of the lung.The area under the curve was 0.873.When the best cut-off value was 3.367,the sensitivity,specificity,positive predictive value and negative predictive value were83.33%,81.48%,66.70%and 91.70%,respectively.The AUCs of the remaining parameters of interest were:K_appSD 0.731（95%CI:0.573-0.890）,K_appskewness 0.821（95%CI:0.676-0.966）,K_appkurtosis 0.812（95%CI:0.674-0.950）and D_appSD 0.750（95%CI:0.586-0.914）.2.3 Regression analysis:Univariate regression analysis showed that D kurtosis（P=0.004）,K_app skewness（P=0.015）,K_appkurtosis（P=0.036）and D_appSD（P=0.009）were predictors of squamous cell carcinoma（P<0.05）.Multivariate regression analysis showed that D kurtosis（OR=7.245,P=0.004）and D_appSD（OR=2.214,P=0.008）were independent predictors of lung squamous cell carcinoma.2.4 Joint model:Will the multi-factor Logistic regression analysis of independent related D kurtosis and Dapp standard deviation of the analysis,the results showed that the combined model of AUC value of 0.935（95%CI:0.857 0.100）,take the best cutoff value>0.597,the sensitivity was 75%,100%.Conclusion:1.Histograms of parameters of monoexponential model,biexponential model and diffusion kurtosis model are valuable in the differential diagnosis of lung cancer,infectious lesions and lung cancer subtypes.2.In the lung cancer and infectious lesion groups,ADC skewness（AUC=0.678,OR=7.061）and D median（AUC=0.777,OR=0.044）showed good predictive power for lung cancer and were significant independent predictors.3.In subgroup analysis of lung squamous cell carcinoma and adenocarcinoma,D kurtosis （AUC=0.873,OR=7.245）and D_appSD（AUC=0.750,OR=2.214）were independent predictors of lung squamous cell carcinoma.4.The parameter D has significant clinical application value for the differentiation of lung cancer,infectious lesions and lung cancer subtypes,and has a good application prospect.