The Value Of Spectral CT Multi-parameter Quantitative Analysis And CT Texture Analysis In Distinguishing Pericolic Infiltration In Colorectal Adenocarcinoma

Part Ⅰ:A radiomic nomogram based on morphological CT features combined with quantitative dual-source CT dual-energy in distinguishing pericolic infiltration in rectal adenocarcinomaObjective To explore the value of Logistic regression based Nomogram prediction model based on morphological CT features combined with quantitative dual-source CT dual-energy in distinguishing pericolic infiltration in rectal cancer,so as to improve the accuracy of T staging before the operation of rectal adenocarcinoma.Methods:62 patients who had adenocarcinoma of rectum confirmed by pathological examination and the signs of perirectal fat infiltration indicated by energy spectrum contrast CT scan were collected and divided into two groups according to pathological result:cancerous infiltration group(N=32);inflammatory infiltration group(N=30).(1)The differences in CT signs between the two groups were compared:to observe whether the serosal layer of the intestinal wall where the lesion was located was coarse,whether there was a shadow of cords around the intestine,whether there were nodal projections outside the intestinal wall,whether there were enlarged lymph nodes in the rectal mesentery(with a short diameter of 5mm),and to measure the length of the intestine involved by the rectal adenocarcinoma,and the thickest diameter of the intestinal wall where the lesion was located,(2)Rrespectively,comparing the difference between the two groups of patients with energy spectrum parameters:two groups of lesions in the three phases enhanced scan of adipose tissue and normal adipose tissue mass concentration difference(Iodine difference,Water difference,Calcium difference and HAP difference)and standardized Iodine concentration difference(NIC difference)differences in two groups of adipose tissue lesions mean Effective atomic number(the Effective-Z)and energy spectrum curve slope K value difference.(3)Multivariate Logistic regression models of quantitative parameters related to CT signs and energy spectrum of arterial and venous delay were constructed,and all parameters were included in the multivariate Logistic regression model to screen out features of great significance for differential diagnosis.ROC curve was drawn to calculate the area under the curve(AUC),so as to measure the diagnostic efficiency of the model and parameters.(4)Draw ROC curve and calibration curve of the nomogram of the multivariate Logistic regression model.Results:(1)Compared with inflammatory infiltration group,the intestinal wall in cancerous infiltration group was thicker with inhomogeneous enhancement,nodules mostly protruded the intestinal wall,and perirectal enlarged lymph nodes(diameter>5mm)were more likely to occur.(2)Material concentration differences(Iodine diffberence,Water difference,Calcium difference and HAP difference)and standard iodine concentration difference(NIC difference)between diseased fat tissue and normal fat tissue,and Effective-Z and energy spectrum curve slope K of diseased fat tissue in cancerous infiltration group were higher,showing statistically significant differences(P<0.05).(3)The diagnostic accuracy rates of the multivariate Logistic regression models established for CT signs,arterial phase,venous phase and delayed phase were 71%,90.3%,90.3%and 87.1%respectively.The AUCs of ROC of CT signs,arterial phase,venous phase and delayed phase were 0.709,0.954,0.967 and 0.969 respectively.Based on the multiple Logistic regression model of all meaningful parameters,the feature with great significance in identification was the Water difference of energy spectrum parameters in the delay period of K value in the venous period.(4)The diagnostic accuracy of the Logistic model was 93.5%,and the AUC was 0.975.The calibration curve of predictive power showed that the predicted and actual infiltration had a high consistency.Conclusion The establishment of Nomogram predictive model based on multivariate Logistic analysis of CT signs and quantitative parameters of energy spectrum CT is of great reference value in differentiating whether the signs of perienteric fat infiltration in rectal adenocarcinoma have the occurrence of cancerous invasion,and can improve the accuracy of the staging of rectal adenocarcinoma.Part Ⅱ:Nomogram based on contrast-enhanced CT texture analysis identify pericolic infiltration in colorectal carcinomaObjectives:To explore the value of Multinomial Logistic Regression Nomogram