摘要
目的探索基于磁共振T_(1)WI和对比增强T_(1)WI序列(CE-T_(1)WI)的U-net模型对脑胶质瘤患者的预后评估价值。方法回顾性分析316例脑胶质瘤患者的T_(1)WI、CE-T_(1)WI影像数据和临床资料,其中,低风险组163例,高风险组153例。采用U-net深度学习算法建立预测模型。受试者工作特征曲线下面积(AUC)被用于评价预测性能。结果基于U-net算法的T_(1)WI模型、CE-T_(1)WI模型和多序列联合模型均可用于预测脑胶质瘤患者的生存风险,且与前两者相比,多序列联合模型有最高的诊断效能,训练集中AUC为0.913(95%CI:0.871~0.944),敏感度和特异度分别为90.16%和91.54%,验证集中AUC为0.883(95%CI:0.778~0.950),敏感度和特异度分别为90.32%和93.94%。DeLong检验显示,训练集中多序列联合模型与T_(1)WI模型之间AUC的差异具有统计学意义(Z=1.983,P<0.05)。结论基于T_(1)WI和CE-T_(1)WI的U-net深度学习模型能够预测脑胶质瘤患者的预后生存率。
Objective To explore the value of U-net model based on magnetic resonance T_(1)weighted imaging(T_1)WI and contrast enhanced T_(1)weighted imaging sequence(CE-T_(1)WI)for prognostic evaluation of glioma patients.Methods T_(1)WI and CE-T_(1)WI imaging data and clinical information of 316 patients with gliomas were retrospectively analyzed,of which 163 were in the low-risk group and 153 in the high-risk group.We established a prediction model based on the U-net algorithm.The area under the subject operating characteristic curve(AUC)was used to evaluate the predictive performance.Results The T_(1)WImodel based on U-net algorithm,the CE-T_(1)WI model,and the multi sequence combined model could be used to predict the survival risk of patients with glioma,and the multi sequence combined model had the highest diagnostic efficacy compared with the former two,with an AUC of 0.913(95%CI:0.871-0.944),and the sensitivity and specificity of 90.16%and 91.54%,respectively in the training set.The validation set AUC was 0.883(95%CI:0.778-0.950),with sensitivity and specificity of 90.32%and 93.94%,respectively.DeLong's test showed that the difference in AUC between the multi sequence combined model and the T_(1)WI model was statistically significant in training set(Z=1.983,P<0.05).Conclusion The U-net deep learning model based on T_(1)WI and CE-T_(1)WIcan predict the prognostic survival of glioma patients.
作者
段金辉
刘苏娟
张梦
周凤梅
李自强
闫瑞芳
DUAN Jinhui;LIU Sujuan;ZHANG Meng(Department of Magnetic Resonance,the First Affiliated Hospital of Xinxiang Medical University,Weihui,Henan Province 453100,P.R.China)
出处
《临床放射学杂志》
北大核心
2025年第4期606-610,共5页
Journal of Clinical Radiology
基金
河南省医学科技公关计划联合共建项目(编号:LHGJ20230505)
河南省高等学校重点科研项日计划(编号:24B320017)
新乡医学院第一附属医院青年培育基金项目(编号:QN-2022-B11)。
