| Purpose:1.To investigate the value of quantitative data of dual-energy CT spectral imaging to predict microsatellite instability?MSI?status of colorectal cancer.2.To establish imaging radiomics models based on dual-energy CT spectral imaging to predict MSI status of colorectal cancer.3.To establish clinicopathological-imaging radiomics models of dual-energy CT spectral imaging,and to evaluate their value for predicting MSI status of colorectal cancer.Materials and Methods:The 117 colorectal cancer patients confirmed by postoperative pathology were included in our study.All patients have underwent Revolution CT enhanced spectral imaging scan in our hospital within one month before surgery.The post-operative tissues have been tested by immunohistochemical examination to detect the MSI status?MSI colorectal cancer,n=39;MSS colorectal cancer,n=78?.In the first part of the study about“Quantitative parameters of dual-energy CT spectral imaging to predict MSI status of colorectal cancer",we selected the slice with the largest axial diameter of the colorectal cancer and its adjacent upper and lower slices.Two observers measured quantitative parameters of spectral imaging in colorectal cancer patients:?1?the effective atomic number during the pre-contrast CT scan,?2?the CT values on pre-contrast and enhanced phases?arterial,venous and delayed phase?single-energy images?40?140 ke V?,which were used for drawing energy spectrum curve and calculating the slope of the curve:slope?k?=(CT value 40ke V-CT value 100 ke V)/60;?3?the iodine concentration value on enhanced phases.The intraclass correlation coefficient?ICC?was used to evaluate the consistency of the spectral parameters measured by two observers?ICC>0.80 means good consistency?.The independent sample t-test or Mann-Whitney U test was used to analyze the differences between the groups of measurement data.Chi-square test was used to analyze classification data.In order to obtain the spectrum parameters that can effectively predict the MSI status of colorectal cancer,the spectrum parameters were analyzed by univariate and multifactor logistic regression analysis,and Hosmer-Lemeshow test was used to evaluate the goodness of fit of the logistic regression model.Receiver operating characteristic?ROC?analysis was used to evaluate the diagnostic efficacy,and the area under the curve?AUC?,95%confidence interval of AUC,cutoff value,sensitivity,specificity were calculated.The De Long test was used to compare statistical differences between AUCs.P<0.05 indicated that the data were statistically significant.In the second part of the study about“Predicting MSI status of colorectal cancer based on radiomics models on arterial phase of dual-energy CT spectrum imaging”,we reconstructed the arterial phase iodine?water?maps and the 40?100 ke V single-energy images during the arterial phase?because our first part suggested that the arterial phase spectrum parameters were predictors of MSI status in colorectal cancer?.We selected the slice with the largest axial diameter of the colorectal cancer and its adjacent upper and lower slices.Images were exported with Digital Imaging and Communications in Medicine?DICOM?format.The ITK-SNAP software was used to perform lesion segmentation.The Artificial Intelligent Kit software was used to extract 618 radiomics features,including histogram features,texture features,and shape features.All cases?n=117?were randomly assigned as training data set?n=81?and validation data set?n=36?according to 7:3,and feature selecting and model building were performed based on the training data set.The feature selecting steps were as follows:observer consistency test,missing value filling and data standardization,correlation test,least absolute shrinkage and selection operator?LASSO?algorithm.Finally,we built the multi-parameter logistic regression model based on the selected radiomics features,and the 5-fold cross-validation method was used to select the optimal model.The prediction models included:?1?radiomics models based on each single energy image?40,50,60,70,80,90,100 ke V?;and the combined model based on 40?100 ke V single energy images;?2?radiomics model based on iodine?water?map;?3?combined model based on iodine map and 40?100ke V single energy images.Model evaluation:?1?evaluating the diagnostic effectiveness by ROC analysis;?2?evaluating the calibration performance by calibration curves.P>0.05 in the H-L test indicates a good fit;?3?evaluating the clinical application value by decision curves.In the third part of the study about“Predicting MSI status of colorectal cancer based on the clinicopathological-imaging radiomics models on arterial phase of dual-energy CT spectrum imaging”,we recorded the clinicopathological information of the patients,including age,gender,serum carcino-embryonic antigen?CEA?,serum carbohydrate antigen 19-9?CA19-9?,drinking history,smoking history,hypertension history,diabetes history,family history of cancer,tumor size,tumor location,serosal/mesenteric invasion,and lymph node metastasis.The clinicopathological parameters were analyzed by univariate and multifactor logistic