Font Size: a A A

The Study Of Prediction Of Hepatocellular Carcinoma Early Recurrence By Peritumoral And Intratumoral Different ROI Radiomics Features

Posted on:2023-05-17Degree:MasterType:Thesis
Country:ChinaCandidate:W D KangFull Text:PDF
GTID:2544307070998439Subject:Clinical medicine
Abstract/Summary:PDF Full Text Request
Objective: To investigate the value of enhanced CT imaging with different ROI sizes combined with clinical factor models in predicting early recurrence of hepatocellular carcinoma(HCC)after resection(≤2 years).Methods: 160 HCC patients from Xiangya Third Hospital of Central South University from January 2014 to October 2020 were enrolled in this study,and early postoperative recurrence of HCC was determined by follow-up.The original tumor area of interest(OROI)was delineated in enhanced CT image,and then the peritumor ROI of 1-5 ROI was successively extended outward based on OROI to form five ROI of 1-5.Features were extracted,dimension reduction,feature selection and logistic regression model were established.The prediction efficiency of ROI models of different sizes was compared,and combined with clinical factors,the combined model was established.Results: Among 160 HCC patients,97(60.6%)had early recurrence and 63(39.4%)did not.OROI-ROI 5 has 3,4,13,19,13 and 12 features for model building respectively.In the training group,the AUC value and95% confidence interval(CI)of OROI-ROI 5 were: 0.745(0.654-0.836),0.741(0.651-0.832),0.801(0.720-0.882),0.824(0.728-0.893),0.786(0.702-0.8670),0.789(0.705 0.873).The maximum AUC value of ROI 3model was 0.824,which was higher than that of OROI model(0.745)with a significant difference(P<0.05).There was no significant difference between residual ROI and OROI AUC value.In the validation group,the OROI-ROI 5 AUC and 95% confidence interval are: 0.574(0.398-0.750),0.621(0.454-0.788),0.566(0.4005-0.7323),0.725(0.568-0.882),0.605(0.438-0.773),0.635(0.458 0.811).The maximum AUC value of ROI 3model is 0.725,which is higher than OROI AUC value of 0.574(P<0.05)and has a significant difference.There is no significant difference between residual ROI and OROI AUC value.ROI 3 was the best ROI,Rad 3 score was calculated,and image omics label was established.Multivariate logistic regression analysis showed that microvascular invasion(MVI)(P=0.011),satellite nodule(P=0.008)and AFP(P=0.034)were independent risk factors for early postoperative recurrence.Rad 3 score,MVI,AFP and satellite nodule factors were incorporated into the clinical and imaging combined model.The sensitivity,specificity,AUC and 95%CI of the combined model in the training group were 0.923,0.884,0.913(0.862-0.963)(P<0.05),the sensitivity,specificity,AUC and 95%CI of the combined model in the validation group were 0.857,0.829,0.832(0.714-0.950)(P<0.05).Conclusions: In the prediction model of early HCC recurrence(≤2years)based on enhanced CT radiomics,the peritumoral range extending sequentially from the original tumor can improve the prediction efficiency of the model,and different ROI sizes show different prediction abilities in the radiomics model.In this study,intratumoral and periatumoral expansion of 3mm was the best ROI,a clinical-imaging combined model was established and combined with clinical factors,which had the best predictive efficiency.
Keywords/Search Tags:Hepatocellular carcinoma, Early recurrence, Radiomics, Prediction model, Nomogram
PDF Full Text Request
Related items