Radiomics of magnetic resonance imaging has a predictive effect on the progression-free survival of metastatic gastrointestinal stromal tumors after imatinib treatment
DU Juan
ZHANG Duo
YI Huiling
YANG Linsha
ZHENG Tao
LIU Defeng
Abstract：Objective To identify sensitive radiomics features based on multiparametric magnetic resonance(mpMR)imaging for predicting progression-free survival(PFS)in patients with metastatic gastrointestinal stromal tumors(mGIST)following imatinib treatment.Methods A total of 122 pathologically confirmed mGIST patients from two centers were retrospectively collected and divided into a training set(n=83)and a test set(n=39).Three-dimensional regions of interest were outlined on mpMR images,and radiomics features were extracted.Features with poor repeatability or strong collinearity were excluded based on interclass correlation coefficients(ICC)and Pearson correlation coefficients.The least absolute shrinkage and selection operator(LASSO)regression method was applied for feature selection,and cutoff values for the selected features were determined.Kaplan-Meier curves were used to evaluate PFS,and ROC curves assessed the predictive value of the features for 3-year PFS in patients.Results The First-Order 10th percentile@ADC,GLRLM GLN@DWI,and GLRLM RLNN@DWI features showed strong correlations with mGIST PFS.Significant differences in these three features were observed between patients with progressive disease and non-progressive disease(P<0.05).The First-Order 10th percentile@ADC exhibited a negative correlation with PFS,with hazard ratios(HR)of 0.339(95%CI:0.162-0.707)in the training set and 0.354(95%CI:0.146-0.858)in the test set.Both GLRLM GLN@DWI and GLRLM RLNN@DWI were positively correlated with PFS,with HRs of 2.300(95%CI:1.120-4.721)and 4.562(95%CI:2.370-8.783)in the training set,respectively,and HRs of 3.704(95%CI:1.471-9.328)and 7.024(95%CI:2.407-20.498)in the test set.Combining these three features provided strong predictive performance for 3-year PFS,with AUC of 0.909(95%CI:0.861-0.567)in the training set and 0.788(95%CI:0.698-0.877)in the test set.Conclusion Radiomics features of mGIST are associated with patient PFS and can serve as reliable predictors for the prognosis of imatinib-targeted therapy.
Keywords：radiomicssurvival analysisgastrointestinal stromal tumorprogression-free survivalmagnetic resonance imaging
Publication Date：2025-06-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：8( 748-755 )