临床信息与多序列MRI Transformer模型预测胶质瘤异柠檬酸脱氢酶突变状态  

Clinical information and multi-sequence MRI Transformer model predicts isocitrate dehydrogenase mutation status in glioma

作　　者：魏勇[1] 刘跃娜 周凤梅[3] 梁长华[1] WEI Yong;LIU Yuena;ZHOU Fengmei;LIANG Changhua(Department of Radiology,the First Affiliated Hospital of Xinxiang Medical University,Weihui,He’nan Province 453100,China;Department of Magnetic Resonance,the Third Affiliated Hospitalof Xinxiang Medical University,Xinxiang,He’nan Province 453000,China;Department of Magnetic Resonance,the First Affiliated Hospital of Xinxiang Medical University,Weihui,He’nan Province 453100,China)

机构地区：[1]新乡医学院第一附属医院放射科,河南卫辉453100 [2]新乡医学院第三附属医院磁共振科,河南新乡453000 [3]新乡医学院第一附属医院磁共振科,河南卫辉453100

出　　处：《实用放射学杂志》2025年第2期186-189,共4页Journal of Practical Radiology

基　　金：河南省医学科技公关计划联合共建项目(LHGJ20210512)。

摘　　要：目的探讨基于多序列MRI的Transformer模型预测脑胶质瘤患者的异柠檬酸脱氢酶(IDH)突变状态的价值。方法从公开数据集癌症影像档案库中回顾性分析500例脑胶质瘤患者(突变型103例,野生型397例)的多序列MRI资料。通过Transformer深度学习算法进行预测模型构建。受试者工作特征(ROC)曲线的曲线下面积(AUC)被用于评估预测性能,5折交叉被用于预测模型的验证。结果基于Transformer深度学习算法的临床模型、多序列MRI模型以及临床+多序列MRI联合模型可用于预测脑胶质瘤患者IDH突变状态,且与前两者相比,临床+多序列MRI联合模型有最高的诊断效能,AUC为0.904[95%置信区间(CI)0.875~0.928],敏感度和特异度分别为86.41%和86.40%。DeLong检验显示,临床+多序列MRI联合模型与临床模型之间AUC的差异有统计学意义(Z=3.327,P<0.001)。结论基于多序列MRI的Transformer模型能够对IDH突变型和IDH野生型胶质瘤患者进行有效鉴别。Objective To explore the value of the Transformer model based on multi-sequence MRI to predict isocitrate dehydrogenase(IDH)mutation status in patients with glioma.Methods The multi-sequence MRI data of 500 glioma patients(103 mutation-type and 397 wild-type)were analyzed retrospectively from the publicly available dataset Cancer Imaging Archive.The prediction model was constructed through the Transformer deep learning algorithm.Area under the curve(AUC)of the receiver operating characteristic(ROC)curve was used to evaluate the predictive performance,and the five-fold crossover was used for validation of the predictive model.Results The clinical,multi-sequence MRI,and combined clinical+multi-sequence MRI models based on the Transformer deep learning algorithm could be used to predict the IDH mutation status of patients with glioma,and the combined clinical+multi-sequence MRI model had the highest diagnostic efficacy compared with the former two,with an AUC of 0.904[95%confidence interval(CI)0.875-0.928],and the sensitivity and specificity of 86.41%and 86.40%,respectively.DeLong's test showed that the difference in AUC between the combined clinical+multi-sequence MRI model and the clinical model was statistically significant(Z=3.327,P<0.001).Conclusion The Transformer model based on multi-sequence MRI can effectively identify patients with IDH mutation-type glioma and IDH wild-type glioma.

关 键 词：胶质瘤 异柠檬酸脱氢酶 磁共振成像 Transformer模型 

分 类 号：R739.41[医药卫生—肿瘤] R445.2[医药卫生—临床医学]