基于CT平扫影像组学深度学习模型预测自发性脑出血早期血肿扩大  

Prediction of early hematoma expansion in spontaneous intracerebral hemorrhage using a deep learningmodel based on CT radiomics

作　　者：丁俊 陈基明 邵颖 丁治民 DING Jun;CHEN Ji-ming;SHAO Ying(Department of Radiology,the First Affiliated Hospital of Wannan Medical College(Yijishan Hospital),Anhui 241001,China)

机构地区：[1]皖南医学院第一附属医院(弋矶山医院)放射科,安徽芜湖241001

出　　处：《放射学实践》2024年第12期1558-1564,共7页Radiologic Practice

基　　金：安徽省卫生健康科研项目(AHWJ2022b044)。

摘　　要：目的:探讨基于CT平扫的影像组学特征构建的深度学习(DLR)模型对自发性脑出血(sICH)早期血肿扩大(HE)的预测价值。方法:回顾性分析2015年1月-2022年12月在本院就诊的350例sICH患者的临床及影像学资料。所有患者发病6h内接受首次头颅CT平扫,并根据24 h内复查CT图像上血肿体积是否超过基线CT图像上的33%或6 mL,将患者分为HE组(136例)和非HE组(214例)。随机将患者以8∶2的比例分为训练组(n=280)和验证组(n=70)。对临床和影像学资料进行组间差异检验,筛选出有统计学意义的临床和影像特征。沿血肿边缘逐层手动勾画感兴趣区(ROI),并融合得到血肿的三维容积感兴趣区(VOI);然后,借助软件沿勾画ROI边缘自动外扩2 mm,得到血肿周围组织ROI。利用One-key AI软件分别提取血肿周围组织的影像组学特征和深度学习特征(基于ResNet-50卷积神经网络),联合这两类特征并进行特征筛选,得到混合特征集。基于临床-影像特征、混合特征及前两类特征联合,利用逻辑回归(LR)、朴素贝叶斯(NB)、K最近邻(KNN)、自适应增强(AdaBoost)和多层感知器(MLP)这五种分类器共构建了15个机器学习模型,采用ROC曲线下面积(AUC)评价各模型的诊断效能,确定最优模型作为输出模型,应用决策曲线分析(DCA)评价最佳模型的临床效益。结果:在临床和常规影像征象中,血清D-二聚体水平、血肿形态、漩涡征、混合征和卫星征在HE组与非HE组之间的差异有统计学意义(P<0.05)。自血肿周围组织共提取得到29个混合特征(15个影像组学特征和14个深度学习特征)。在训练组或验证组中,基于联合特征构建的5种机器学习预测模型的效能均高于临床-影像特征和混合特征构建的模型,尤其以训练组中KNN分类器构建的联合模型的预测效能最高(AUC=0.947,95%CI:0.924~0.970),作为本研究的最佳输出模型。DCA显示阈值在0.025~0.980时KNN联合模型获得的临床Objective:The purpose of this study was aimed to evaluate the efficacy of a deep learning radiomics(DLR)model based on non-contrast CT scans for predicting early hematoma expansion(HE)in patients with spontaneous intracerebral hemorrhage(sICH).Methods:The comprehensive clinical and imaging data of 350 sICH patients treated at our institution from Jan 2015 to Dec 2022 was retrospectively analyzed.All patients underwent an initial head CT scan within 6 hours of symptom onset.Patients were divided into HE group(136 cases)and non-HE group(214 cases)based on whether the hematoma volume increased by more than 33%or 6mL according to a follow-up CT performed within 24 hours.Patients were randomly assigned in an 8∶2 ratio to a training group(n=280)and a validation group(n=70).Clinical and imaging features were compared between the two groups,and statistically significant features were identified.Regions of interest(ROIs)were manually delineated layer by layer along the hematoma border,and three-dimensional volume ROIs(VOIs)were generated.Software was used to automatically expand the ROIs by 2mm to include peri-lesion tissue.Radiomics features and deep learning features(based on ResNet-50 convolutional neural network)of the peri-lesion tissue were extracted using One-key AI software;and then,the features of the two types were combined and selected to create a set of mixed features.Based on clinical and imaging features,mixed features and their combination respectively,15 machine learning(ML)models were developed using five classifiers,including logistic regression(LR),naive Bayes(NB),K-nearest neighbors(KNN),adaptive boosting(AdaBoost)and multilayer perceptron(MLP).The diagnostic efficacy of each model was assessed by the area under the receiver operating characteristic curve(AUC),and the best-performing model was selected as the final output model.Decision curve analysis(DCA)was conducted to evaluate the clinical utility of the optimal model.Results:In the clinical and conventional imaging features,serum D-dimer level,h

关 键 词：深度学习 分类器 影像组学 血肿周围组织 自发性脑出血 血肿扩大 

分 类 号：R814.42[医药卫生—影像医学与核医学] R743.34[医药卫生—放射医学]