基于元学习和神经架构搜索的半监督医学图像分割方法

doi:10.16180/j.cnki.issn1007-7820.2024.01.003

Abstract

Abstract:

Most medical image segmentation methods mainly focus on training and evaluating in the same or similar medical data domain, which need lots of pixel-level annotations. However, these models face challenges in out-of-distribution medical data set, which is known as "domain shift" problem. A fixed U-shaped segmentation structure is usually used to solve this problem, resulting in it not being better adapted to specific partition tasks.A gradient-based meta-learning and neural architecture search method is proposed in this study, which can adjust the segmentation network according to specific tasks to achieve good performance and have good generalization ability. This method mainly uses the specific task to carry out the architecture search module to further improve the segmentation effect, and then uses the gradient-based meta-learning training algorithm to improve the generalization ability.On the public dataset M&Ms, under the 5% label data, its Dice and Hausdorff distance are 79.62% and 15.38%. Under 2% label data, its Dice and Hausdorff distance are 74.03% and 17.05%.Compared with other mainstream methods, the proposed method has better generalization ability.

Key words: medical image segmentation, meta-learning, neural architecture search, domain generalization, disentangle representations, semi-supervised learning, convolutional neural network, deep learning

CLC Number:

TP391

YU Zhihong,LI Feifei. Semi-Supervised Medical Image Segmentation Method Based on Meta-Learning and Neural Architecture Search[J].Electronic Science and Technology, 2024, 37(1): 17-23.

Figures/Tables 10

Figure 1.

Figure 2.

Table 1.

Figure 3.

Table 2.

Table 3.

Table 4.

Table 5.

Figure 4.

Table 6.

