目的:探讨宫颈癌术前磁共振T2WI矢状位图的影像特征对近期预后的预测作用,构建并验证SVM预测模型.方法:回顾性选取2020年6月至2022年6月在我院接受宫颈癌手术治疗的患者80例,统计患者2年内预后情况,根据患者预后情况分为良好组(n=40)和不良组(n=40),再按7:3比例分为建模集(n=56)和验证集(n=24),收集患者术前MR-T2WI矢状位图的影像组学特征,使用单变量曲线下面积(Area under curve,AUC)分析及五折交叉验证的最低绝对收缩和选择算子LASSO回归算法进行特征筛选,以此构建SVM支持向量机预测模型.结果:SVM支持向量机结果显示,影响近期预后不良的前6位特征是灰度游程矩阵运行熵、灰度尺寸区域数量、灰度共生矩阵差异熵、一阶特征平均绝对偏差、运行长度不均匀度标准化、最大行2D直径,模型AUC为0.765,最佳截断值0.536对应的灵敏度、特异度分别为0.667、0.828.结论:基于宫颈癌术前T2WI影像组学特征构建的SVM支持向量机模型具有较好的预测效能,可为临床预防宫颈癌术后预后不良提供参考.
Prediction of Short-term Prognosis Based on Preoperative T2WI Imaging Features of Cervical Cancer
Objective:To investigate the predictive effect of sagittal bitmap of preoperative MRI T2WI on the short-term prognosis of cervical cancer,and construct and validate SVM prediction model.Methods:Eighty patients who received surgical treatment for cervical cancer in our hospital from June 2020 to June 2022 were selected retro-spectively.The prognosis of the patients within 2 years was statistically analyzed,and the patients were divided in-to good group(n=40)and bad group(n=40)according to their prognosis,and then divided into modeling set(n=56)and validation set(n=24)according to the ratio of 7:3.The image omics features of the preoperative MR-T2WI sagittal bitmap were collected,and the LASSO regression algorithm,the minimum absolute contrac-tion and selection operator of univariate Area under curve(AUC)analysis and 50％cross validation,was used for feature screening,and SVM support vector machine prediction model was constructed.Results:SVM support vector machine results showed that the top 6 features that affected the short-term prognosis were the running en-tropy of gray run matrix,the number of gray scale size regions,the difference entropy of gray scale co-occur-rence matrix,the average absolute deviation of first-order features,the standardization of run length inhomoge-neity,and the maximum 2D row diameter.The model AUC was 0.899.The sensitivity and specificity correspond-ing to the optimal cut-off value 0.536 were 0.667 and 0.828,respectively.Conclusion:The SVM support vector machine model constructed based on the preoperative T2WI imaging features of cervical cancer has good predictive efficacy,and can provide a reference for clinical prevention of poor prognosis after cervical cancer surgery.
cervical cancerT2WIimaging omicsimmediate prognosisprediction model
万方数据
维普
