摘要
目的 探讨生境分析在预测高强度聚焦超声(HIFU)对子宫肌瘤消融疗效评估中的价值。方法 选取110例子宫肌瘤患者的影像学及临床资料,其中充分消融组32例、非充分消融组78例,按照7∶3划分为训练集(77例)、测试集(33例)。在ITK-SNAP(http://www.itksnap.org/)平台上对子宫肌瘤进行逐层勾画并在此基础上生成生境区域,生境区域聚类数为3类,提取肌瘤瘤内(ROI_(intra))影像组学特征及3个亚区域Habitat1、Habitat2、Habitat3影像组学特征,利用相关系数、t检验及最小绝对收缩和选择算子算法对特征进行筛选,构建放射组学模型,最终使用受试者工作特征(ROC)曲线下面积(AUC)、准确度、灵敏度、特异度及F1分数进行模型效能评价,DeLong检验用于评估不同模型之间AUC差异,并绘制决策曲线(DCA)比较模型临床效能。结果 两组肿瘤最大径差异有统计学意义(P=0.004),其他影像学及临床资料差异无统计学意义(P>0.05),Habitat3建立的模型优于Habitat1、Habitat2及ROI_(intra)建立的模型,其中Habitat3基于逻辑回归建立的模型最佳,DeLong检验显示Habitat3与ROI_(intra)在LR训练集AUC比较差异无统计学意义(P>0.05),在测试集中比较差异有统计学意义(P=0.005),在训练集和测试集中的AUC、准确度、灵敏度、特异度和F1分数分别为0.903、0.805、0.774、0.875、0.845和0.875、0.788、0.760、0.875、0.844,且DCA显示Habitat3利用逻辑回归模型的临床净获益率最大。结论 生境分析在预测子宫肌瘤HIFU后消融疗效评估中具有良好的效能。
Objective To investigate the application value of habitat analysis in predicting the therapeutic effect of high-intensity focused ultrasound in uterine fibroids.Methods A retrospective analysis was conducted on the imaging and clinical data of 110 patients with uterine fibroids admitted to our hospital.Among them,32 and 78 patients were included in the sufficient and insufficient ablation groups,respectively.They were then divided into a training set(77 cases)and a test set(33 cases)at a ratio of 7:3.The images were preprocessed and the uterine fibroids were outlined on the ITK-SNAP platform for ROI segmentation.These ROI segmentations were clustered into three subregions.Radiomics features were extracted from the the fibroid(ROIintra)and three subregions:Habitat1,Habitat2,and Habitat3.Features were screened using the correlation coefficient,t-test,minimum absolute shrinkage,and selection operator,and a radiation genomics model was constructed.The model performance was evaluated using the area under the ROC curve,accuracy,sensitivity,specificity,and F1 score.The DeLong test was used to evaluate the differences between area under the ROC curves among different models,and decision-making curves were used to compare the clinical efficacy of the models.Results There was a statistically significant difference in the maximum tumor diameter between the two groups(P=0.004),whereas no statistically significant differences were observed in other imaging and clinical information(P>0.05).The model constructed using Habitat3 outperformed those established using Habitat1,Habitat2,and ROIintra.Among them,Habitat3,based on logistic regression(LR),was the best model.The DeLong test showed no significant difference in the area under the ROC curve between Habitat3 and ROIintra in the LR training set(P>0.05)and significant difference in the testing set(P=0.005).The area under the ROC curve,accuracy,sensitivity,specificity,and F1 score in the training and test sets were 0.903,0.805,0.774,0.875,0.845,and 0.788,0.760,0.875,0.844,respectively.The decision curve analysis curve showed that Habitat3 had the maximum net clinical benefit rate using the LR model.Conclusion Habitat analysis combined with LR machine learning can effectively predict the postoperative ablation of high-intensity focused ultrasound on uterine fibroids.
作者
陶响
陈芳
陆惠娟
刘梦玥
杨雅敏
顾清华
TAO Xiang;CHEN Fang;LU Huijuan;LIU Mengyue;YANG Yamin;GU Qinghua(Department of Radiology,Yongding Hospital,Suzhou 212500,China;Department of Gynecology and obstetrics,Yongding Hospital,Suzhou 212500,China)
出处
《医学影像学杂志》
2025年第7期109-114,共6页
Journal of Medical Imaging
关键词
子宫肌瘤
生境分析
影像组学
消融疗效
Uterine fibroids
Habitat analysis
Imaging genomics
Ablation
