摘要
目的:旨在评估CT和MRI影像在诊断肝癌射频消融术后复发情况的有效性。方法:通过对27名肝癌患者进行射频消融术后的31个结节进行回顾性分析,本文对增强CT和MRI T1序列延迟期图像提取影像组学特征,并进行数据处理和特征选择。采用影像组学方法,结合多种机器学习模型,对增强CT和MRI T1序列图像进行分析。通过准确率、F1值、精确率、召回率/敏感度、特异度和ROC曲线下面积(AUC)对模型进行评估。结果:CT图像的1334个组学特征中,经过删除和特征降维,最终保留6个关键特征。而MRI图像则从相同数量的特征中保留10个关键特征。这一对比结果显示出CT在特征简洁性上的优势。在测试集上,随机森林模型表现最佳,CT图像的ROC曲线下面积为0.92,而MRI图像为0.83。结论:增强CT和MRI在对于肝癌射频消融术后患者复发情况的诊断均具有重要价值。在影像组学特征中CT图像相较于MRI T1序列对肝癌射频消融术后复发情况的诊断有相对更高的准确率和应用价值。
Objective:The study aims to evaluate the effectiveness of computed tomograph(CT)and magnetic resonance imaging(MRI)imaging in diagnosing the recurrence of hepatocellular carcinoma after radiofrequency ablation.Method:A retrospective analysis was conducted on 31 nodules from^(2)7 patients with hepatocellular carcinoma who underwent radiofrequency ablation.This study extracted radiomic features from enhanced CT and MRI T1 sequence delayed phase images and performed data processing and feature selection.Feature reduction was conducted using the absolute shrinkage and selection operator(LASSO)regression model,and hyperparameters of various machine learning models(Logistic regression,support vector machine,random forest,LightGBM)were adjusted using five-fold cross-validation and Bayesian optimization methods.The models were evaluated based on accuracy,F1 score,precision,recall/sensitivity,specificity,and the area under the receiver operator characteristic(ROC)curve(AUC).Results:Of the 1334 radiomic features of CT images,after deletion and feature reduction,6 key features were retained,showcasing their comparative simplicity compared to the 10 features retained from MRI images from the same number of initial features.The random forest model performed best in the test set,with the AUC for CT images being 0.92 and for MRI images 0.83.Conclusion:Enhanced CT and MRI are both valuable in diagnosing the recurrence of hepatocellular carcinoma after radiofrequency ablation.CT images show relatively higher accuracy and practical value in the diagnosis of post-ablation recurrence compared to MRI T1 sequences.
作者
陈飞
胡羽澄
吴桐
郑凯文
张雪君
郭丽
CHEN Fei;HU Yucheng;WU Tong;ZHENG Kaiwen;ZHANG Xuejun;GUO Li(Tianjin Medical University,Tianjin 300203,P.R.China)
出处
《影像科学与光化学》
CAS
2024年第3期177-186,共10页
Imaging Science and Photochemistry
基金
天津市教委科研项目(2019KJ181)。
关键词
肝癌
射频消融
诊断
影像组学
机器学习
hepatocellular carcinoma
radiofrequency ablation
diagnosis
radiomics
machine learning
