| 马立勇,董梦超,李广涵,刘健,武敬平,陆海涛,邹古明,卓莉,牟姗,郑敏.基于剪切波弹性成像和卷积神经网络建立深度学习模型预测肾脏病变[J].中国医学影像技术,2021,37(6):919~922 |
| 基于剪切波弹性成像和卷积神经网络建立深度学习模型预测肾脏病变 |
| Predicting renal diseases with deep learning model based on shear wave elastography and convolutional neural network |
| 投稿时间：2021-01-31 修订日期：2021-05-09 |
| DOI：10.13929/j.issn.1003-3289.2021.06.029 |
| 中文关键词: 肾疾病 弹性成像技术 深度学习 |
| 英文关键词:kidney diseases elasticity imaging techniques deep learning |
| 基金项目:国家重点研发计划政府间国际科技创新合作重点专项（2017YFE0110500）、山东省自然科学基金（ZR2018MF026）、山东省重点研发计划项目（2019GGX101054）。 |
| 作者 | 单位 | E-mail | | 马立勇 | 哈尔滨工业大学(威海)信息科学与工程学院, 山东 威海 264209 | | | 董梦超 | 哈尔滨工业大学(威海)信息科学与工程学院, 山东 威海 264209 | | | 李广涵 | 中日友好医院超声医学科, 北京 100029 | | | 刘健 | 中日友好医院超声医学科, 北京 100029 | | | 武敬平 | 中日友好医院超声医学科, 北京 100029 | | | 陆海涛 | 中日友好医院肾病科, 北京 100029 | | | 邹古明 | 中日友好医院肾病科, 北京 100029 | | | 卓莉 | 中日友好医院肾病科, 北京 100029 | | | 牟姗 | 上海交通大学医学院附属仁济医院肾脏科, 上海 200127 | | | 郑敏 | 中日友好医院超声医学科, 北京 100029 | zhengmin16@163.com |
|
| 摘要点击次数: 3039 |
| 全文下载次数: 769 |
| 中文摘要: |
| 目的 基于剪切波弹性成像（SWE）量化参数和卷积神经网络建立深度学习（DL）模型预测肾脏病变。方法 采集94例肾脏病变患者（病例组）和109名健康人（对照组）的肾脏超声SWE量化参数。利用卷积神经网络建立DL模型，比较DL模型和支持向量机、随机森林模型预测肾脏病变的敏感度、特异度、准确率和曲线下面积（AUC）。结果 DL模型对预测肾脏病变的敏感度为90.48%，特异度为100%，准确率为95.12%，AUC为0.93；支持向量机模型的敏感度、特异度、准确率和AUC分别为80.74%、80.71%、80.98%、0.90，随机森林模型分别为82.22%、77.87%、80.33%和0.88。DL模型预测敏感度、特异度、准确率和AUC均高于支持向量机和随机森林模型，与支持向量机模型和随机森林模型预测肾脏病变差异均有统计学意义（P均<0.05）。结论 基于SWE量化参数和卷积神经网络的DL模型预测肾脏疾病性能良好，具有一定临床价值。 |
| 英文摘要: |
| Objective To observe the value of a deep learning (DL) model based on shear wave elastography (SWE) quantitative parameters and convolution neural network for predicting renal diseases. Methods The quantitative parameters of kidney ultrasound shear wave elastography were collected from 94 cases of kidney diseases (case group) and 109 healthy subjects (control group). DL model was built by convolutional neural network. The accuracy, sensitivity, specificity and area under the curve (AUC) for predicting renal diseases were compared among DL model, support vector machine and random forest models. Results The sensitivity, specificity, accuracy and AUC of DL model for predicting renal diseases was 90.48%, 100%, 95.12% and 0.93, of support vector machines was 80.74%, 80.71%, 80.98% and 0.90, while of random forest models was 82.22%, 77.87%, 80.33% and 0.88, respectively. The sensitivity, specificity, accuracy and AUC of DL model were all better than those of support vector machines and random forest models. There were significant differences between DL model and support vector machine model, also between DL model and random forest model for predicting renal diseases (both P<0.05). Conclusion DL model based on SWE quantitative parameters and convolution neural network had good performances in predicting renal diseases, therefore having certain clinical value. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
