摘要
目的:探讨基于术前双期增强CT的影像组学分析结合支持向量机(SVM)方法对肝细胞癌(HCC)术后早期复发的预测价值。方法:回顾性分析经病理证实且术前行上腹部CT增强扫描的130例HCC患者的临床及影像资料。对所有患者在术后进行定期随访,术后2年内有复发或转移者61例(早期复发组),无早期复发者69例。对2组患者的临床资料进行组间单因素及多因素分析。应用ITK-SNAP软件对目标病灶进行分割,采用美国GE AK软件提取病灶的组学特征。对所有特征之间进行Pearson相关分析以降低特征冗余,随后采用t检验及Wilcoxon秩和检验降低特征维度。将所有患者按2∶1的比例分为训练集(84例)和验证集(46例),利用SVM算法构建基于临床资料、动脉期及静脉期组学特征的预测模型,使用ROC曲线评价各模型预测HCC术后早期复发的效能。结果:单因素及多因素分析筛选出临床资料中肿瘤最大径、微血管侵犯(MVI)为早期复发的独立危险因素。经特征降维后共筛选出4个动脉期和2个静脉期组学特征参数,以动脉期模型的预测效能更佳,在验证集中的AUC为0.858。结合有效临床因子和动脉期最优特征建立的组合模型的预测效能最高,其在验证集中的AUC为0.938。结论:基于术前增强CT图像的影像组学模型能较好地预测HCC切除术后早期复发,可作为临床个性化治疗的辅助工具。
Objective:To explore the value of radiomics analysis based on preoperative dual-phases enhanced CT combined with support vector machine(SVM)method in predicting early recurrence after resection of HCC.Methods:The clinical and imaging data of 130 patients with HCC confirmed by pathology and underwent preoperative contrast enhanced CT scan of upper abdomen were analyzed in our study retrospectively.All patients were followed up regularly after surgery.61 patients were found recurrence or metastasis within 2 years after surgery(early recurrence group),and no early recurrence or metastasis was found in 69 patients(non-recurrence group).Univariate and multivariate analyses were used to analyze the clinical data.The target lesions were segmented using ITK-SNAP software,and AK software of the United States was used to extract the features.Pearson correlation analysis was performed for features,then t-test and Wilcoxon U-test were used to reduce the dimension of the features.According to the ratio of 2∶1,the patients were divided into training set(n=84)and test set(n=46).Predictive models were established by SVM method based on features of clinical,arterial and venous phases.ROC curve was used to evaluate the diagnostic efficacy of the model in predicting early recurrence of HCC.Results:Univariate and multivariate analysis showed that the maximum tumor diameter and microvascular invasion(MVI)were independent risk factors for early recurrence.Four radiomics features of arterial phase and two of venous phase were finally selected out.The arterial phase model showed better prediction efficacy,with AUC of 0.858 in the validation set.Finally,the efficacy of the combined model integrated with the effective clinical factors and features of arterial phase was improved,with the AUC of 0.938 in the validation set.Conclusion:Preoperative enhanced CT radiomics model can be used to predict early postoperative recurrence of HCC,and as a useful aid for clinical personalized treatment.
作者
熊亮霞
方淇民
李淑豪
彭云
龚良庚
XIONG Liang-xia;FANG Qi-min;LI Shu-hao(Department of Radiology,the Second Affiliated Hospital of Nanchang University,Nanchang 330006,China)
出处
《放射学实践》
CSCD
北大核心
2022年第4期432-436,共5页
Radiologic Practice
基金
国家自然科学基金(81860316)。
关键词
肝细胞癌
早期复发
影像组学
支持向量机
体层摄影术
X线计算机
Hepatocellular carcinoma
Early recurrence
Radiomics
Support vector machine
Tomography,X-ray computed