References 26

[1]	缪冉, 李菲菲, 陈虬. 基于卷积神经网络与多尺度空间编码的场景识别方法[J]. 电子科技, 2020, 33(12):54-58.
	Miao Ran, Li Feifei, Chen Qiu. Scene recognition algorithm based on convolutional neural networks and multi-scale space encoding[J]. Electronic Science and Technology, 2020, 33(12):54-58.
[2]	左斌, 李菲菲. 基于注意力机制和Inf-Net的新冠肺炎图像分割方法[J]. 电子科技, 2023, 36(2):22-28.
	Zuo Bin, Li Feifei. COVID-19 image segmentation method based on attention mechanism and INF-NET[J]. Electronic Science and Tecnology, 2023, 36(2):22-28.
[3]	Zhang L, Wang X, Yang D, et al. Generalizing deep learning for medical image segmentation to unseen domains via deep stacked transformation[J]. IEEE Transactions on Medical Imaging, 2020, 39(7):2531-2540. doi: 10.1109/TMI.2020.2973595 pmid: 32070947
[4]	Tao Q, Yan W, Wang Y, et al. Deep learning-based method for fully automatic quantification of left ventricle function from cine MR images:A multivendor,multicenter study[J]. Radiology, 2019, 290(1):81-88. doi: 10.1148/radiol.2018180513
[5]	Bian C, Yuan C, Wang J, et al. Uncertainty-aware domain alignment for anatomical structure segmentation[J]. Medical Image Analysis, 2020, 64(8):101732-101746. doi: 10.1016/j.media.2020.101732
[6]	Li D, Yang Y, Song Y Z, et al. Learning to generalize:Meta-learning for domain generalization[C]. New Orleans: Proceedings of the AAAI Conference on Artificial Intelligence, 2018:3533-3541.
[7]	Zhang L, Wang X, Yang D, et al. Generalizing deep learning for medical image segmentation to unseen domains via deep stacked transformation[J]. IEEE Transactions on Medical Imaging, 2020, 39(7):2531-2540. doi: 10.1109/TMI.2020.2973595 pmid: 32070947
[8]	Liu C, Wang L, Li K, et al. Domain generalization via feature variation decorrelation[C]. New York: Proceedings of the Twenty-ninth ACM International Conference on Multimedia, 2021:1517-1526.
[9]	Li H, Pan S J, Wang S, et al. Domain generalization with adversarial feature learning[C]. Salt Lake City: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018:620-631.
[10]	Dou Q, Daniel C, Kamnitsas K, et al. Domain generalization via model-agnostic learning of semantic features[C]. Curran: Advances in Neural Information Processing Systems, 2019:2331-2339.
[11]	Liu Q, Dou Q, Heng P A. Shape-aware meta-learning for generalizing prostate MRI segmentation to unseen domains[C]. Lima: Medical Image Computing and Computer Assisted Intervention, 2020:852-863.
[12]	Khandelwal P, Yushkevich P. Domain generalizer:A few-shot meta learning framework for domain generalization in medical imaging[J]. Domain Adaptation and Representation Transfer,and Distributed and Collaborative Learning, 2020, 11(9):73-84.
[13]	Liu X, Thermos S, O'Neil A, et al. Semi-supervised meta-learning with disentanglement for domain-generalised medical image segmentation[C]. Strasbourg: Medical Image Computing and Computer Assisted Intervention, 2021:3592-3603.
[14]	Elsken T, Metzen J H, Hutter F. Neural architecture search: A survey[J]. Journal of Machine Learning Research, 2019, 20(1):1997-2017.
[15]	Hospedales T, Antoniou A, Micaelli P, et al. Meta-learning in neural networks:A survey[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(9): 5149-5169.
[16]	Li D, Zhang J, Yang Y, et al. Episodic training for domain generalization[C]. Long Beach: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019:78-86.
[17]	Weng Y, Zhou T, Li Y, et al. NAS-Unet:Neural architecture search for medical image segmentation[J]. IEEE Access, 2019, 7(3):44247-44257. doi: 10.1109/Access.6287639
[18]	Zhou Z, Siddiquee M M R, Tajbakhsh N, et al. UNet++: Redesigning skip connections to exploit multiscale features in image segmentation[J]. IEEE Transactions on Medical Imaging, 2020, 39(6):1856-1867. doi: 10.1109/TMI.2019.2959609 pmid: 31841402
[19]	Liu H, Simonyan K, Yang Y. DARTS:Differentiable architecture search[C]. Vancouver: International Conference on Learning Representations, 2018:931-938.
[20]	Chartsias A, Joyce T, Papanastasiou G, et al. Disentangled representation learning in cardiac image analysis[J]. Medical Image Analysis, 2019, 58(C):101535-101549.
[21]	Huang X, Liu M Y, Belongie S, et al. Multimodal unsupervised image-to-image translation[C]. Munich: Proceedings of the European Conference on Computer Vision, 2018:690-701.
[22]	Ma W D K, Lewis J P, Kleijn W B. The HSIC bottleneck: Deep learning without back-propagation[C]. New York: Proceedings of the AAAI Conference on Artificial Intelligence, 2020:529-537.
[23]	Li H, Wang Y, Wan R, et al. Domain generalization for medical imaging classification with linear-dependency regularization[C]. Vancouver: Advances in Neural Information Processing Systems, 2020:1017-1029.
[24]	Paszke A, Gross S, Massa F, et al. PyTorch:An imperative style,high-performance deep learning library[C]. Vancouver: Advances in Neural Information Processing Systems, 2019:998-1007.
[25]	Isensee F, Jaeger P F, Kohl S A A, et al. nnU-Net:A self-configuring method for deep learning-based biomedical image segmentation[J]. Nature Methods,Nature Publishing Group, 2021, 18(2):203-211.
[26]	Liu X, Thermos S, Chartsias A, et al. Disentangled representations for domain-generalized cardiac segmentation[C]. Lima: International Workshop on Statistical Atlases and Computational Models of the Heart, 2020:806-811.

