Prediction Of IDH1 Status In Adult Diffuse Glioma Based On Preoperative T1CE Clinical-imaging Model

Objective: To evaluate the predictive value of clinical imaging model based on preoperative T1CE for IDH1 status of adult diffuse glioma.Methods: Methods: A total of 154 patients with adult diffuse glioma confirmed by pathology from the First Affiliated Hospital of China Medical University from January2013 to August 2022 were retrospectively included,including 77 patients with IDH1 mutant type and 77 patients with IDH1 wild type.Randomly stratified sampling was used to divide the patients into training set(107 cases)and verification set(47 cases)in a ratio of 7:3.All patients had complete preoperative routine MRI images(including T1WI,T2WI,FLAIR,T1CE sequences).The T1CE images were registered to T2WI images using ITK-SNAP software,and then the area of interest(including tumor parenchyma and peritumoral edema)was delineated layer by layer on T2WI images.Finally,the ROI in the T2WI image was copied to the registered T1CE image.Feature extraction was carried out by A.K software,correlation > 0.7 was eliminated,and univariate Logistic regression p < 0.1.Feature screening was carried out by gradient lifting decision tree(GBDT)method,and then multi-factor logistic regression model was established.Three prediction models were constructed by independent image omics features and combined with different clinical features:(1)Radscore model;(2)Radscore+ age combined model;(3)Radscore+ age + location combined clinical-imaging model.The predictive efficacy of each model for IDH1 expression in adult diffuse glioma was evaluated by receiver operating characteristic curve(ROC).Mann-Whitney U test was used for continuous variables with non-normal distribution,and Chi-square test was used for categorical variables.Statistical analysis was conducted on the differences of clinical data such as age,gender,machine classification,tumor location and selected features of each model between the IDH1 mutant group and the IDH1 wild group,and p<0.05 was considered statistically significant.Results: There was no statistical difference in machine classification and gender between IDH1 mutant group and IDH1 wild group,but there was statistical significance in age and tumor location.1319 image features were extracted from T1CE images,and 12 most significant features were finally obtained after feature reduction screening.The selected features were compared between the IDH1 mutant group and IDH1 wild group,and the predictive efficiency of the model was evaluated by ROC curve.In addition,the three models constructed were compared.The AUC values of Radscore,Radscore+age and Radscore+age+ location in the training set were 0.868,0.877 and 0.923,respectively.Therefore,the clinics-imaging omics model with Radscore+age+location was better.Conclusions: This study shows that the clinical-imaging model based on preoperative T1CE can effectively and noninvasive preliminarily predict the status of adult diffuse glioma IDH1,which is expected to provide certain clinical reference value for the prognosis of adult diffuse glioma patients.