基于MRI影像组学对评估术前直肠癌周边纤维脂肪组织浸润的应用价值

Application value of fibrous adipose tissue infiltration in rectal cancer using MRI-based radiomics

目的 评估基于MRI影像组学模型对术前直肠癌周边纤维脂肪组织浸润的应用价值。方法 回顾性分析2019年12月~2022年3月于铜陵市人民医院术后经病理证实直肠癌75例患者术前MRI图像资料,根据病理结果分为纤维脂肪组织浸润组(n=33)及纤维脂肪组织未浸润组(n=42)。基于T2WI及动脉期对比增强T1WI(CE-T1WI)提取影像组学特征,将数据按照7:3的比例分为训练组(n=54)与验证组(n=21),采用最大相关最小冗余和最小绝对收缩和选择算子回归的方法对训练集数据进行特征降维,并将特征降维后作为特征向量,筛选出最佳的影像组学特征,构建出T2WI模型、CE-T1WI模型及T2WI+CE-T1WI模型3种影像组学模型,根据ROC曲线确定其中最优模型;根据肿瘤上下径、肿瘤最厚径、CE-T1WI相对信号强度比值(RCE-T1WI)、T2WI相对信号强度比值(RT2)、DWI相对信号强度比值(RDWI)及ADC值多因素Logistic回归分析构建出联合模型;评估最优模型与联合模型二者效能及临床效益。结果 基于T2WI和CE-T1W1各提取851个影像组学特征,最终分别获得10个和14个最佳特征,用于构建T2WI+CE-T1WI模型。基于T2WI模型、CE-T1W1模型及T2WI+CE-T1WI模型在训练组中对判定直肠癌周边纤维脂肪组织浸润的AUC分别为0.87(95%CI:0.75~0.95)、0.82(95%CI:0.70~0.91)和0.84(95%CI:0.71~0.92),在验证组中分别为0.76(95%CI:0.53~0.92)、0.73(95%CI:0.50~0.90)和0.76(95%CI:0.52~0.92)。采用联合模型在训练组中对直肠癌纤维脂肪组织浸润的AUC为0.91(95%CI:0.80~0.97),在验证组中AUC为0.80(95%CII:0.57~0.94)。决策曲线分析提示,阈值取0~1时,联合模型的净收益高于T2WI模型。结论 基于MRI影像组学的联合模型对评估术前直肠癌周边纤维脂肪组织浸润的效能较高。

Objective To evaluate the value of preoperative prediction of fibrous adipose tissue infiltration around rectal cancer using MRI-based radiomics.Methods The preoperative MRI of 75 patients with rectal cancer confirmed by pathology in The People's Hospital of Tongling City from December 2019 to March 2022 were analyzed retrospectively.According to the pathological results,the patients were divided into two groups:the fibrous adipose tissue infiltration group(n=33)and the fibrous adipose tissue noninfiltration group(n=42).The image group features for T2WI and arterial phase contrast-enhanced T1-weighted imaging(CE-T1WI)were extracted.The data were divided into a training group(n=54)and a verification group(n=21)at a ratio of 7:3.The training set data were reduced by maximum correlation minimum redundancy and LASSO regression,and feature dimension reduction was used to select the optimal imaging features.Three image group models(T2WI model,CE-T1WI model,T2WI+CE-T1WI model)were constructed.The best model had been selected in the three models.Based on the ROC curves and multivariate Logistic regression analysis in accordance with such parameters as upper and lower diameter of tumour,thickest diameter of tumour,unsmoothed edges,blurred boundary,CE-T1WI relative signal intensity(RCE-T1WI),T2WI relative signal intensity(RT2),DWI relative signal intensity(RDWI)and ADC,a combined model was constructed,and its predictive efficiency and clinical benefits were evaluated.Results In total,851 imaging features were extracted from T2WI and CE-T1WI,and 10 and 14 optimal features were obtained,respectively,which were used to construct the combined model.In the training group,the AUC for the T2WI model,CE-T1WI model and T2WI+CE-T1WI model for predicting fibrous adipose tissue infiltration around rectal cancer were 0.87(95%CI:0.75-0.95),0.82(95%CI:0.70-0.91)and 0.84(95%CI:0.71-0.92),respectively.In the verification group,the AUC were 0.76(95%CI:0.53-0.92),0.73(95%CI:0.50-0.90)and 0.76(95%CI:0.52-0.92),respectively.The T2WI modle was the best model.For the combined model,the AUC for predicting fibrous adipose tissue infiltration of rectal cancer was 0.91(95%CI:0.80-0.97)in the training group and 0.80(95%CI:0.57-0.94)in the verification group.Decision curve analysis suggested that when the threshold was 0:1,the net return of combined model was higher than that of the T2WI model.Conclusion The combined model of MRI-based radiomics is effective in inevaluateing the preoperative of fibrous adipose tissue infiltration in rectal cancer.