基于细节增强的MR序列转换方法

doi:10.11959/j.issn.2096-6652.202110

智能科学与技术学报 ›› 2021, Vol. 3 ›› Issue (1): 93-100.doi: 10.11959/j.issn.2096-6652.202110

基于细节增强的MR序列转换方法

严凡¹, 邸奕宁², 张建伟¹, 陈为¹

¹ 浙江大学计算机辅助设计与图形学国家重点实验室，浙江杭州 310058
² 浙江大学伊利诺伊大学厄巴纳香槟校区联合学院，浙江杭州 314400

修回日期:2021-02-08 出版日期:2021-03-15 发布日期:2021-03-01
作者简介:严凡（1995- ），男，浙江大学计算机辅助设计与图形学国家重点实验室硕士生，主要研究方向为医学影像、深度学习。
邸奕宁（1998- ），男，浙江大学伊利诺伊大学厄巴纳香槟校区联合学院本科生，主要研究方向为智能交通系统、交通控制与优化、计算机可视分析。
张建伟（1995- ），男，浙江大学计算机辅助设计与图形学国家重点实验室博士生，主要研究方向为医学影像、深度学习。
陈为（1976- ），男，博士，浙江大学计算机辅助设计与图形学国家重点实验室教授、博士生导师，中国计算机学会高级会员，主要研究方向为可视化、可视分析、大数据分析、人机混合智能。

Conversion method of magnetic resonance sequence based on detail enhancement

Fan YAN¹, Yining DI², Jianwei ZHANG¹, Wei CHEN¹

¹ State Key Lab of CAD＆CG , Zhejing University, Hangzhou 310058, China
² Zhejiang University/University of Illinois at Urbana-Champaign Institute, Hangzhou 314400, China

Revised:2021-02-08 Online:2021-03-15 Published:2021-03-01

摘要/Abstract

摘要：

磁共振（MR）是一种被广泛使用的医学成像方式。针对某些序列的 MR 影像不易直接获得、需要由其他序列的MR影像转换的问题，提出一种基于细节增强的MR序列转换方法。该方法基于条件生成对抗网络构建了残差模板合成模块 ResGAN和细节增强合成模块 EnGAN，设计了一种端到端的网络结构。在一组配准过的神经纤维瘤T1序列影像和STIR序列影像上进行了测试，结果表明，相比于现有方法，提出的方法更好地还原了边界细节和信号强度细节。

关键词: 磁共振序列转换, 生成对抗网络, 医学影像处理

Abstract:

Magnetic resonance (MR) is a widely used medical imaging method.However, MR images of certain modalities are not easy to obtain directly and thus need to be converted from other modality.To solve these problem, a conversion method of MR sequence based on detail enhancement was proposed.The proposed method constructed an end-to-end network with two conditional GAN-based modules: resiguals template generation module (ResGAN) and detail enhancement module (EnGAN).The proposed method are tested on a set of registered neurofibroma T1 and STIR images, which demonstrate that the proposed method can restore boundary details and signal intensity details better than other existing methods.

Key words: cross-modality magnetic resonance, generative adversarial network, medical image processing

中图分类号:

R445.2

严凡, 邸奕宁, 张建伟, 等. 基于细节增强的MR序列转换方法[J]. 智能科学与技术学报, 2021, 3(1): 93-100.

Fan YAN, Yining DI, Jianwei ZHANG, et al. Conversion method of magnetic resonance sequence based on detail enhancement[J]. Chinese Journal of Intelligent Science and Technology, 2021, 3(1): 93-100.

图/表 10

图1

图2

图3

图4

图5

表1

模型在神经纤维瘤上的准确率"

方法	SSIM	PSNR	MSE
Pix2Pix	0.785 5	17.209 1	0.004 9
VAE	$0 . 8124$	$18 . 1298$	$0 . 0041$
Ea-GAN	0.789	16.977 9	0.005 2
本文提出的方法	0.798 8	17.357 7	0.004 8

表1

表2

消融实验结果"

方法	SSIM	PSNR	MSE
without edge loss	0.791 3	16.952 6	0.005 3
without res loss	0.792 8	17.193 3	0.005
本文提出的方法	$0 . 7988$	$17 . 3577$	$0 . 0048$

