DWI影像组学模型预测子宫内膜癌微卫星不稳定状态:与ADC值的对比研究

Value of diffusion-weighted imaging radiomics model in the prediction of microsatellite instability in endometrial cancer:a comparative study with ADC

目的:探讨基于磁共振DWI的影像组学模型对子宫内膜癌(EC)微卫星不稳定(MSI)状态的预测价值。方法:回顾性分析2019年5月-2023年1月在本院确诊为EC的81例患者的DWI资料。其中,MSI组29例,微卫星稳定组(MSS)52例。在DWI图像上沿病变边缘逐层勾画ROI后生成容积ROI(VOI)并提取影像组学特征,并在生成的ADC图像上测量病灶的ADC值。采用最小绝对收缩和选择算子(LASSO)和SelectKBest算法进行组学特征的筛选,然后采用决策树(DT)分析方法构建组学预测模型。使用受试者工作特征(ROC)曲线评估模型的诊断效能,采用Delong检验比较影像组学模型与ADC之间诊断效能的差异。基于1000次采样的Bootstrap算法和校准曲线来验证预测模型的临床应用价值。结果:MSI组的ADC值小于MSS组(P=0.008)。模型构建方面,共筛选出5个最优DWI影像组学特征(2个一阶统计特征、1个直方图灰度共生矩阵特征、1个灰度共生矩阵特征和1个灰度游程长度矩阵特征)用于建立预测模型。诊断效能方面,DWI组学预测模型的AUC为0.927(95%CI:0.847~0.973),较ADC值的诊断效能(AUC=0.771,95%CI:0.664~0.857)显著增加(Z=2.436,P=0.015)。模型验证方面,在基于Bootstrap算法的验证中,DWI组学预测模型亦显示出较高的效能,AUC为0.904(95%CI:0.885~0.916);同时,校准曲线显示该模型的预测值与实际观测值之间有较好的一致性。结论:基于DWI影像组学特征构建的预测模型较ADC值能更好地对EC患者的MSI状态进行术前评估,有望为临床诊疗提供一种新的选择。

Objective:To explore the ability of diffusion weighted imaging(DWI)radiomics model to predict microsatellite-instability(MSI)status in endometrial cancer(EC).Methods:The DWI data of 81 patients[29 in the MSI group and 52 in the microsatellite stability(MSS)group]diagnosed as EC at the 7th P”eople's Hospital of Zhengzhou from May 2019 to January 2023 were retrospectively analyzed.ROIs were outlined layer by layer along the lesion edge on the DWI images,”radiomics features were extracted,and the apparent diffusion coefficient(ADC)values of the lesions were measured on the generated ADC images.The least absolute shrinkage and selection operator(LASSO)and SelectKBest algorithm were used for feature screening.Decision tree(DT)analysis was used for predictive model,and receiver operating characteristic(ROC)curve was used for diagnostic efficacy assessment.The Delong test was used to compare the difference in diagnostic efficacy between radiomics model and ADC values.A Bootstrap algorithm based on 1000 samples and calibration curves were used to validate the clinical application value of the predictive model.Results:For parameter comparison,the ADC value of the MSI group was smaller than that of the MSS group(P=0.008);for model construction,a total of five optimal DWI radiomics features including two first-order statistical features,one histogram grey-scale covariance matrix feature,one grey-scale covariance matrix feature,and one grey-scale tour length matrix feature histological features were screened to build a prediction model;in terms of diagnostic efficacy,the diagnostic efficacy of the DWI radiomics prediction model(AUC=0.927,95%CI:0.847~0.973)was significantly increased(Z=2.436,P=0.015)compared to the ADC value(AUC=0.771,95%CI:0.664~0.857);and in terms of model validation,the predictive model also showed high performance in Bootstrap algorithm-based validation,with an AUC of 0.904(95%CI:0.885~0.916),while the calibration curve showed good agreement between the predicted and actual observed values of the model.Conclusion:The prediction model constructed on the basis of DWI radiomics features provides a better preoperative assessment of MSI status in EC patients than ADC values,and is expected to provide a new option for clinical management.