Multi-regional radiomics based on T1WI enhanced images for predicting brain invasion of meningioma
Objective To explore the application value of multi-region radiomics based on T1WI enhanced images in the preoperative prediction of brain invasion in patients with meningiomas.Methods A retrospective analysis was conducted on the clinical imaging data of patients with meningioma confirmed by pathology at the Fifth Affiliated Hospital of Wenzhou Medical University(Center 1)and Lishui People's Hospital(Center 2)from January 2016 to February 2024.A total of 297 patients in Center 1 were randomly divided into a training set(n=208,152 cases without brain invasion and 56 cases with brain invasion)and an internal testing set(n=89,65 cases without brain invasion and 24 cases with brain invasion)in a 7∶3 ratio;117 cases in Center 2 were set as an external test set,including 91 cases without brain invasion and 26 cases with brain invasion.Independent risk factors for meningioma brain invasion were screened using univariate analysis and multivariate logistic regression analysis,and a clinical model was constructed.The best radiomics features were extracted and screeed from T1WI enhanced images based on the intratumoral region[gross tumor volume(GTV)]and peritumoral region[gross and peritumoural volume(GPTV),GPTV1,GPTV2,GPTV3,GPTV4,GPTV5,GPTV10,GPTV15,and GPTV20],and corresponding models were established using support vector machines.The highest average AUC between the internal and external test sets was selected as the optimal radiomics model,and the results were converted into radiomics scores(Rad-score).Subsequently,a combination model was constructed based on clinical risk factors and Rad-score,and a nomogram was developed for visualization.Results Among clinical factors,gender(OR=4.073,P＜0.001)and peritumoral edema(OR=4.202,P＜0.001)were independent risk factors for predicting meningioma brain invasion.In the internal test set,the AUC of GTV and GPTV ranged from 0.679 to 0.833;in the external test set,the AUC of GTV and GPTV ranged from 0.612 to 0.808.Among them,GPTV10 had the best predictive performance(AUC=0.821).The combination of gender and peritumoral edema with Rad-score was applied to develop a nomogram model.The nomogram exhibited good performance in predicting meningioma brain invasion,with AUC of 0.937,0.879,and 0.845 in the training set,internal test set and external test set,respectively.Conclusion The GPTV10 radiomics model based on T1WI enhanced images can well predict the brain invasion of meningiomas before surgery,and a nomogram in combination of radiomics and clinical risk factors can further improve performance.
万方数据
