The value of enhanced T1WI radiomics in predicting the IDH 1 genotype of mutant and wild type in highgrade gliomas
XIA Shuiwei, ZHOU Yongjin, CHEN Chunmiao, CHEN Jiajun, HUI Junguo, CHEN Minjiang,KONG Chunli, WANG Zufei, JI Jiansong.
Department of Radiology, Key Laboratory of Imaging Diagnosis and Minimally Invasive Intervention Research of Zhejiang Province, the Fifth Affiliated Hospital of Wenzhou Medical University, Lishui 323000, China
Abstract：Objective: To explore the value of radiomics predictive model based on preoperative enhanced T1WI images in predicting the mutant and wild type of isocitrate dehydrogenase-1 (IDH 1) genotype in patients with high-grade gliomas. Methods: A retrospective analysis was conducted of 89 high-grade glioma patients with
complete preoperative craniocerebral enhanced T1WI images in the Fifth Affiliated Hospital of Wenzhou Medical University from June 2018 to December 2020, including 32 cases of IDH 1 mutant type (15 cases of WHO grade III, 17 cases of WHO grade IV) and 57 cases of IDH 1 wild type (12 cases of WHO grade III, 45 cases of WHO grade IV). The two groups of patients were randomly divided into training group and validation group with a ratio of 7:3. A.K software was used to extract the texture features from the original enhanced T1WI images,and Kruskal-Wallis test, Spearman correlation analysis, LASSO-regression and ten-fold cross validation were
performed for feature dimension reduction and identifying the most characteristic parameters to build the predictive model. Receiver operating characteristic (ROC) curve was used to evaluate the diagnostic performance of the model in identifying IDH 1 mutant and IDH 1 wild-type high-grade gliomas. Decision curve analysis (DCA) was used to evaluate the clinical benefit of the model. Results: A total of 396 texture parameters were extracted from theenhanced T1WI images of each patient. Five texture parameters were selected ultimately by LASSO-regression and ten-fold cross validation, and the corresponding radiomics scores (Rad-scroes) were calculated to construct the prediction models of the training group and the validation group. The area under curve (AUC) of ROC of the training group was 0.902 (95%CI: 0.826-0.978), with the sensitivity and specificity being 84.6% and 81.8%, respectively. The AUC of the validation group was 0.844 (95%CI: 0.676-1.000), with the sensitivity and specificity being 77.8% and 80.1%, respectively. DCA showed that the net benefit of the radiomics model was better than that of the untreated model and the all-treated model with a risk threshold ranges from 0.1 to 1.0. Conclusion: The radiomics model based on enhanced T1WI images can effectively identify the IDH 1 mutant and wild type of high-grade glioma.
