基于生成对抗网络的医学数据域适应研究

doi:10.11959/j.issn.2096-0271.2020043

摘要/Abstract

摘要：

在医疗影像辅助诊断研究中，研究者通常使用不同医院（多域）的数据，但当其中一个域的训练样本较少时，模型在该域的测试集上的分类结果将会很差。针对此问题，提出一种基于生成对抗网络的分类方法进行男女脑影像差异的域适应研究，首先使用生成对抗网络学习不同域的数据分布，并提取关键特征，然后基于提取的关键特征研究不同域的男女脑影像差异。实验表明，该方法在仅有少量数据参与训练的域上也能取得80%以上的分类准确度。

关键词: 深度学习, 生成对抗网络, 域适应, 医疗影像

Abstract:

In the study of medical imaging aided diagnosis,researchers often collect a lot of training data coming from different hospitals (named variety fields).But because of the certain field has insufficient training data,the deep learning model would get very poor performance on the test data of this field.To mitigate this problem,a method to study domain adaptation of the difference between male and female brain images based on the generative adversarial network was proposed.The data distribution of different domains was learned and the key features were extracted by using the generative adversarial network,and then the differences between male and female brain images in different domains were studied based on the extracted key features.Experiments show that the method can also achieve more than 80% recognition accuracy in the domain with only a small amount of data involved in training.

Key words: deep learning, generative adversarial network, domain adaptation, medical image

中图分类号:

TP399

于胡飞, 温景熙, 辛江, 唐艳. 基于生成对抗网络的医学数据域适应研究[J]. 大数据, 2020, 6(5): 45-54.

Hufei YU, Jingxi WEN, Jiang XIN, Yan TANG. Study on domain adaptation of medical data based on generative adversarial network[J]. Big Data Research, 2020, 6(5): 45-54.

图/表 7

图1

图2

图3

图4

图5

图6

图7

参考文献 23

[1]	QIU T , WEN C , XIE K ,et al. Efficient medical image enhancement based on CNN-FBB model[J]. IET Image Processing, 2019,13(10): 1736-1744.
[2]	YOU K C , LONG M S , CAO Z J ,et al. Universal domain adaptation[C]// The IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2019: 2720-2729.
[3]	GANIN Y , LEMPITSKY V . Unsupervised domain adaptation by backpropagation[C]// International Conference on Machine Learning.[S.l.:s.n. ], 2015: 1180-1189.
[4]	LONG M S , WANG J M . Learning transferable features with deep adaptation networks[C]// International Conference on Machine Learning.[S.l.:s.n. ], 2015: 97-105.
[5]	GOODFELLOW I , POUGET-ABADIE J , MIRZA M ，et al. Generative adversarial nets[M]. [S.l.:s.n.]. 2014.
[6]	KAZEMINIA S , BAUR C , KUIJPER A ,et al. GANs for medical image analysis[J]. arXiv preprint, 2018,arXiv:1809.06222.
[7]	HE K M , ZHANG X Y , REN S Q ,et al. Deep residual learning for image recognition[C]// IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 770-778.
[8]	RONNEBERGER O , FISCHER P , BROX T . U-net:convolutional networks for biomedical image segmentation[C]// International Conference on Medical Image Computing and Computer-Assisted Intervention.[S.l.:s.n]. 2015: 234-241.
[9]	ZHANG L , GOOYA A , FRANGI A F . Semi-supervised assessment of incomplete LV coverage in cardiac MRI using generative adversarial nets[C]// International Workshop on Simulation and Synthesis in Medical Imaging.[S.l.:s.n. ], 2017: 61-68.
[10]	GLOROT X , BORDES A , BENGIO Y . Deep sparse rectifier neural networks[C]// The 14th International Conference on Artificial Intelligence and Statistics.[S.l.:s.n]. 2011: 315-323.
[11]	IOFFE S , SZEGEDY C . Batch normalization:accelerating deep network training by reducing internal covariate shift[J]. arXiv preprint, 2015,arXiv:1502.03167.
[12]	HE K M , ZHANG X Y , REN S Q ,et al. Delving deep into rectifiers:Surpassing human-level performance on imagenet classification[C]// IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2015: 1026-1034.
[13]	KINGMA D P , BA J . Adam:a method for stochastic optimization[J]. arXiv preprint, 2014,arXiv:1412.6980.
[14]	CETIN M S , CHRISTENSEN F , ABBOTT C C ,et al. Thalamus and posterior temporal lobe show greater internetwork connectivity at rest and across sensory paradigms in schizophrenia[J]. Neuroimage, 2014,97: 117-126.
[15]	SHAFTO M A , TYLER L K , DIXON M ,et al. The cambridge centre for ageing and neuroscience (Cam-CAN) study protocol:a cross-sectional,lifespan,multidisciplinary examination of healthy cognitive ageing[J]. BMC Neurology, 2014,14(1):204.
[16]	LIU W , WEI D T , CHEN Q L ,et al. Longitudinal test-retest neuroimaging data from healthy young adults in southwest China[J]. Scientific Data, 2017,4: 1-9.
[17]	ESSEN D C V , SMITH S M , BARCH D M ,et al. The WU-Minn human connectome project:an overview[J]. Neuroimage, 2013,80(80): 62-79.
[18]	MAREK K , JENNINGS D , LASCH S ,et al. The parkinson progression marker initiative (PPMI)[J]. Progress in Neurobiology, 2011,95: 629-635.
[19]	SMITH S M , JENKINSON M , WOOLRICH M W ,et al. Advances in functional and structural MR image analysis and implementation as FSL[J]. Neuroimage, 2004,23: 208-219.
[20]	SAHINER B , PEZESHK A , HADJIISKI L M ,et al. Deep learning in medical imaging and radiation therapy[J]. Medical Physics, 2019,46(1): 1-36.
[21]	LI Q , BAI I Y , LIU T F . Deep learning classification of diabetic retinal images[J]. Journal of Image and Graphics, 2018,23: 1594-1603.
[22]	PANG H , WANG Z . Deep learning model for detection of diabetic retinopathy[J]. Journal of Software, 2017,28: 3018-3029.
[23]	HAN G H , LIU X B , ZHENG G Y . Detection method of lesion area in lung CT image[J]. Acta Automatica Sinica, 2017,43: 2071-2090.