Evaluation Of Microvessel Density And Microvascular Invasion Of Hepatocellular Carcinoma Based On Multi-sequence MRI Histogram And Nomogram Model

Part 1 Evaluation of Microvessel Density of Hepatocellular Carcinoma Based on Multi-sequence MRI HistogramObjective: To explore the value of multi-sequence MRI histogram in predicting microvessel density(MVD)ofhepatocellular carcinoma(HCC).Methods: 93 cases of solitary HCC confirmed by pathology were analyzed retrospectively,including 57 cases with high MVD(MVD≥50)and 36 cases with low MVD(MVD(27)50).All patients underwent 1.5T EOB-MRI before surgery.Histogram analysis was performed in nine sequences including T1 WI,T2WI,fat suppressed T2 WI,DWI,ADC map and dynamic contrast enhancement(early arterial phase,late arterial phase,portal venous phase and hepatobiliary phase).The mean,variance,minimum value,maximum value,skewness,kurtosis and the 1st,5th,25 th,50th,75 th,95th,99 th percentiles were obtained.All parameters were compared between the two groups,and receiver operating characteristic curve(ROC)was used to analyze the efficiency of high and low MVD.Results1.Among the above 9 sequences of high and low MVD,the histogram parameters of 7 sequences were statistically different.The mean,maximum value and the 1st,5th,25 th,50th,75 th,95th percentiles of T2 WI were statistically significant between the two groups.The mean,minimum value,maximum value and the 1st,5th,25 th,50th,75 th,95th percentiles of fat suppressed T2 WIwere statistically significant between the two groups.The mean,variance,maximum value and the 1st,5th,25 th,50th,75 th,95th,99 th percentiles of DWI were statistically significant between the two groups.The skewness,kurtosis and the50 th percentile of the early arterial phase were statistically significant between the two groups.The mean,minimum value and the 1st,5th,25 th,50th,75 th,95th,99 th percentiles of the late arterial phase were statistically significant between the two groups.The mean and the 5th,25 th,50th,75 th,95th,99 th percentiles of the portal venous phase were statistically significant between the two groups.The kurtosis of the hepatobiliary phase was statistically significant between the two groups.2.Among the parameters with statistical differences in each sequence,the maximum area under curve(AUC)for the classification of high and low MVDwere as follows.AUC of the 75 th percentile of T2 WI was 0.681.AUC of the mean of fat suppressed T2 WI was 0.725.AUC of the 50 th percentile of DWI was 0.726.AUC of the skewness of early arterial phase was 0.696.AUC of the 25 th percentile of late arterial phase was 0.707.AUC of the 25 th percentile of portal venous phase was 0.665.And AUC of the kurtosis of hepatobiliary phase was 0.639.3.The combination of the above-mentioned high diagnostic efficacy parameters had higher diagnostic efficacy for high and low MVD,AUC was0.821,which waspartial statistically significant compared with the diagnostic efficacy of each single parameter.Conclusion: The parameters of multi-sequence MRI histogram can provide reference for MVD diagnosis and evaluation of HCC,and the combined application of multi parameters can further improve the diagnosis efficiency.Part 2 To Establish and Validate a Nomogram Model for Preoperative Prediction of Microvessel Density in Hepatocellular CarcinomaObjective: To establish and validate a nomogram model for preoperative prediction of MVD in HCC.Methods: The clinical information and MRI images feature of 93 patients of solitary HCC confirmed by pathology were collected and analyzed.Combined with histogram joint probability as histogram label to establish a nomogram model for preoperative prediction of MVD.According to the principle of random,56 cases were selected as the modeling group and 37 cases as the verification group.Logistic regression analysis was used to screen out the independent risk factors to predict high MVD and establish a prediction nomogram model.The C index and calibration curve were used to evaluate the accuracy and consistency of the model prediction.Results: The independent risk factors for high MVD were the MRI tumor maximum diameter and histogram labels.Based on these two independent risk factors,the nomogram model was established.The C indexes of the modeling group and the verification group were 0.846 and 0.791,respectively.The calibration curve indicated that the model was in good agreement.Conclusion: The nomogram model can predict the MVD status of HCC and has clinical utility.Part 