Gray-scale ultrasound and shear-wave elastography for predicting Ki-67 expression in hepatocellular carcinoma
Objective To investigate the predictive value of grayscale ultrasound (US) and shear wave elastography (SWE) for Ki-67 expresson in hepatocellular carcinoma (HCC).Methods A total of 277 HCC patients confirmed by surgical pathology admitted in Lishui Central Hospital from January 2016 to March 2024 were retrospectively analyzed.Patients were randomly divided into a training set and a validation set in a ratio of 7︰3.Patients were categorized into high and low Ki-67 expression groups based on immunohistochemistry results.The 3D Slicer software was used to manually outline the regions of interest on US and SWE images.Radiomics features were extracted using Pyradiomics,and the optimal features were selected using the intraclass correlation coefficient,Pearson correlation analysis,and least absolute shrinkage and selection operator (lasso) regression.Three radiomics models (US,SWE,and US+SWE) were constructed using the random forest algorithm,and radiomics scores were obtained.Independent predictors of Ki-67 expression from clinical indicators were identified to construct clinical model using multivariate logistic regression analysis.The best-performing radiomics scores and clinical independent predictors were combined to construct a nomogram.The predictive performance of prediction models was assessed with ROC curve.Results There were no statistically significant differences in gender,age,liver cirrhosis,HBsAg,ALT,AST,albumin,international standardized ratio,Tbil,and Dbil between patients with high Ki-67 expression and low Ki-67 expression in both training set and validation set (all P>0.05).However,there were significant differences in tumor size and alpha fetoprotein (AFP) levels (all P<0.05).Finaly 9,7,and 12 optimal radiomics features were selected through lasso regression to construct the US,SWE,and US+SWE radiomics models,respectively.The areas under ROC curve (AUCs) of these models were 0.829,0.792,0.870 in the training set and 0.779,0.751,0.829 in the validation set.Tumor size and AFP were identified as independent predictors to construct clinical model with AUCs of 0.754 and 0.740 in the training and validation sets,respectively.The AUCs of nomogram based on the combination of US+SWE radiomics scores with tumor size and AFP were 0.926 and 0.867 in the training and validation sets,respectively.Conclusion The nomogram based on the combination of US+SWE radiomics scores with clinical factors has good predictive value for Ki-67 expression levels in HCC patients,which may has the potential to guide clinical decision-making.
万方数据
维普
