Preliminary Study Of Clinical Application Prediction Of HCC-GPC3 Based On EOB-MRI Image-Clinical Features Nomogram

Objective:To explore the value of nomogram model for predicting the expression of phosphatidyl inositol proteoglycan 3（GPC3）in hepatocellular carcinoma（HCC）based on MRI imaging features and clinical indicators enhanced by gadolinium serate disodium.Methods:Clinical,pathological and EOB-MRI imaging data of 129 HCC patients were retrospectively analyzed,including 101 GPC3（+）cases and 28 GPC3（-）cases.Kolmogorov-smirnoff method was used to test the normality of continuous variables.Independent sample t test was used to analyze those consistent with normal distribution,and Mann-Whitney U test was used to analyze those inconsistent with normal distribution.For categorical variables,Chi-square test or Fisher exact test were used to analyze the clinical and imaging data of the two groups of patients.Kappa coefficient was used to evaluate the consistency among observation groups.Multivariate Logistic regression analysis was performed to determine independent risk factors for GPC3 expression.Based on the above independent risk factors,nomogram model was constructed using R4.2 to predict positive GPC3 expression.The C index was used to evaluate the nomogram model’s ability to distinguish GPC3（+）and GPC3（-）,and the corrected C index was obtained through bootstrap resampling for internal verification.Calibration curves were used to assess the consistency of predicting positive GPC3 expression,and decision curve analysis（DCA）was used to evaluate the clinical efficacy of the nomogram model.Area under the curve（AUC）of the ROC curve was used to evaluate the predictive power of the nomogram model.Results:1、There were significant differences between GPC3（+）and GPC3（-）groups in age,hepatitis B,alpha-fetoprotein（AFP）,non-peripheral washout,lesion edge morphology,hepatobiliary low signal,apparent diffusion coefficient（ADC）,portal tumor signal intensity/liver signal intensity,hepatobiliary tumor signal intensity/liver signal intensity（P<0.05）.2、MultivariateanalysisresultsshowedthatAFP≥20ug/L,ADC≤1.06×10^-3mm²/s,non-peripheral washout,and unsmooth tumor edge had statistical significance in predicting GPC3 expression in HCC（P<0.05）.3、A nomogram model for predicting positive GPC3 expression in HCC was established based on independent risk factors,and the C index was 0.921,and the corrected C index was 0.897.The calibration curve showed that the nomogram model was consistent in predicting positive GPC3 expression in HCC.DCA curves showed that nomogram models had a net clinical benefit in the 0-1 threshold probability range.Nomogram predicted that the AUC of positive GPC3 expression was 0.921,the sensitivity was 79.2%,and the specificity was 92.9%.Conclusion:Nomogram models based on AFP≥20ug/L,ADC≤1.06×10^-3mm²/s,non-peripheral washout,and non-smooth tumor edge can predict the positive expression of GPC3 in HCC,providing a means for clinical noninvasive prediction of GPC3 before surgery.