3 Evaluation of Microvascular Invasion of Hepatocellular Carcinoma Based on Multi-sequence MRI HistogramObjective: To explore the value of multi-sequence MRI histogram in predicting microvascular invasion(MVI)of HCC.Methods: 101 cases of solitary HCC confirmed by pathology were analyzed retrospectively,including 59 cases of MVI positive and 42 cases of MVI negative.All patients underwent 1.5T EOB-MRI before surgery.Histogram analysis was performed in nine sequences including T1 WI,T2WI,fat suppressed T2 WI,DWI,ADC map and dynamic contrast enhancement(early arterial phase,late arterial phase,portal venous phase and hepatobiliary phase).The mean,variance,minimum value,maximum value,skewness,kurtosis and the 1st,5th,25 th,50th,75 th,95th,99 th percentiles were obtained.All parameters were compared between the two groups,and curve was used to analyze the efficiency ofdifferential MVI-positive.Results1.Among the above 9 sequences of MVI positive and negative groups,the histogram parameters of 8 sequences were statistically different.The mean,maximum value and the 1st,5th,25 th,50th,75 th,95th percentiles of T2 WI were statistically significant between the two groups.The maximum value and 75 th percentiles of fat suppressed T2 WIwere statistically significant between the two groups.The mean,variance,maximum value and the 1st,5th,25 th,50th,75 th,95th,99 th percentiles of DWI were statistically significant between the two groups.The mean,minimum value,and the 1st,5th,25 th,50th percentiles of ADC map were statistically significant between the two groups.The mean,minimum value,skewness,kurtosis and the 1st,5th,25 th,50th,75 th,95th,99 th percentile of the early arterial phase were statistically significant between the two groups.The mean,minimum value and the 1st,5th,25 th,50th,75 th,95th,99 th percentiles of the late arterial phase were statistically significant between the two groups.The mean,minimum value and the 1st,5th,25 th,50th,75 th,95th,99 th percentiles of the portal venous phase were statistically significant between the two groups.The minimum value of the hepatobiliary phase was statistically significant between the two groups.2.Among the parameters with statistical differences in each sequence,the maximum AUCfor the diagnosis of MVI positive were as follows.AUC of the 1st percentile of T2 WI was 0.687.AUC of the maximum value of fat suppressed T2 WI was 0.648.AUC of the maximum value of DWI was 0.711.AUC of the minimum value of ADC map was 0.691.AUC of the mean of early arterial phase was 0.736.AUC of the 50 th percentile of late arterial phase was 0.743.AUC of the 50 th percentile of portal venous phase was 0.711.And AUC of the minimum value of hepatobiliary phase was 0.633.3.The combination of the above-mentioned high diagnostic efficacyparameters had a higher diagnostic efficacy for MVI positive,AUC was 0.833,which was significantly different from the diagnostic efficacy of each singleparameter.Conclusion: The parameters of multi-sequence MRI histogram can provide reference for the MVI diagnosis and evaluation of HCC,and the combined application of multi parameters can further improve the diagnosis efficiency.Part 4 To Establish and Validate a Nomogram Model for Preoperative Prediction of Microvascular Invasion in Hepatocellular CarcinomaObjective: To establish and validate a nomogram model for preoperative prediction of MVI in HCC.Methods: The clinical information and MRI images feature of 101 patients of solitary HCC confirmed by pathology were collected and analyzed.Combined with histogram joint probability as histogram label to establish a nomogram model for preoperative prediction of MVI.According to the principle of random,61 cases were selected as the modeling group and 40 cases as the verification group.Logistic regression analysis was used to screen out the independent risk factors to predict MVI positive and establish a prediction nomogram model.The C index and calibration curve were used to evaluate the accuracy and consistency of the model prediction.Results: The independent risk factors for MVI positive were gender,tumor protruding liver margin or exogenous growth,histogram label.Based on these three independent risk factors,the nomogram model was established.The C indexes of the modeling group and the verification group were 0.845 and 0.907 respectively.The calibration curve indicated that the model was in good agreement.Conclusion: The nomogram model can predict the MVI status of HCC and has clinical utility.