Department of Radiology, The Institute of Medical Science, The University of Tokyo, 4-6-1 Shirokanedai, Minato-ku, Tokyo, 108-8639, Japan.
Department of Radiology, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-8655, Japan.
Eur Radiol. 2018 Nov;28(11):4578-4585. doi: 10.1007/s00330-018-5499-7. Epub 2018 May 14.
To investigate whether liver fibrosis can be staged by deep learning techniques based on CT images.
This clinical retrospective study, approved by our institutional review board, included 496 CT examinations of 286 patients who underwent dynamic contrast-enhanced CT for evaluations of the liver and for whom histopathological information regarding liver fibrosis stage was available. The 396 portal phase images with age and sex data of patients (F0/F1/F2/F3/F4 = 113/36/56/66/125) were used for training a deep convolutional neural network (DCNN); the data for the other 100 (F0/F1/F2/F3/F4 = 29/9/14/16/32) were utilised for testing the trained network, with the histopathological fibrosis stage used as reference. To improve robustness, additional images for training data were generated by rotating or parallel shifting the images, or adding Gaussian noise. Supervised training was used to minimise the difference between the liver fibrosis stage and the fibrosis score obtained from deep learning based on CT images (F score) output by the model. Testing data were input into the trained DCNNs to evaluate their performance.
The F scores showed a significant correlation with liver fibrosis stage (Spearman's correlation coefficient = 0.48, p < 0.001). The areas under the receiver operating characteristic curves (with 95% confidence intervals) for diagnosing significant fibrosis (≥ F2), advanced fibrosis (≥ F3) and cirrhosis (F4) by using F scores were 0.74 (0.64-0.85), 0.76 (0.66-0.85) and 0.73 (0.62-0.84), respectively.
Liver fibrosis can be staged by using a deep learning model based on CT images, with moderate performance.
• Liver fibrosis can be staged by a deep learning model based on magnified CT images including the liver surface, with moderate performance. • Scores from a trained deep learning model showed moderate correlation with histopathological liver fibrosis staging. • Further improvement are necessary before utilisation in clinical settings.
利用 CT 图像深度学习技术对肝纤维化进行分期。
本临床回顾性研究经机构审查委员会批准，纳入 286 例因肝脏评估而行动态对比增强 CT 检查且具有肝纤维化分期组织病理学信息的患者 496 次 CT 检查。利用患者的 396 次门静脉期图像（年龄和性别数据，F0/F1/F2/F3/F4=113/36/56/66/125）训练深度卷积神经网络（DCNN）；另外 100 次图像（F0/F1/F2/F3/F4=29/9/14/16/32）用于测试训练网络，以组织病理学纤维化分期作为参考。为了提高稳健性，通过旋转或平行移动图像或添加高斯噪声来生成训练数据的附加图像。采用监督训练来最小化模型基于 CT 图像输出的肝纤维化分期和纤维化评分（F 评分）与组织病理学纤维化分期之间的差异。将测试数据输入训练后的 DCNN 以评估其性能。
F 评分与肝纤维化分期显著相关（Spearman 相关系数=0.48，p<0.001）。使用 F 评分诊断显著纤维化（≥F2）、进展性纤维化（≥F3）和肝硬化（F4）的受试者工作特征曲线下面积（95%置信区间）分别为 0.74（0.64-0.85）、0.76（0.66-0.85）和 0.73（0.62-0.84）。
利用 CT 图像深度学习模型可以对肝纤维化进行分期，其性能中等。
利用包括肝脏表面在内的放大 CT 图像，通过深度学习模型可以对肝纤维化进行分期，其性能中等。
经过训练的深度学习模型评分与组织病理学肝纤维化分期有中等相关性。
在临床应用前，还需要进一步改进。