表2

图6

图7

图8

参考文献 26

[1]	JACK C R , PETERSEN R C , XU Y C ,et al. Prediction of AD with MRI-based hippocampal volume in mild cognitive impairment[J]. Neurology, 1999,52(7): 1397-1397.
[2]	HANRAHAN C J , SHAH L M . MRI of spinal bone marrow:part 2,T1-weighted imaging-based differential diagnosis[J]. American Journal of Roentgenology, 2011,197(6): 1309-1321.
[3]	EUSTACE S , TELLO R , DECARVALHO V ,et al. Whole body turbo STIR MRI in unknown primary tumor detection[J]. Journal of Magnetic Resonance Imaging, 1998,8(3): 751-753.
[4]	DADAR M , PASCOAL T A , MANITSIRIKUL S ,et al. Validation of a regression technique for segmentation of white matter hyperintensities in Alzheimer’s disease[J]. IEEE Transactions on Medical Imaging, 2017,36(8): 1758-1768.
[5]	PURANIK R , MUTHURANGU V , CELERMAJER D S ,et al. Congenital heart disease and multi-modality imaging[J]. Heart Lung and Circulation, 2010,19(3): 133-144.
[6]	KINGMA D P , WELLING M . Auto-encoding variational bayes[J]. arXiv preprint, 2013,arXiv:1312.6114.
[7]	GOODFELLOW I , POUGET-ABADIE J , MIRZA M ,et al. Generative adversarial nets[C]// Proceeding of the Advances in Neural Information Processing Systems.[S.l.:s.n.], 2014: 2672-2680.
[8]	MIRZA M , OSINDERO S . Conditional generative adversarial nets[J]. arXiv preprint, 2014,arXiv:1411.1784.
[9]	YI X , WALIA E , BABYN P . Generative adversarial network in medical imaging:a review[J]. Medical Image Analysis, 2019,58: 101552.
[10]	MOUNSAVENG S , VAZQUEZ D , AYED I B ,et al. Adversarial learning of general transformations for data augmentation[J]. arXiv preprint, 2019,arXiv:1909.09801.
[11]	LEDIG C , THEIS L , HUSZáR F ,et al. Photo-realistic single image super-resolution using a generative adversarial network[C]// Proceeding of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2017: 4681-4690.
[12]	BROCK A , DONAHUE J , SIMONYAN K . Large scale GAN training for high fidelity natural image synthesis[J]. arXiv preprint, 2018,arXiv:1809.11096.
[13]	LITJENS G , KOOI T , BEJNORDI B E ,et al. A survey on deep learning in medical image analysis[J]. Medical Image Analysis, 2017,42: 60-88
[14]	NIE D , TRULLO R , LIAN J ,et al. Medical image synthesis with context-aware generative adversarial networks[C]// Proceeding of the 20th International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham:Springer, 2017: 417-425.
[15]	CHARTSIAS A , JOYCE T , DHARMAKUMAR R ,et al. Adversarial image synthesis for unpaired multi-modal cardiac data[C]// Proceeding of the 2nd International Workshop on Simulation and Synthesis in Medical Imaging. Cham:Springer, 2017: 3-13.
[16]	HIASA Y , OTAKE Y , TAKAO M ,et al. Cross-modality image synthesis from unpaired data using CycleGAN[C]// Proceeding of the 3rd International Workshop on Simulation and Synthesis in Medical Imaging. Cham:Springer, 2018: 31-41.
[17]	WOLTERINK J M , DINKLA A M , SAVENIJE M H F ,et al. Deep MR to CT synthesis using unpaired data[C]// Proceeding of the 3rd International Workshop on Simulation and Synthesis in Medical Imaging. Cham:Springer, 2017: 14-23.
[18]	WELANDER P , KARLSSON S , EKLUND A . Generative adversarial networks for image-to-image translation on multi-contrast MR images-a comparison of CycleGAN and UNIT[J]. arXiv preprint, 2018,arXiv:1806.07777.
[19]	OLUT S , SAHIN Y H , DEMIR U ,et al. Generative adversarial training for MRA image synthesis using multi-contrast MRI[C]// Proceeding of the 1st International Workshop on PRedictive Intelligence in Medicine. Cham:Springer, 2018: 147-154.
[20]	HAN C , HAYASHI H , RUNDO L ,et al. GAN-based synthetic brain MR image generation[C]// Proceeding of the 2018 IEEE 15th International Symposium on Biomedical Imaging. Piscataway:IEEE Press, 2018: 734-738.
[21]	DAR S U H , YURT M , KARACAN L ,et al. Image synthesis in multi-contrast MRI with conditional generative adversarial networks[J]. IEEE Transactions on Medical Imaging, 2019,38(10): 2375-2388.
[22]	YU B T , WANG L P , SHI L ,et al. Ea-GANs:edge-aware generative adversarial networks for cross-modality MR image synthesis[J]. IEEE Transactions on Medical Imaging, 2019,38(7): 1750-1762.
[23]	RONNEBERGER O , FISCHER P , BROX T . U-Net:Convolutional networks for biomedical image segmentation[C]// Proceeding of the 18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham:Springer, 2015: 234-241.
[24]	ULYANOV D , VEDALDI A , LEMPITSKY V . Instance normalization:The missing ingredient for fast stylization[J]. arXiv preprint, 2016,arXiv:1607.08022.
[25]	MAAS A L , HANNUN A Y , NG A Y . Rectifier nonlinearities improve neural network acoustic models[C]// Proceeding of the 30th International Conference on Machine Learning.[S.l.:s.n.], 2013.
[26]	WANG Z , BOVIK A C , SHEIKH H R ,et al. Image quality assessment:from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004,13(4): 600-612.

基于细节增强的MR序列转换方法

Conversion method of magnetic resonance sequence based on detail enhancement

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 26

相关文章 2

Metrics

推荐阅读 0

[1]	李莉, 曾伟良, 黄永慧, 孙为军. 基于半监督学习的人脸识别反欺骗方法研究[J]. 智能科学与技术学报, 2021, 3(3): 370-380.
[2]	余翀. 基于半监督生成对抗网络的三维重建云工作室[J]. 智能科学与技术学报, 2019, 1(1): 70-82.