摘要
目的 研究基于临床-超声特征及MRI影像组学参数建模用于术前预测胎盘植入程度的能力,开发基于组合模型的预测工具并评估其诊断效能。方法 回顾性分析2015年3月—2021年11月本院124例确诊胎盘植入孕妇的临床病理学特征、产前超声图像和MRI影像学特征,并提取MRI影像组学纹理参数;依据病理结果分为胎盘植入组(PI组)69例,胎盘穿透组(PP组)55例。以7/3的比例建立训练集和测试集,在训练集中,采用二元logistic回归分析建立超声/MRI影像学模型、临床资料模型、MRI影像组学模型及组合模型预测胎盘植入程度,并用Delong非参数检验比较分析不同模型的预测效能,然后建立决策曲线测试模型净收益;通过以上4种模型评估测试集患者的预测效果,最终验证开发模型工具的预测效能。结果 二元logistic回归分析确定了流产史、子宫内膜损伤史、胎盘和子宫肌层之间的边界模糊、子宫浆膜与膀胱边界模糊、MRI纹理特征弧度区域大小矩阵(SmallAreaHighGrayLevelEmphasis和RunVariance),邻域灰度差矩阵(Contrast)等为风险因素,在训练集中,分别构建了预测胎盘植入程度的组合模型(曲线下面积AUC=0.885;95%CI 0.817~0.954,P<0.05),临床模型(AUC=0.744;95%CI 0.644~0.844,P=0.0007),超声/MRI影像学模型(AUC=0.748;95%CI 0.646~0.831,P=0.0071),MRI影像组学模型(AUC=0.728;95%CI 0.626~0.831,P=0.0045),证实组合模型预测效能最高。在测试集中,组合模型也显示出更高的预测效能。结论 超声/MRI的影像学特征及纹理分析是预测胎盘植入程度的有力指标。基于临床-影像学资料建立的组合模型可提高预测胎盘植入程度的准确性。
Objective To study the ability of clinical data-ultrasound features-MRI radiomics based model for predicting the degree of placental implantation before operation,and develop a prediction tool based on combined model and evaluate its diagnostic efficiency.Methods From March 2015 to November 2021,the clinicopathological features,prenatal ultrasound images and MRI images of 124 pregnant women with placental implantation confirmed in Xiangyang First People′s Hospital were analyzed retrospectively.According to the pathological results,they were divided into 69 cases in placenta increta group(PI Group) and 55 cases in placenta percreta group(PP group).And then the training set and test set were constructed with the proportion of 7/3.In training set,the ultrasound/MRI imaging model,clinical data model,MRI radiomics model and combined model were established by binary logistic regression analysis to predict the degree of placental implantation.The prediction efficiency of different models was compared and analyzed by Delong nonparametric test,and then the net benefit of decision curve test model was obtained;Prediction effect were also confirmed in the test set using the above 4 models,and finally the prediction efficiency of the developed model tool was verified.Results Binary logistic regression analysis identified the history of abortion,the history of endometrial injury,the blurred boundary between placenta and myometrium,and the blurred boundary between uterine serosa and bladder as risk factors.In the training set,The area under the curve(AUC) of combined models was confirmed best [(AUC) =0.885;95%CI 0.817-0.954,P<0.05],clinical model[(AUC)=0.744;(95%CI 0.644-0.844),P=0.0007],ultrasound/MRI imaging model [(AUC)=0.748;(95%CI 0.646-0.831),P=0.0071] and MRI radiomics model[(AUC)=0.728;(95%CI 0.626-0.831),P=0.0045].In the test set,the combined model also showed higher prediction efficiency.Conclusion The imaging features and texture analysis of ultrasound/MRI are a powerful index to predict the degree of placental implantation.The combined model based on clinical-imaging data can improve the accuracy of predicting the degree of placental implantation.
作者
宋萍
王勇
安鹏
SONG Ping;WANG Yong;AN Peng(Department of Radiology,Xiangyang First People′s Hospital Affiliated to Hubei university of Medicine,Xiangyang 441000,Hubei,China)
出处
《西部医学》
2024年第1期135-141,共7页
Medical Journal of West China
基金
襄阳市科技局2020年卫生科技项目(2020YL30)。
关键词
MRI
影像组学
产前超声
预测模型
胎盘植入
粘连性胎盘
穿透性胎盘
MRI
Radiomics
Prenatal ultrasound
Prediction model
Placenta implantation
Placenta accreta
Placenta percreta