上采样操作	下采样操作	正常采样操作
上采样深度卷积	下采样深度卷积	跳跃连接
上采样卷积	下采样卷积	无连接
上采样空洞卷积	下采样空洞卷积	空洞卷积
	最大池化	深度卷积
	平均池化	卷积

	源域	目标域	nnUNet^[25]	SDNet+Aug^[26]	LDDG^[23]	SAML^[11]	DGNet^[13]	本文方法
5%	BCD	A	65.30_17.0	71.21_13.0	66.22_9.1	67.11_10.0	72.40_12.0	75.12_11.3
	ACD	B	79.73_10.0	77.31_10.0	69.49_8.3	76.35_7.9	80.30_9.1	80.93_8.2
	ABD	C	78.06_11.0	81.40_8.0	73.40_9.8	77.43_8.3	82.51_6.6	80.86_6.1
	ABC	D	81.25_8.3	79.95_7.8	75.66_8.5	78.64_5.8	83.77_5.1	81.60_8.3
	平均值		76.09_6.3	77.47_3.9	71.29_3.6	74.88_4.6	79.75_4.4	79.62_8.5

	源域	目标域	nnUNet^[25]	SDNet+Aug^[26]	LDDG^[23]	SAML^[11]	DGNet^[13]	本文方法
2%	BCD	A	52.87_19.0	54.48_18.0	59.47_12.0	56.31_13.0	66.01_12.0	67.95_14.6
	ACD	B	64.63_17.0	67.81_14.0	56.16_14.0	56.32_15.0	72.72_10.0	74.07_10.1
	ABD	C	72.97_14.0	76.46_12.0	68.21_11.0	75.70_8.7	77.54_10.0	79.78_8.7
	ABC	D	73.27_11.0	74.35_11.0	68.56_10.0	69.94_9.8	75.14_8.4	74.33_12.5
	平均值		65.94_8.3	68.28_8.6	63.16_5.4	64.57_8.5	72.85_4.3	74.03_11.5

	源域	目标域	nnUNet^[25]	SDNet+Aug^[26]	LDDG^[23]	SAML^[11]	DGNet^[13]	本文方法
5%	BCD	A	23.04_6.7	22.84_6.3	23.35_5.7	23.10_5.9	22.55_6.6	18.38_5.4
	ACD	B	18.18_4.7	20.26_5.5	20.56_4.7	18.97_4.9	19.37_6.4	15.95_4.5
	ABD	C	16.44_4.2	16.22_3.9	17.14_3.3	16.29_3.2	15.77_3.8	13.77_2.8
	ABC	D	15.24_4.2	15.15_3.3	15.80_3.2	15.58_3.2	14.24_2.8	13.43_3.3
	平均值		18.22_3.0	18.62_3.1	19.21_3.0	18.49_2.9	17.98_3.2	15.38_4.0

	源域	目标域	nnUNet^[25]	SDNet+Aug^[26]	LDDG^[23]	SAML^[11]	DGNet^[13]	本文方法
2%	BCD	A	26.48_7.5	24.69_7.0	25.56_5.9	25.57_5.7	23.55_6.5	20.42_6.0
	ACD	B	23.11₆.₈	21.84_6.2	25.44_5.2	24.91_5.5	19.95_6.3	17.83_5.2
	ABD	C	16.75_4.6	16.57_4.2	18.98_3.9	16.46_3.5	16.29_4.0	14.19_3.1
	ABC	D	17.51_4.9	17.57_4.1	18.08_3.8	17.94_3.8	17.48_4.7	15.76_4.1
	平均值		20.96_4.0	20.17_3.3	22.02_3.5	21.22_4.1	19.32_2.8	17.05_4.6

Semi-Supervised Medical Image Segmentation Method Based on Meta-Learning and Neural Architecture Search

