| 王珍,张声旺,顾潜彪,刘晓,容鹏飞,杨晓,刘晟.CT影像组学评估门静脉高压及预测曲张静脉出血[J].中国医学影像技术,2020,36(12):1896~1899 |
| CT影像组学评估门静脉高压及预测曲张静脉出血 |
| CT radiomics for assessing portal hypertension and predicting variceal veins hemorrhage |
| 投稿时间：2019-08-06 修订日期：2020-11-28 |
| DOI：10.13929/j.issn.1003-3289.2020.12.029 |
| 中文关键词: 高血压,门静脉 出血 影像组学 体层摄影术,X线计算机 |
| 英文关键词:hypertension, portal hemorrhage radiomics tomography, X-ray computed |
| 基金项目: |
|
| 摘要点击次数: 2006 |
| 全文下载次数: 694 |
| 中文摘要: |
| 目的 探讨基于CT影像组学标签评估门静脉高压（PH）的可行性及其预测曲张静脉出血的价值。方法 回顾性分析112例确诊或疑似PH患者，分别以肝静脉压力梯度（HVPG）10、12和20 mmHg作为界值，将患者分为HVPG<10 mmHg组和≥10 mmHg组，HVPG<12 mmHg组和≥12 mmHg组，HVPG<20 mmHg组和≥20 mmHg组，以MaZda软件提取并分析肝脏增强CT门脉期图像纹理特征，观察影像组学判断不同程度PH及曲张静脉出血的效能。结果 CT征象中，在HVPG<10及≥10、HVPG<12及≥12组间有无腹腔积液差异具有统计学意义（P均<0.05）；单层面勾画ROI判别HVPG≥10 mmHg、HVPG≥12 mmHg、HVPG≥20 mmHg及曲张静脉出血的平均最小误诊率分别为12.50%、14.74%、10.27%及22.70%；多层面勾画平均最小误诊率分别为10.27%、11.16%、5.36%及20.49%（P均>0.05）。HVPG预测PH患者曲张静脉出血的曲线下面积（AUC）为0.813[95% CI（0.730，0.897）]，以12 mmHg为HVPG的截断值，误诊率为20.54%，敏感度为77.22%，特异度为69.70%。结论 基于CT影像组学可评估PH程度，且对预测曲张静脉出血有一定价值。 |
| 英文摘要: |
| Objective To explore the feasibility of CT radiomics in evaluating portal hypertension (PH) and the predictive value of variceal hemorrhage. Methods Totally 112 cases of diagnosed or suspected PH were retrospectively collected. Taken 10, 12 and 20 mmHg as the cut-off values of hepatic venous pressure gradient (HVPG), the patients were divided into HVPG<10 mmHg group and ≥10 mmHg group, HVPG<12 mmHg group and ≥12 mmHg group, HVPG <20 mmHg group and ≥20 mmHg group in order. MaZda software was used to extract the texture features of enhanced abdominal CT (portal phase) images, and the radiomics for judging PH and variceal hemorrhage were observed. Results There were significant differences of the presence or absence of ascites between of HVPG<10 mmHg and ≥10 mmHg groups, also between HVPG<12 mmHg and ≥12 mmHg groups (both P<0.05). The average minimum misclassification rate of single-layer ROI delineation for distinguishing HVPG≥10 mmHg, HVPG≥12 mmHg, HVPG≥20 mmHg and variceal hemorrhage was 12.50%, 14.74%, 10.27% and 22.70%, respectively, while the average minimum misclassification rate of multi-layer ROI delineation was 10.27%, 11.16%, 5.36% and 20.49%, respectively (all P>0.05). The area under the curve (AUC) of HVPG for predicting hemorrhage was 0.813 (95% CI[0.730, 0.897]), the misclassification rate was 20.54%, the sensitivity was 77.22% and the specificity was 69.70%. Conclusion The radiomics signatures based on enhanced CT images could assess the degree of PH, therefore having a certain predictive value for variceal hemorrhage. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