based on dynamic contrast-enhanced CT texture analysis in diffrentiating colorectal carcinoma pericolic infiltration and improving the accuracy of preoperative staging of colon cancer.Methods:100 patients who were diagnosed of colorectal carcinoma through postoperative pathological examination and received abdominal and pelvic three-phase contrast scan in our hospital one week before operation were retrospectively analyzed,including 50 cases of pericolic cancerous infiltration and 50 cases of inflammatory infiltration.MaZda software was used to extract histogram and Gray Level Co-occurrence Matrix(GLCM)texture features of pericolic adipose tissue.Comparison between two groups was done with independent sample T-test.The receiver operating characteristic(ROC)curve was drawn to calculate the cut-off value.The texture features selected were used to establish a multivariate Logistic regression.Results:Highly repeatable texture features with statistical significance after eliminating those with high correlation(|r|≥0.9)were Perc.90%(venous phase),Entropy(venous phase),DifVarnc(delayed phase)showed the best differential diagnostic efficiency.The three texture features were included into a multivariate Logistic regression model.Nomogram was drawn to visualize the model.The diagnostic accuracy rate of this model was 84%,and the AUC of ROC was 0.911.The calibration curve shows a high consistency between predicted and actual cancerous infiltration.Conclusion:The quantitative information provided by conventional CT texture analysis can be used for differentiating pericolic cancerous infiltration in colorectal carcinoma.Part Ⅲ:The value of conventional CT texture analysis in the differentiation of adipose tissue invasion in adenocarcinoma of colon.Objective To explore the application value of conventional CT texture analysis(CTTA)in differentiating cancerous invasion of pericolic fat in adenocarcinoma of colon.Methods 87 patients who were diagnosed of adenocarcinoma of colon through postoperative pathological examination and received abdominal and pelvic CT plain scan+three-phase contrast scan in our hospital one week before operation were retrospectively analyzed,including 42 cases with cancerous invasion of pericolic fat and 45 with no cancerous invasion.MaZda software was used to extract the texture features of pericolic fat to manually delineate the region of interest(ROI).Fisher coefficient,mutual information(MI),classification error probability combined average correlation coefficients(POE+ACC)and the combination of the three methods(Fisher、POE+ACC、MI,FPM)provided by MaZda software were used to screen out the texture features with differential diagnostic value.The texture features selected were classified by the following statistical analysis methods for texture feature classification:raw data analysis(RDA),principal component analysis(PCA),linear discriminant analysis(LDA)and non-linear discriminant analysis(NDA).The effect of differential diagnosis was rated by misclassification rate(MCR).A Logistic regression model was built by using the texture features selected based on the classification sequence with the minimum MCR.The diagnostic efficacy of the model was evaluated by receiver operating feature(ROC)curve.Furthermore,the best texture features for differentiating cancerous invasion of pericolic fat were obtained by reducing the dimensions of the Logistic regression model.Results Arterial phase texture features were the most significant for differentiating cancerous invasion of pericolic fat in adenocarcinoma of colon.FPM had the maximum and minimum MCR in differentiating the two lesions among the four featuer screening methods.NDA had the minimum MCR in differentiating the two lesions among the four feature classification methods.The 30 texture features selected from arterial phase images by FPM were used to establish a Logistic regression model.The accuracy,sensitivity,specificity and area under curve(AUC)of the model in differentiating cancerous invasion of pericolic fat in adenocarcinoma of colon were 93.1%,91.1%,95.2%and 0.967 respectively.By dimensionality reduction we found that Perc.99%,entropy and wavelet energy(WavEnHL_s-2)had great implications for differentiation.Conclusion The quantitative information provided by conventional CT texture analysis can be used for differentiating cancerous invasion of pericolic fat in adenocarcinoma of colon,improving the accuracy of staging of colorectal adenocarcinoma and providing more significant reference value for the selection of preoperative treatment scheme for patients.