|本期目录/Table of Contents|

[1]菅影超,付东山,王伟.基于3D深度卷积神经网络依据MRI生成伪CT的研究[J].天津医科大学学报,2020,26(02):133-137.
　JIAN Ying-chao,FU Dong-shan,WANG Wei.Study on generating pseudo-CT image based on MRI using 3D deep convolutional neural network[J].Journal of Tianjin Medical University,2020,26(02):133-137.
点击复制

基于3D深度卷积神经网络依据MRI生成伪CT的研究(PDF)
分享到：

《天津医科大学学报》[ISSN:1006-8147/CN:12-1259/R]

卷:
26卷
期数:
2020年02期
页码:
133-137
栏目:
临床医学
出版日期:
2020-04-30

文章信息/Info

Title:
Study on generating pseudo-CT image based on MRI using 3D deep convolutional neural network
文章编号:
1006-8147(2020)02-0138-04
作者:
菅影超; 付东山; 王伟
（天津医科大学肿瘤医院放疗科，国家肿瘤临床医学研究中心，天津市“肿瘤防治”重点实验室，天津市恶性肿瘤临床医学研究中心，天津 300060）
Author(s):
JIAN Ying-chao;  FU Dong-shan;  WANG Wei
（Department of Radiation Oncology, Tianjin Medical University Cancer Institute and Hospital,National Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, Tianjin’s Clinical Research Center for Cancer, Tianjin 300060, China）
关键词:
MRI; 伪CT; 深度卷积神经网络; U-net; 平均绝对误差
Keywords:
MRI;  pseudo-CT;  deep convolutional neural network;  U-net;  mean absolute error
分类号:
R815.2
DOI:
-
文献标志码:
A
摘要:
目的：研究一种依据MRI生成伪CT的方法，从而减少放疗过程中额外CT的使用，降低患者辐射剂量，提高放疗精准度。方法：提出一种基于3D深度卷积神经网络（DCNN）的预测算法，利用单张图像的解剖特征以及相邻图像层之间的关联信息，从而提高了图像特征提取的准确性。采用U-net网络结构，通过编码部分的卷积层、池化层和解码部分的上采样、卷积层，对MRI和对应的CT进行端到端转换的学习。采集13例患者图像数据，应用留一交叉验证的方法，分别对3D DCNN和2D DCNN的伪CT结果与原始CT进行对照比较。结果：提出的3D DCNN算法的平均绝对误差（MAE）为86 HU，远小于2D DCNN的136 HU。结论：3D DCNN算法能更准确的生成伪CT，明显改善了骨骼、空气与软组织之间的误转化。
Abstract:
Objective: To study a method of generating pseudo-CT based on MRI, for reducing the use of extra CT in the course of radiotherapy, reducing the radiation dose of patients and improving the accuracy of radiotherapy. Methods: A prediction algorithm based on 3D deep convolutional neural network(DCNN) was proposed. The accuracy of image feature extraction was improved by utilizing the anatomical features of a single image and the correlation information between adjacent image slices. Using U-net network structure, the end-to-end conversion from MRI to the corresponding CT was studied by convolution layers and pooling layers in the encoding portion, and upsampling layers and convolution layers in the decoding portion. The image data of 13 patients were collected and tested. The results of generated pseudo-CT for 3D DCNN and 2D DCNN were compared with original CT using the method of leave-one-out cross validation, respectively. Results: A mean absolute error（MAE） of the proposed 3D DCNN algorithm was 86 HU, which was much smaller than the 136 HU of the 2D DCNN. Conclusion: The 3D DCNN algorithm can more accurately generate pseudo-CT and significantly improve the mis-conversion among bone, air and soft tissue.

