Value of a radiomics-based nomogram using 18F-FDG PET in predicting KRAS gene mutation in colorectal cancer

PANG Dinghua; YANG Hong; ZHANG Shilai; LIU Ziya; YANG Zhi; WEI Linlin; WEI Hongjiao; XIAO Guoyou

doi:10.19300/j.2025.L21312

Abstract：Objective To explore the predictive value of a radiomics-based nomogram using 18F-FDG PET for KRAS gene mutation in colorectal cancer(CRC).Methods A total of 115 patients with CRC who underwent 18F-FDG PET/CT before treatment were enrolled retrospectively,with an average age of 56.40±1.17 years.According to the gene phenotype,the patients were divided into a KRAS mutant group(56 cases)and a KRAS wild-type group(59 cases).All patients were randomly assigned into a training set of 80 cases and a validation set of 35 cases at a 7:3 ratio.Lesion regions of interest(ROI)were delineated on PET images using ITK-SNAP,and radiomics features were extracted using the Pyradiomics package.Feature selection was performed using the least absolute shrinkage and selection operator(LASSO),and radiomics scores(Rad-scores)were calculated to construct the radiomics model.Univariate analysis was conducted on clinical data,and statistically significant variables were combined with Rad-scores for multivariate logistic regression to identify independent predictors of KRAS mutations.Clinical models and combined models were developed,with the latter presented as a nomogram.The predictive performance of the models was evaluated using receiver operating characteristic(ROC)curves,and the differences in performance were compared using the Delong test.Decision curve analysis(DCA)was used to assess the clinical utility of the nomogram,and a calibration curve was employed to evaluate its calibration accuracy.Results In both the training and validation sets,the combined model demonstrated the highest predictive performance,with area under the curve(AUC)values of 0.787 and 0.727,respectively,which were significantly higher than those of the clinical model(Z=-2.230,P=0.026;Z=-2.223,P=0.026).DCA showed that the nomogram provided clinical net benefits when the threshold probability ranged from 0.13 to 0.92 in the training set and from 0.08 to 0.93 in the validation set.The calibration curve indicated good agreement between predicted and actual values for the nomogram.Conclusions The radiomics-based nomogram using 18F-FDG PET imaging holds potential predictive value for KRAS gene mutation in CRC,and may assist clinicians in devising individualized treatment strategies.