基于MRI动态增强影像组学鉴别肝硬化基础肝内结节良恶性的可行性研究

The Feasibility of Differentiating Benign and Malignant Intrahepatic Nodules in Cirrhosis Based on Dynamic Enhanced MRI

目的探究磁共振成像(MRI)动态增强影像组学鉴别肝硬化(LC)基础肝内结节良恶性的可行性。方法随机选择2015年1月~2022年6月西安医学院第二附属医院收治的慢性肝炎并LC患者(80例)为研究对象,共141个病灶。将病灶按6:4分为训练组(85个)和测试组(56个),各组病灶病理类型无显著差异(P>0.05)。所有患者均于术前4周接受MRI检查。根据病灶的病理结果,分为恶性(89个)和良性(52个)。分析良恶性病灶的临床影像学特征,并建立临床影像模型。应用最小绝对收缩和选择运算符(LASSO)的降维方法获得最优组学特征,应用支持向量机(SVM)建立影像组学模型。最佳诊断模型通过比较临床影像模型、基于MRI平扫图像和MRI增强图像的组学模型的受试者工作特征(ROC)曲线获得。结果恶性病灶组与良性病灶组患者的动脉期高信号、门脉期低/等信号、平衡期低/等信号、肝胆期低信号差异显著(P<0.05),这4种特征联合的诊断模型在训练组和测试组的AUC分别为0.861(95%CI:0.654~1.068)和0.872(95%CI:0.611~1.133)。影像组学模型在训练组和测试组的AUC分别为:基于MRI平扫图像:0.890(95%CI:0.681~1.099)和0.923(95%CI:0.755~1.091);基于MRI动态增强图像:0.954(95%CI:0.872~1.036)和0.901(95%CI:0.746~1.056)。结论对于鉴别LC基础肝内结节的良恶性,基于MRI平扫以及动态增强图像获得影像组学模型是一种高效、非侵入性的诊断方法。这种方法弥补了传统形态学影像的不足,提高了鉴别诊断的准确性,辅助临床治疗,进而改善患者预后。其中基于MRI动态增强图像的影像组学模型的诊断效能最高。

Objective To explore the feasibility of dynamic contrast-enhanced magnetic resonance imaging(MRI)in differentiating benign and malignant intrahepatic nodules in liver cirrhosis(LC).Methods Randomly select 80 patients with chronic hepatitis and LC admitted to The Second Affiliated Hospital of Xian Medical University from January 2015 to June 2022 as the study object,with a total of 141 lesions.The lesions were divided into training group(85 lesions)and test group(56 lesions)according to 6:4,and there was no significant difference in pathological types among the groups(P>0.05).All patients received MRI examination 4 weeks before operation.According to the pathological results,the lesions were divided into malignant(89 lesions)and benign(52 lesions).Analyze the clinical imaging characteristics of benign and malignant lesions,and establish a clinical imaging model.The dimension reduction method of least absolute shrinkage and selection operator(LASSO)was used to obtain the optimal omic features,and support vector machine(SVM)was used to establish the imageomic model.The best diagnostic model was obtained by comparing the receiver operating characteristic(ROC)curves of clinical imaging models,histology models based on MRI plain scan images and MRI enhanced images.Results There were significant differences in arterial phase high signal,portal phase low/equal signal,balance phase low/equal signal and hepatobiliary phase low signal between patients with malignant lesions and patients with benign lesions(P<0.05),and the AUC of the combined diagnostic model of these four characteristics in the training group and the test group was 0.861(95%CI:0.654~1.068)and 0.872(95%CI:0.611~1.133),respectively.The AUC of the imageomics model in the training group and the test group were:Based on MRI plain scan images:0.890(95%CI:0.681~1.099)and 0.923(95%CI:0.755~1.091);Based on MRI dynamic enhanced images:0.954(95%CI:0.872~1.036)and 0.901(95%CI:0.746~1.056).Conclusion To differentiate the benign and malignant of LC based intrahepatic nodules,it is an efficient and non-invasive diagnostic method to obtain an imageomic model based on MRI plain scan and dynamic enhanced images.This method makes up for the shortcomings of traditional morphological images,improves the accuracy of differential diagnosis,assists clinical treatment,and thus improves the prognosis of patients.Among them,the diagnostic efficiency of the imageomics model based on dynamic enhanced MRI images is the highest.