参考文献/References:


[1] Hsu S H, Cao Y, Huang K, et al. Investigation of a method for generating synthetic CT models from MRI scans of the head and neck for radiation therapy[J]. Phys Med Biol, 2013, 58(23):8420
[2] Hofmann M, Pichler B, Bernhard Sch?觟lkopf, et al. Towards quantitative PET/MRI: A review of MR-based attenuation correction techniques[J]. Eur J Nucl Med Mol Imaging, 2009, 36 (Suppl 1):S94
[3] Chen L, Nguyen T B, Jones E, et al. Magnetic resonance-based treatment planning for prostate intensity-modulated radiotherapy: creation of digitally reconstructed radiographs[J]. Int J Radiat Oncol Biol Phys, 2007, 68(3):904
[4] Edmund J M, Kjer H M, Leemput K V, et al. A voxel-based investigation for MRI-only radiotherapy of the brain using ultra short echo times[J]. Phys Med Biol, 2014, 59(23):7504
[5] Schreibmann E, Nye J A, Schuster D M, et al. MR-based attenuation correction for hybrid PET-MR brain imaging systems using deformable image registration[J]. Med Phys, 2010, 37(5):2104
[6] Dowling J A, Lambert J, Parker J, et al. An atlas-based electron density mapping method for magnetic resonance imaging (MRI)-alone treatment planning and adaptive MRI-based prostate radiation therapy[J]. Int J Radiat Oncol Biol Phys, 2012, 83(1):e6
[7] Uh J, Merchant T E, Li Y, et al. MRI-based treatment planning with pseudo CT generated through atlas registration[J]. Med Phys, 2014, 41(5):051711
[8] Demol B, Boydev C, Korhonen J, et al. Dosimetric characterization of MRI-only treatment planning for brain tumors in atlas-based pseudo-CT images generated from standard T1-weighted MR images[J]. Med Phys, 2016, 43(12):6566
[9] Gudur MSR, Hara W, Le Q T, et al. A unifying probabilistic Bayesian approach to derive electron density from MRI for radiation therapy treatment planning[J]. Phys Med Biol, 2014, 59(21):6597
[10] Han X. MR-based synthetic CT generation using a deep convolutional neural network method[J]. Med Phys, 2017, 44(4):1409
[11] Sled J G, Zijdenbos A P, Evans A C. A nonparametric method for automatic correction of intensity nonuniformity in MRI data[J]. IEEE T Med Imaging, 1998, 17(1):89
[12] Otsu N. A threshold selection method from gray-level histograms [J]. IEEE Trans Sys Man Cybern, 2007, 9(1):62
[13] 周飞燕, 金林鹏, 董军.卷积神经网络研究综述[J].计算机学报,2017, 40(7):5
[14] Kingma D P, Ba J. Adam: A method for stochastic optimization[C]//LeCun Y. Proceedings of the 3rd international conference on learning representations. San Diego: OALib, 2015:2
[15] Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//Blei D. Proceedings of the 32nd international conference on international conference on machine learning. Lille: JMLR, 2015:449
[16] 胡永生,张立毅.基于非局部自相似图像块字典学习的伪CT图像预测[J].信号处理, 2017,33(3):347

相似文献/References:

[1]孙静,蔡跃增.肋骨少见单发膨胀性病变的影像学诊断[J].天津医科大学学报,2013,19(05):421.
[2]侯 萍,张 虹.泪腺腺样囊性癌93例临床分析[J].天津医科大学学报,2016,22(06):528.
[3]徐学鑫,赵 滨,胡晓丽.儿童颌面部Kimura病1例[J].天津医科大学学报,2019,25(01):84.
[4]刘广宇,孙浩然,白人驹.多序列MRI检查在胆系结石诊断中的应用价值[J].天津医科大学学报,2021,27(03):301.
　LIU Guang-yu,SUN Hao-ran,BAI Ren-ju.The clinical value of multiple sequence MRI in detecting gallstones[J].Journal of Tianjin Medical University,2021,27(02):301.
[5]温鹏,韩玉娟,杨剑,等.肝细胞癌患者缩短普美显强化MRI检查中肝胆期时间的探索[J].天津医科大学学报,2022,28(04):433.
　WEN Peng,HAN Yu-juan,YANG Jian,et al.Exploration of shortening the duration of hepatobiliary phase in EOB enhanced MRI for patients with hepatocellular carcinoma[J].Journal of Tianjin Medical University,2022,28(02):433.

备注/Memo

备注/Memo:
基金项目 国家重点研发计划（2017YFC0113100）
作者简介 菅影超（1994-），女，硕士在读，研究方向：医学图像处理；
通信作者:付东山，E-mail：dongshan_fu@hotmail.com；王伟，E-mail：weiwang_2@126.com。
更新日期/Last Update: 2020-06-02