LAUSR

4081 Background: IDH1/2 mutations have a high prevalence (20%) in intrahepatic cholangiocarcinoma (iCCA) and can be associated with therapeutic benefit from IDH inhibitors. Radiomics, a developing field within imaging, has shown its ability to discriminate between tumors of distinct genomic profiles and mutational status. Methods: We developed a radiogenomic signature to robustly predict IDH1/2 mutation status (mutated versus wild-type [WT]) in 22 patients with iCCA using the pretreatment CT scans. The triphasic hepatic CT scan was used to segment the lesion. After semiautomatic segmentation of the tumor, the extracted volume of interest (VOI) was imported into our in-house radiomic pipeline and 610 radiomic features were extracted. The least absolute shrinkage and selection operator regression (LASSO) and minimum redundancy and maximum relevance (mRMR) were used for feature selection. Selected features were used to build a classification model for prediction of IDH1/2 mutation status (XGboost). The performance of the radiomics model was assessed using leave-one-out cross-validation (LOOCV). Results: Of 22 patients, 16 patients (male, 6; female, 10; average age, 55.5 years) had IDH1 (N = 14) or IDH2 (N = 2) mutations and 6 patients (male, 4; female, 2; average age = 55.5 years) had IDH1/2 WT.The CT-derived radiomic signature robustly predicted presence of IDH1/2 mutations versus WT with an area under the curve (AUC), sensitivity and specificity of 98.4%, 83.3% and 93.8%, respectively ( P = 0.037) and in a subgroup analysis presence of IDH1 mutation versus WT with an AUC, sensitivity and specificity of 98.2%, 83.3% and 92.8%, respectively ( P = 0.035). Conclusions: To our knowledge, this is the first study investigating the ability of radiogenomics as a potential method to predict the IDH1/2 mutation status in iCCA patients. Our data suggest that radiogenomic signature may correlate with IDH1/2 mutations and represent a promising non-invasive tool to stratify the patients based on molecular alterations.