Study On The Accurate Puncture Model Of Lung Cancer Using Energy CT And Radiomics

Objective:In view of the difficulty in determining the location of accurate puncture biopsy of lung cancer,this study used energy CT images and radiomics methods to construct a comprehensive model to guide accurate puncture biopsy of lung cancer,so as to ensure the acquisition of the most effective tissue specimens to meet the needs of biomarker detection.Materials and Methods:In the first part of this study,patients with suspected lung cancer were prospectively enrolled to undergo energy CT enhanced scans and CT-guided transthoracic lung biopsy（TTLB）.Forty-one patients with lung cancer were enrolled to identify the high tumor cell proportion region（HTPR）and low tumor cell proportion region（LTPR）in the tumor.Another 15 lung cancer patients were included to validate the accuracy of HTPR.In the 41 patients who met inclusion and exclusion criteria,suspected regions with high tumor cell proportion（HTS）[iodine density（Io D）>0.5 mg/m L])and low tumor cell proportion（LTS）（Io D≤0.5 mg/m L）were punctured according to the local Io D value,and the tumor cell proportion of 82specimens was evaluated by pathologists,with a threshold of 20%to classify the tumor cell proportion into high and low categories.The Spearman correlation analysis was used to analyze the correlation between energy CT parameters and tumor proportion.The energy CT parameter values,including conventional enhanced CT(CT_conventional)value,Io D value,normalized Io D（NIo D）value,CT_{40ke V}value,and effective atomic number（Zeff）value.The independent sample t test and Wilcoxon test were used to compare the tumor proportion and energy CT parameters of HTPR and LTPR.The receiver operating characteristics（ROC）curve was used to calculate the area under the curve（AUC）value,the sensitivity,the specificity,and the accuracy to analyze the diagnostic efficacy of each energy CT parameter in distinguishing HTPR（tumor proportion≥20%）and LTPR（tumor proportion<20%）.The Chi-square test was used to analyze the relationship between tumor proportion and puncture area（HTS and LTS）.The parameter with the highest AUC value was selected as the best energy CT parameter,and its cutoff value was used to guide the puncture biopsy of HTPR in 15 suspected lung cancer tumors for further verification,and then its accuracy was calculated.In the second part of the study,80 ROIs of high and low tumor cell proportion puncture biopsy regions in the 40 patients from the first part of the study were used as the training cohort.A total of 205 patients who were pathologically confirmed as lung cancer by CT guided TTLB in our hospital were prospectively enrolled,and the tumor cell proportions of tissue specimens were analyzed by pathologists.Then the ROIs of puncture biopsy regions were outlined and were used as the validation cohort.The consistency analysis,univariate rank-sum test,and multivariate logistic regression analysis were used for the training and verification cohorts.The multiple logistic regression analysis was used to establish models based on energy CT parameter values and radiomics respectively to predict the high and low tumor proportion region in lung cancer tumors.Based on the values of energy CT parameters and radiomics features,two models were established to predict high tumor cell proportion regions and low tumor cell proportion regions inside lung cancer tumor using the multivariate logistic regression analysis.A combine model was built by integrating the energy CT parameter model with radiomics model.The predictive efficacy of each model was evaluated using the ROC curve.Result:In the first part,through correlation analysis between tumor cell proportion and energy CT parameters,it was found that except for the CT_conventionalvalue in the venous phase,which was not related to tumor cell proportion,all other energy CT parameters were significantly correlated with tumor cell proportion.Through independent sample t-test and Wilcoxon test,it was found that there was no statistically significant difference between the CT_conventionalvalues of high and low tumor cell proportion regions in the venous phase,while the other energy CT parameters in the arterial and venous phases showed significant statistical differences between high and low tumor cell proportion regions.ROC curve analysis was used to evaluate the diagnostic performance of energy CT parameters in identifying HTPR in arterial and venous phases,and it was found that the AUC value of the Io D value in the arterial phase was the highest（0.83）,with a cutoff value of 0.59 mg/m L in the arterial phase,and a sensitivity of 70.8%and a specificity of 97.1%for identifying HTPR.The AUC values of the energy CT parameters(Io D,NIo D,CT_{40ke V},and Zeff)in the arterial and venous phases were higher than those of CT_conventional.Chi-square analysis was used to analyze the relationship between tumor cell proportion and biopsy regions,and for tumor cell proportion≥20%,the number of cases with arterial phase Io D value≥0.59 mg/m L in the biopsy region exceeded the number of cases with arterial phase Io D value<0.59 mg/m L（34 vs.14）.For tumor cell proportion<20%,the number of cases with arterial phase Io D value<0.59 mg/m L in the biopsy region exceeded the number of cases with arterial phase Io D value≥0.59 mg/m L（33vs.1）,and there was a statistical difference.The cutoff value of the arterial phase Io D value of 0.59 mg/m L was used to identify high tumor cell proportion region in 15cases of lung cancer,with an accuracy of 100%.In the second part,four energy CT parameters(venous phase Io D,venous phase Zeff,arterial phase CT_conventional,and arterial phase CT_{40ke V})were selected by multivariate logistic regression analysis to be independently(venous phase Io D,OR value=0.026,95%CI=0.004-0.175,P<0.001;venous phase Zeff,OR value=0.009,95%CI=0.001-0.126,P<0.001;arterial phase CT_conventional,OR value=0.960,95%CI=0.927-0.995,P=0.009;arterial phase CT_{40ke V},OR value=0.960,95%CI=0.954-0.984,P<0.001)that were independently correlated with high and low tumor cell proportion.Three radiomic features（Zone Entropy,Maximum,and 90Percentile）were significantly correlated with high and low tumor cell proportion,and among the three models（energy CT parameter model,radiomics model,and energy CT parameter-radiomics combined model）,the energy CT parameter-radiomics combined model had the best ability to distinguish between high and low tumor cell proportion region,and showed stable performance in the validation cohort,with AUCs of 0.916（95%CI,0.858-0.974）and 0.807（95%CI,0.702-0.911）in the training and validation cohorts,respectively.The confusion matrix showed that the positive predictive values of the training cohort and the verification cohort were 0.744（95%CI,0.694-0.756）and0.433（95%CI,0.315-0.548）,and the negative predictive values were 0.946（95%CI,0.909-0.953）and 0.952（95%CI,0.937-0.967）.Conclusion:Energy CT parameters can identify regions with tumor proportion≥20%in lung cancer tumors,which can be used to accurately plan the puncture path before CT-guided lung biopsy.The comprehensive model based on energy CT parameters and radiomics has the highest performance in predicting the high tumor cell proportion region and low tumor cell proportion region in the training cohort,and also has a good performance in the verification cohort.This comprehensive model has potential clinical application value in accurate puncture path planning before the CT-guided TTLB.