| Noninvasive prediction of Ki-67 expression in hepatocellular carcinoma using machine learning-based ultrasomics:a multicenter studyObjective:The study investigated the ability of machine learning-based ultrasomics to predict Ki-67 expression in hepatocellular carcinoma(HCC).Methods:A total of 244 patients from three hospitals were retrospectively recruited.First,patients from hospitals 1 and 2 were mixed.Stratified random sampling was performed to divide the patients into the training dataset(n=168)and testing dataset(n=43)at a ratio of 8:2.Patients from hospital 3(n=33)were classified into the independent validation dataset.Then,lesion segmentation were manually on the ITK-SNAP software,and ultrasomics features were extracted using the pyradiomics.Feature selection was conducted using intraclass correlation coefficient,variance threshold,mutual information method,and recursive feature elimination plus eXtreme Gradient Boosting.Finally,support vector machine was combined with grid search parameter tuning and the learning curve to construct the clinical,ultrasomics,and combined models.The predictive performance of the models was evaluated by the area under the receiver operating characteristic curve(AUC)with the 95%confidence interval(95%CI),accuracy,sensitivity and specificity.Results:In the training,test and validation datasets,the AUC of the ultrasomics model reached 0.955,0.861and 0.665,respectively.The performance of combined model had been improved in the three datasets.The corresponding AUC(95%CI),sensitivity,specificity,and accuracy were 0.986(0.955-0.998),0.973,0.840,and 0.869 on the training dataset,0.871(0.734-0.954),0.750,0.654,and 0.814 on the testing dataset,and 0.742(0.560-0.878),0.714,0.808 and 0.788 on the validation dataset,respectively.Conclusions:Ultrasomics was proved to be a potential noninvasive method for preoperative stratification of Ki-67 expression in HCC patients.The performance of the combined model composed of ultrasound omics features and clinical features has been improved to some extent.Ultrasomics prediction for CK19 expression in hepatocellular carcinoma:A multicenter studyObjective:To investigate the value of machine learning-based ultrasomics for preoperative prediction of Cytokeratin 19(CK19)expression in patients with HCC.Methods:We retrospectively analyzed 214 patients with pathologically confirmed HCC and received CK19 immunohistochemical staining.Through random stratified sampling(ratio,8:2),patients from hospital 1 and 2 were assigned to the training dataset(n=143)and the test dataset(n=36),and patients from hospital 3 served as external validation dataset(n=35).All gray-scale ultrasound images were preprocessed,and then the regions of interest were manually segmented by two sonographers.A total of 1409 ultrasomics features were extracted from the original and derived images.Next,the intraclass correlation coefficient,variance threshold,mutual information,and embedding method were applied to feature dimension reduction.Finally,the clinical model,ultrasomics model,and combined model were constructed by eXtreme Gradient Boosting.The predictive performance of models was assessed by AUC,accuracy,sensitivity,and specificity.Results:A total of 12 ultrasomics signatures and 21 clinical features were finally used to construct the predictive models.The AUC of the ultrasomics model were 0.789 and 0.787 in the test and validation datasets,respectively.However,the performance of the combined model improved significantly.The AUC,accuracy,sensitivity,and specificity were 0.867(0.712-0.957),0.861,0.750,0.875,and 0.862(0.7030.955),0.857,0.833,0.862 in the test dataset and external validation dataset,respectively.Conclusions:Ultrasomics signatures could be used to predict the expression of CK19 in HCC patients.The combination of clinical features and ultrasomics signatures showed excellent effects,which not only improved the predictability,but also further enhanced the generalization ability. |
PDF Full Text Request