摘要
目的准确的T分期有助于直肠癌个体化治疗,探讨MRI多参数影像组学模型在预测直肠癌新辅助治疗后T分期(pT)中的预测价值。方法回顾性分析基线期进行3 T MRI检查、接受新辅助治疗后行直肠癌根治性切除手术的171例直肠癌患者资料,搜集基线期临床特征及术后病理T分期。依据病理T分期将患者分为分期较低组(pT 0~2)和分期较高组(pT 3~4)。使用ITK-SNAP软件分别在高分辨率T_(2)WI、轴位T_(1)WI增强扫描图像上,采用最小轮廓法逐层手动绘制直肠癌瘤灶作为感兴趣区(ROI),使用Pyradiomics软件提取ROI中所有影像组学特征,通过组内相关系数(ICC)分析保留稳定性较好(ICC≥0.75)的特征。采用最小绝对紧缩与选择算子(LASSO)方法,分别从T_(1)WI、T_(2)WI及融合特征(包括T_(1)WI、T_(2)WI、临床特征)中筛选出与pT相关的特征。统计学分析筛选出与pT具有相关性的临床特征。将筛选得到的T_(1)WI影像组学特征、T_(2)WI影像组学特征、临床特征及融合特征通过逻辑回归(LR)方法分别构建pT的预测模型,包括临床模型、T_(1)WI影像组学模型、T_(2)WI影像组学模型及融合模型,融合模型纳入4个临床特征(最大淋巴结短径、壁外血管侵犯、基线期T、N分期)、9个T_(1)WI组学特征、12个T_(2)WI组学特征。使用受试者工作特征曲线(ROC)和校准曲线评估模型的性能;Delong检验比较模型之间的差异;决策分析曲线(DCA)模型的临床应用价值。结果训练集中T_(1)WI、T_(2)WI、临床模型及融合模型的曲线下面积(AUC)分别是0.868、0.921、0.713、0.967,测试集中分别为0.761、0.842、0.689、0.932。T_(1)WI及T_(2)WI模型效能相当且高于临床模型,融合模型具有最佳的预测效能。结论直肠癌的临床影像学表现、T_(1)WI、T_(2)WI影像组学特征均可预测直肠癌pT分期,基于MRI影像组学多参数模型融合临床特征,可以提高预测pT分期的准确性。
Objective Accurate T staging is helpful for individualized treatment of rectal cancer.The purpose of this study was to investigate the value of MRI multiparameter radiomics model in predicting posttreatment T(pT)stage of rectal cancer after neoadjuvant therapy.Methods This study retrospectively analyzed the pretreatment clinical featuresn and pT stage of 171 rectal cancerpatients,which confirmed by pathological who underwent 3 T MRI examination before neoadjuvant therapy and radical resection after neoadjuvant therapy.According to the pathological T stage,the patients were divided into lower stage group(pT 0-2)(n=71)and higher stage group(pT 3-4)(n=100).ITK-SNAP software was used to manually draw the rectal cancer tumor as the region of interest(ROI)layer by layer on the images of T_(2)WI and enhanced T_(1)WI,respectively.Pyradiomics software was used to extract all radiomics features in ROI.The characteristics with good retention stability(ICC≥0.75)were analyzed by intraclass correlation coefficient(ICC).The minimum absolute constriction and selection operator(LASSO)method was used to screen the most relevant features of pT from T_(1)WI,T_(2)WI and fusion fea tures including T_(1)WI,T_(2)WI and clinical features.Clinical features associated with pT were screened out by statistical analysis.The selected T_(1)WI radiomics features,T_(2)WI radiomics features,clinical features and fusion features were used to construct the pT prediction models by logistic regression(LR)method,including clinical model,T_(1)WI radiomics model,T_(1)WI radiomics model and fusion model.The fusion model was constructed by nine T_(1)WI,twelve T_(2)WI radiomics features and four clinical features[including maximum lymph node short diameter(Lnmax),extra-intestinal vascular invasion(EMVI),baselineT stage(rT)and N stage(rN)].The performance of the models were evaluated by receiver operating characteristic(ROC)curve and calibration curve.Delong test was used to compare the differences between the models.The decision curve was used to evaluate the clinical application value of the models.Results In the training set,the area under the curve(AUC)of T_(1)WI,T_(2)WI,clinical model and fusion model were 0.868,0.921,0.713 and 0.967,respectively,and the corresponding AUC were 0.761,0.842,0.689 and 0.932 in the test set.The efficacy of T_(1)WI and T_(2)WI models were comparable and higher than that of clinical model,and the fusion model achieved the best predictive efficacy.Conclusion The clinical imaging findings,T_(1)WI and T_(2)WI radiomics features of rectal cancer can predict pTstage after neoadjuvant therapy,and the accuracy of predicting pTstage can be improved by multi-parameter MRI radiomics model incorporating clinical features.
作者
郭小芳
袁子龙
汪博泉
徐海波
聂婷婷
GUO Xiaofang;YUAN Zilong;WANG Boquan(Department of Radiology,Hubei Cancer Hospital,Wuhan,Hubei Province 430000,P.R.China)
出处
《临床放射学杂志》
北大核心
2023年第12期1939-1945,共7页
Journal of Clinical Radiology
基金
国家癌症中心攀登基金临床研究重点课题项目(编号:NCC201917B05)
湖北省肿瘤医院生物医学中心专项科研基金项目(编号:2022SWZX06)。
关键词
磁共振成像
影像组学
直肠癌
T分期
Magnetic resonance imaging
Radiomics
Rectal cancer
T staging
