| 翟亚楠,庄辛,李建林,郭顺林.构建模型预测二维自旋回波-平面回波MR肝脏弹性成像图像质量[J].中国医学影像技术,2023,39(11):1684~1688 |
| 构建模型预测二维自旋回波-平面回波MR肝脏弹性成像图像质量 |
| Construction model for predicting imaging quality of liver two-dimensional spin-echo and planar-echo MR elastography |
| 投稿时间：2023-06-27 修订日期：2023-10-13 |
| DOI：10.13929/j.issn.1003-3289.2023.11.020 |
| 中文关键词: 肝 弹性成像技术 磁共振成像 图像质量 |
| 英文关键词:liver elasticity imaging techniques magnetic resonance imaging image quality |
| 基金项目:兰州大学第一医院院内基金资助项目(ldyyyn2021-104)。 |
| 作者 | 单位 | E-mail | | 翟亚楠 | 兰州大学第一医院放射科, 甘肃 兰州 730000
甘肃省放射影像临床医学研究中心, 甘肃 兰州 730000
甘肃省智能影像医学工程研究中心, 甘肃 兰州 730000
甘肃省智能影像医学行业技术中心, 甘肃 兰州 730000
精准影像协同创新甘肃省国际科技合作基地, 甘肃 兰州 730000 | | | 庄辛 | 兰州大学第一医院放射科, 甘肃 兰州 730000
甘肃省放射影像临床医学研究中心, 甘肃 兰州 730000
甘肃省智能影像医学工程研究中心, 甘肃 兰州 730000
甘肃省智能影像医学行业技术中心, 甘肃 兰州 730000
精准影像协同创新甘肃省国际科技合作基地, 甘肃 兰州 730000 | | | 李建林 | 兰州大学第一医院放射科, 甘肃 兰州 730000
甘肃省放射影像临床医学研究中心, 甘肃 兰州 730000
甘肃省智能影像医学工程研究中心, 甘肃 兰州 730000
甘肃省智能影像医学行业技术中心, 甘肃 兰州 730000
精准影像协同创新甘肃省国际科技合作基地, 甘肃 兰州 730000 | | | 郭顺林 | 兰州大学第一医院放射科, 甘肃 兰州 730000
甘肃省放射影像临床医学研究中心, 甘肃 兰州 730000
甘肃省智能影像医学工程研究中心, 甘肃 兰州 730000
甘肃省智能影像医学行业技术中心, 甘肃 兰州 730000
精准影像协同创新甘肃省国际科技合作基地, 甘肃 兰州 730000 | guoshunlin@msn.com |
|
| 摘要点击次数: 1272 |
| 全文下载次数: 633 |
| 中文摘要: |
| 目的 建立预测肝脏二维自旋回波-平面回波MR弹性成像(2D-SE-EPI-MRE)图像质量模型,评估其临床价 值。方法 回顾性分析116例肝脏病变患者的上腹部2D-SE-EPI-MRE资料,以MRI显示可测量肝脏容积百分比(pMLV)中位数为界,将其分为质量优良(优良组)或不良(不良组),并按7 ∶ 3比例划分训练集(n=82)与测试集(n=34)。以MRI肝脏左叶平均硬度(MSLL)、肝实质T2WI信号(SIT2-liver)、特异性增强肝胆期相对强化比值(RE)、扫描时驱动器振幅(DA)及患者屏气状态(BHS)为自变量,以可测量肝脏容积百分比(pMLV)为因变量,基于训练集建立预测MR弹性成像图像质量的随机森林和线性回归模型;以均方误差(MSE)、均方根误差(RMSE)、平均绝对误差(MAE)及决定系数(R2)评估2种模型对于测试集的预测效能。结果 优良组59例、不良组57例,组间MSLL、DA、BHS差异均有统计学意义(P均＜0.05),而RE、SIT2-liver差异均无统计学意义(P均＞0.05)。测试集中,随机森林模型的R2值均高于线性回归模型、而MSE、RMSE及MAE均低于线性回模型。结论 所获随机森林模型可用于预测肝脏2D-SE-EPI-MRE图像质量。 |
| 英文摘要: |
| Objective To establish a prediction model for imaging quality of liver two-dimensional spin-echo and planar-echo MR elastography (2D-SE-EPI-MRE), and to observe its clinical value. Methods Abdomen 2D-SE-EPI-MRE examination of 116 patients with liver diseases were retrospectively collected and divided into good or poor quality groups according to the median percent measurable liver volume (pMLV) shown on MRI. The patients were divided into training set (n=82) and test set (n=34) at the ratio of 7 ∶ 3. Taken the mean stiffness of left liver lobe (MSLL), signal of hepatic parenchyma on T2WI (SIT2-liver), relative enhancement ratio (RE) of hepatobiliary phase from specific enhanced obtained from MRI, driver amplitude (DA) and patient's breath holding state (BHS) as independent variables and pMLV as dependent variable, random forest and linear regression models for predicting imaging quality of MR elastography were established based on training set. The mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE) and coefficient of determination (R2) were used to evaluate the prediction efficacy of the models in test set. Results There were 59 cases in good quality group and 57 cases in poor quality group. Significant differences of MSLL, DA and BHS were found (all P＜0.05), but not of RE nor SIT2-liver between groups (both P＞0.05). In test set, R2 value of random forest model was higher, while MSE, RMSE and MAE were lower than those of linear regression model. Conclusion The obtained random forest model could be used to predict imaging quality of liver 2D-SE-EPI-MRE. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
