基于磁共振平衡式稳态自由进动序列影像组学诊断直肠癌周围淋巴结良恶性效能的研究  

Diagnostic efficacy of MRI balanced steady-state free precession-based radiomics for benign and malignant diagnosis of peripheral lymph nodes in rectal cancer

作　　者：章俊 张海青 徐薇 赵慧[1] 夏鹏 李君君 Zhang Jun;Zhang Haiqing;Xu Wei;Zhao Hui;Xia Peng;Li Junjun(Department of Imaging,Lujiang Branch,Oriental Hospital,Tongji University,Hefei 231500,Anhui,China;Department of General Surgery,Lujiang Branch,Oriental Hospital,Tongji University,Hefei 231500,Anhui,China;Department of Imaging,The First Affiliated Hospital of University of Science and Technology of China,Hefei 230001,Anhui,China)

机构地区：[1]同济大学附属东方医院庐江分院影像科,安徽合肥231500 [2]同济大学附属东方医院庐江分院普外科,安徽合肥231500 [3]中国科学技术大学附属第一医院影像科,安徽合肥230001

出　　处：《结直肠肛门外科》2024年第5期601-606,共6页Journal of Colorectal & Anal Surgery

基　　金：安徽省科技计划项目(1704f0804009);中国红十字基金会2022医学赋能—领航菁英科研项目基金(XM_LHJY2022_05_20);合肥市卫生健康应用医学科研项目(Hwk2023zc022)。

摘　　要：目的探讨基于磁共振平衡式稳态自由进动序列影像组学模型对直肠癌周围淋巴结良、恶性的诊断效能。方法回顾性选取2020年12月至2023年12月于同济大学附属东方医院庐江分院接受治疗的288例直肠癌患者作为研究对象。患者淋巴结良、恶性均经术后病理证实,根据淋巴结性质分为良性淋巴结组(n=129)和恶性淋巴结组(n=159),并按7∶3的比例分为训练组(n=201)和验证组(n=87),分别于常规T2WI序列和平衡式稳态自由进动序列图像进行手动勾画病灶感兴趣区并提取影像组学特征,通过数据降维和筛选出有效特征构建影像组学预测模型;然后,绘制受试者工作特征曲线,并计算曲线下面积、特异度和灵敏度,以量化模型对直肠癌周围淋巴结良、恶性的诊断效能;最后应用决策曲线分析评估其临床应用价值。结果与常规T2WI序列相比,基于核磁共振平衡式稳态自由进动序列影像组学模型能够高效能诊断直肠癌周围淋巴结良、恶性。其中常规T2WI序列影像组学训练组曲线下面积为0.825,特异度为0.726,灵敏度为0.755;验证组曲线下面积为0.877,特异度为0.829,灵敏度为0.832;平衡式稳态自由进动序列影像组学训练组曲线下面积为0.932,特异度为0.950,灵敏度为0.868;验证组曲线下面积为0.884,特异度为0.902,灵敏度为0.845。基于磁共振平衡式稳态自由进动序列影像组学模型各项参数诊断效能均高于常规T2WI序列,同时具有良好的校准度,差异有统计学意义(P<0.05)。结论基于磁共振平衡式稳态自由进动序列影像组学模型能够有效预测直肠癌周围淋巴结的良、恶性,为术前的治疗决策提供了重要的参考依据。Objectives To investigate the diagnostic efficacy of an MRI balanced steady-state free precession-based radiomics model for distinguishing between benign and malignant peripheral lymph nodes in rectal cancer.Methods A retrospective study was conducted on 288 rectal cancer patients treated at Lujiang Branch,Oriental Hospital,Tongji University from December 2020 to December 2023.All patients had histologically confirmed lymph node status,which was used to classify them into the benign lymph node group(n=129)and the malignant lymph node group(n=159).The patients were randomly divided into a training group(n=201)and a validation group(n=87)at a 7∶3 ratio.Radiomic features were extracted from both conventional T2WI sequences and balanced steady-state free precession sequences images by manually delineating region of interest around the lymph nodes.An effective radiomics model was constructed through data dimensionality reduction and feature selection.Receiver operating characteristic curve were plotted,and area under the curve,specificity,and sensitivity were calculated to quantify the diagnostic efficacy of the model in benign and malignant diagnosis of peripheral lymph nodes in rectal cancer.Decision curve analysis were used to evaluate the clinical utility of the model.Results The radiomics model based on balanced steady-state free precession sequences demonstrated superior diagnostic performance compared to conventional T2WI sequences.For the conventional T2WI sequences,the area under the curve,specificity,and sensitivity were 0.825,0.726,and 0.755,respectively,in the training group,and 0.877,0.829,and 0.832,respectively,in the validation group.For the balanced steady-state free precession sequences,the area under the curve,specificity,and sensitivity were 0.932,0.950,and 0.868,respectively,in the training group,and 0.884,0.902,and 0.845,respectively,in the validation group.The MRI balanced steady-state free precessionbased radiomics model showed statistically significant higher diagnostic performance across all

关 键 词：直肠癌 周围淋巴结 磁共振成像 影像组学 平衡式稳态自由进动序列 

分 类 号：R735.3[医药卫生—肿瘤]