regression analysis,and the clinicopathological model for predicting the MSI status was established.We further combined it with the three radiomics models established in the second part,respectively.The combined clinicopathological-imaging radiomics models were obtained.Model evaluation:?1?evaluating the diagnostic effectiveness by ROC analysis;?2?evaluating the calibration performance by calibration curves.P>0.05 in the H-L test indicates a good fit;?3?evaluating the clinical application value by decision curves.Results:In the first part,the quantitative parameters of spectral imaging had good consistency,and the ICC values were all greater than 0.80.In the MSI group,the effective atomic number in the pre-contrast phase,the slopes of the energy spectrum curves in the pre-contrast phase and the enhanced phases,the iodine concentration values in the enhanced phases,the CT values of the 70 ke V single energy images in the pre-contrast phase,arterial and venous phases,were all significantly lower than those of the MSS group?P<0.005?.Logistic regression analysis showed that the spectral parameters on arterial phase?the slope of the energy spectrum curve,the CT value of the 70 ke V single energy image,and the iodine concentration value?were all predictors of the MSI status of colorectal cancer.The diagnostic values were 0.779,0.749,and 0.785,respectively.The combination of arterial phase spectrum parameters had a significantly improved efficiency with an AUC value of 0.852.In the second part of the study about“Predicting MSI status of colorectal cancer based on radiomics models on arterial phase of dual-energy CT spectrum imaging”,we established the radiomics models:?1?radiomics model based on each single energy image?40,50,60,70,80,90,100 ke V?:in the training group,AUC was between 0.706?0.795;in the validation group,AUC was between 0.687?0.816.The combined radiomics model based on 40?100ke V single energy images:in the training group,the AUC was 0.798;in the validation group,the AUC was 0.837.?2?radiomics model based on iodine?water?maps:in the training group,the AUC was0.846;in the validation group,the AUC was 0.882.?3?combined model of iodine map and 40?100 ke V images:in the training group,AUC was 0.862;in the validation group,AUC was 0.920.There was no statistical difference between the AUCs of training group and validation group of the above models.There was no statistical difference between the AUCs of radiomics models based on combined single images,iodine?water?maps,and their combination.The calibration curves indicated that the models had good calibration effects.The P values in the Hosmer-Lemeshow test were all greater than 0.05.The decision curve confirmed that the models had clinical application value.In the third part of the study about“Predicting MSI status of colorectal cancer based on the clinicopathological-imaging radiomics models on arterial phase of dual-energy CT spectrum imaging”,the clinicopathological parameters including age,CEA,and tumor location were used as predictors of MSI status.The above parameters were used to establish a clinicopathological model.The AUC values of the training group and the validation group were 0.824 and 0.740,respectively.We further establishedclinicopathological-imagingradiomicsmodels:?1?clinicopathological-combined single images model:in the training group,AUC was0.917;in the validation group,AUC was 0.903;?2?clinicopathological-iodine model:in the training group,AUC was 0.903;in the validation group,AUC was 0.892;?3?clinicopathological-combined single images-iodine model:in the training group,AUC was 0.922;in the validation group,AUC was 0.951.There was no statistical difference between the AUCs of training group and validation group of the above models.There was no statistical difference between the AUCs of above three clinicopathological-imaging radiomics models.The calibration curve indicated that the models had good calibration effects.The P values in the Hosmer-Lemeshow test were all greater than 0.05.The decision curves confirmed that the model had clinical application value.Conclusions:1.Among the spectrum parameters of dual-energy CT,the arterial phase parameters?spectrum curve slope,70 ke V CT value,and iodine concentration value?can be used as potential imaging markers of MSI colorectal cancer.2.The imaging radiomics models based on arterial phase images?single energy image combination,iodine concentration map,iodine-single energy combination?had value for predicting MSI status of colorectal cancer,but their diagnostic effects had no statistical difference.3.The three clinicopathological-imaging radiomics models based on on arterial phase images had value for predicting MSI status of colorectal cancer,but their diagnostic effects had no statistical difference.The iodine?water?map was more feasible.Therefore,we recommend the“clinicopathological-iodine model”as the optimal model for predicting the MSI status of colorectal cancer.4.This study established radiomics models based on dual-energy CT spectral imaging,which provided new ideas for predicting MSI status of colorectal cancer. |
PDF Full Text Request