• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Workshop on Predictive Intelligence in Medicine;International Conference on Medical Image Computing and Computer Assisted Intervention >Longitudinal Prediction of Radiation-Induced Anatomical Changes of Parotid Glands During Radiotherapy Using Deep Learning
【24h】

Longitudinal Prediction of Radiation-Induced Anatomical Changes of Parotid Glands During Radiotherapy Using Deep Learning

机译：深深学习在放疗过程中辐射诱导腮腺诱导的解剖学变化的纵向预测

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

During a course of radiotherapy, patients may have weight loss and radiation induced anatomical changes. To avoid delivering harmful dose to normal organs, the treatment may need adaptation according to the change. In this study, we proposed a novel deep neural network for predicting parotid glands (PG) anatomical changes by using the displacement fields (DFs) between the planning CT and weekly cone beam computed tomography (CBCT) acquired during the treatment. Sixty three HN patients treated with volumetric modulated arc therapy of 70 Gy in 35 fractions were retrospectively studied. We calculated DFs between week 1-3 CBCT and the planning CT by a B-spline deformable image registration algorithm. The resultant DFs were subsequently used as input to a novel network combining convolutional neural networks and recurrent neural networks for predicting the DF between the Week 4-6 CBCT and the planning CT. Finally, we reconstructed the warped PG contour using the predicted DF. For evaluation, we calculated DICE coefficient and mean volume difference by comparing the predicted PG contours, and manual contours at weekly CBCT. The average DICE was 0.82 (week 4), 0.81 (week 5), and 0.80 (week 6) and the average of volume difference between predict contours and manual contours was 1.85 cc (week 4), 2.20 cc (week 5) and 2.51 cc (week 6). In conclusion, the proposed deep neural network combining CNN and RNN was capable of predicting anatomical and volumetric changes of the PG with clinically acceptable accuracy.
机译：在放射治疗过程中，患者可能具有减肥和辐射诱导的解剖改变。为避免将有害剂量递送到正常器官，治疗可能需要根据变化进行适应。在这项研究中，我们提出了一种新的神经网络，用于通过在治疗期间获取的规划CT和每周锥梁计算断层摄影（CBCT）之间的位移场（DF）来预测腮腺（PG）解剖学改变。回顾性研究了六十三个患有70 Gy的体积调节弧治疗的患者，进行了回顾性研究。我们通过B样条曲线可变形图像配准算法计算第1-3 CBCT和规划CT之间的DFS。随后将得到的DFS用作新颖网络的输入组合卷积神经网络和经常性神经网络，用于预测星期间4-6 CBCT和规划CT之间的DF。最后，我们使用预测的DF重建翘曲的PG轮廓。为了评估，我们通过将预测的PG轮廓进行比较，计算骰子系数和平均体积差，以及每周CBCT的手动轮廓。平均骰子为0.82（第4周），0.81（第5周）和0.80（第6周），预测轮廓和手动轮廓之间的体积差异为1.85cc（第4周），2.20cc（第5周）和2.51 CC（第6周）。总之，所提出的CNN和RNN的深神经网络能够以临床上可接受的精度预测PG的解剖和体积变化。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号