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 26

Related Articles 15

Metrics

Comments

Recommended 0

	结点0	结点1	结点2	结点3
下采样单元结构	跳跃连接	空洞卷积	空洞卷积	空洞卷积
下采样单元结构	下采样空洞卷积	下采样卷积	空洞卷积	空洞卷积
正常采样单元结构	跳跃连接	空洞卷积	空洞卷积	空洞卷积
正常采样单元结构	卷积	跳跃连接	空洞卷积	空洞卷积
上采样单元结构	跳跃连接	空洞卷积	空洞卷积	空洞卷积
上采样单元结构	上采样可分离卷积	上采样可分离卷积	上采样可分离卷积	空洞卷积

[1]	LI Zenghui,WANG Wei. Research Progress of Medical Image Segmentation Method Based on Deep Learning [J]. Electronic Science and Technology, 2024, 37(1): 72-80.
[2]	WANG Jiawei,YU Xiao. Review of Text Classification Research Based on Deep Learning [J]. Electronic Science and Technology, 2024, 37(1): 81-86.
[3]	HUANG Zixuan,LI Qiaoxing. Research on Identification of Urban Illegal Vehicles Based on Random Forest Model [J]. Electronic Science and Technology, 2024, 37(1): 66-71.
[4]	Bin ,WANG Sen. Visual Detection of Structural Cracks Using Depth Deformable Contour ModelLAI [J]. Electronic Science and Technology, 2023, 36(9): 35-40.
[5]	YUE Shengyao,XU Baiqiang,XU Guidong,XU Chenguang,ZHANG Sai. Super-Resolution Imaging of Laminate Debonding Defects via Deconvolutional Neural Network and Ultrasound Guided Waves [J]. Electronic Science and Technology, 2023, 36(8): 7-13.
[6]	SUN Hong,ZHAO Yingzhi. Lightweight Generative Adversarial Networks Based on Multi-Scale Gradient [J]. Electronic Science and Technology, 2023, 36(7): 32-38.
[7]	OU Jingyi,TIAN Ying,XIANG Xin,SONG Qizhe. Fault Diagnosis of Few Shot Industrial Process Based on Transfer BN-CNN Framework [J]. Electronic Science and Technology, 2023, 36(7): 49-55.
[8]	ZENG Xinxin,ZHANG Hongyan. A Farmland Parcel Extraction Network Based on Multi-Scale Semantic Information Enhancement [J]. Electronic Science and Technology, 2023, 36(7): 70-74.
[9]	LI Keran,CHEN Sheng,KE Panpan. A Method of Facial Mask Segmentation Based on CA-Net [J]. Electronic Science and Technology, 2023, 36(6): 64-71.
[10]	SHI Jianke,QIAO Meiying,LI Bingfeng,ZHAO Yan. Underwater Occlusion Target Detection Algorithm Based on Attention Mechanism [J]. Electronic Science and Technology, 2023, 36(5): 62-70.
[11]	LIU Yuwei,CAO Min,FENG Haojia. CNAS Recognition Criteria Automatic Benchmarking System Based on Natural Language Processing [J]. Electronic Science and Technology, 2023, 36(5): 28-33.
[12]	ZHENG Yuheng,FU Dongxiang. UAV Detection Based on Slim-YOLOv4 with Embedded Device [J]. Electronic Science and Technology, 2023, 36(5): 55-61.
[13]	LUO Ruipeng,FENG Mingke,HUANG Xin,ZOU Renling,LI Dan. A Review of Research on EEG Signal Preprocessing Methods [J]. Electronic Science and Technology, 2023, 36(4): 36-43.
[14]	CUI Zhuodong,CHEN Wei,YIN Zhong. Helmet Wearing Detection Based on Enhanced Feature Fusion Network [J]. Electronic Science and Technology, 2023, 36(4): 44-51.
[15]	SUN Hong,ZHANG Yuxiang. Super-Resolution Image Reconstruction Algorithm Based on Multi-Feature Gated Feedback Residual Network [J]. Electronic Science and Technology, 2023, 36(4): 65-70.