基于置信学习和协同训练的医学图像分割方法

doi:10.13229/j.cnki.jdxbgxb.20230833

吉林大学学报(工学版) ›› 2025, Vol. 55 ›› Issue (5): 1675-1681.doi: 10.13229/j.cnki.jdxbgxb.20230833

基于置信学习和协同训练的医学图像分割方法

赵宏伟^1,²(),周明珠¹,刘萍萍^1,²(),周求湛³

^1.吉林大学计算机科学与技术学院，长春 130012
^2.吉林大学符号计算与知识工程教育部重点实验室，长春 130012
^3.吉林大学通信工程学院，长春 130022

收稿日期:2023-08-07 出版日期:2025-05-01 发布日期:2025-07-18
通讯作者: 刘萍萍 E-mail:zhaohw@jlu.edu.cn;liupp@jlu.edu.cn
作者简介:赵宏伟（1962-），男，教授，博士. 研究方向：嵌入式人工智能. E-mail：zhaohw@jlu.edu.cn
基金资助:
吉林省产业关键核心技术攻关项目(20230201085GX)

Medical image segmentation based on confident learning and collaborative training

Hong-wei ZHAO^1,²(),Ming-zhu ZHOU¹,Ping-ping LIU^1,²(),Qiu-zhan ZHOU³

^1.College of Computer Science and Technology，Jilin University，Changchun 130012，China
^2.Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education，Jilin University，Changchun 130012，China
^3.College of Communication Engineering，Jilin University，Changchun 130022，China

Received:2023-08-07 Online:2025-05-01 Published:2025-07-18
Contact: Ping-ping LIU E-mail:zhaohw@jlu.edu.cn;liupp@jlu.edu.cn

摘要/Abstract

摘要：

在医学图像低质量标注数据的训练中，置信学习发挥了重要的作用，但目前置信学习的应用均基于平均教师模型，并未探讨在其它网络上的可能性。针对该问题，提出了基于置信学习和协同训练的分割模型。该模型使用两个不同的网络，鼓励两网络的输出结果一致，接着将其中一网络的输出利用置信学习同原低质量标注进行对比，修正低质量标注数据从而使其提供有效训练参考。将提出模型在3个不同模态的医学图像数据集上进行对比，实验结果表明：提出模型的分割效果均优于已有的置信学习模型。

关键词: 计算机应用, 医学图像分割, 置信学习, 协同训练

Abstract:

Confident learning plays an important role in the training of low-quality labeled data of medical images， but the current application of confident learning is based on the mean teacher model， and the possibility on other networks is not discussed. To solve this problem， a segmentation model based on confident learning and collaborative training is proposed in this paper. The model uses two different networks， encourages the output of the two networks to be consistent， and then compares the output of one network with the original low-quality label by using confident learning to modify the low-quality labeled data so as to provide an effective training reference. The proposed model has been compared on three different modal medical image datasets， and the experimental results show that the segmentation effect of the model is better than that of the existing confident learning model.

Key words: computer application, medical image segmentation, confident learning, collaborative training

中图分类号:

TP391

赵宏伟,周明珠,刘萍萍,周求湛. 基于置信学习和协同训练的医学图像分割方法[J]. 吉林大学学报(工学版), 2025, 55(5): 1675-1681.

Hong-wei ZHAO,Ming-zhu ZHOU,Ping-ping LIU,Qiu-zhan ZHOU. Medical image segmentation based on confident learning and collaborative training[J]. Journal of Jilin University(Engineering and Technology Edition), 2025, 55(5): 1675-1681.

图/表 7

图1

图2

表1

表2

表3

图3

表4

参考文献 12

[1]	Zhang Y, Lu Y, Chen W, et al. MSMANet: a multi-scale mesh aggregation network for brain tumor segmentation[J]. Applied Soft Computing, 2021, 110: 107733.
[2]	Zhou Y, Yu H, Shi H. Study group learning: improving retinal vessel segmentation trained with noisy labels[C]∥Medical Image Computing and Computer Assisted Intervention-MICCAI 2021: 24th International Conference, Strasbourg, France, 2021: 57-67.
[3]	戴振晖, 简婉薇, 朱琳, 等. 基于3D UNet结合Transformer的肝脏及肝肿瘤自动分割[J]. 中国医疗设备, 2023, 38(1): 42-47.
	Dai Zhen-hui, Jian Wan-wei, Zhu Lin, et al. Automatic segmentation of the liver and liver tumor based on 3D unet combined with transformer[J]. China Medical Devices, 2023,38(1): 42-47.
[4]	Ren M, Zeng W, Yang B, et al. Learning to reweight examples for robust deep learning[C]∥International Conference on Machine Learning, Stockholm, Sweden, 2018: 4334-4343.
[5]	Goldberger J, Benreuven E. Training deep neural-networks using a noise adaptation layer[C]∥International Conference on Learning Representations, San Juan, Puerto Rico, 2016: 1-9.
[6]	Northcutt C, Jiang L, Chuang I. Confident learning: estimating uncertainty in dataset labels[J]. Journal of Artificial Intelligence Research, 2021, 70: 1373-1411.
[7]	Luo W, Yang M. Semi-supervised semantic segmentation via strong-weak dual-branch network[C]∥Computer Vision-ECCV 2020: 16th European Conference, Glasgow, UK, 2020: 784-800.
[8]	Xu Z, Lu D, Wang Y, et al. Noisy labels are treasure: mean-teacher-assisted confident learning for hepatic vessel segmentation[C]∥Medical Image Computing and Computer Assisted Intervention-MICCAI 2021: 24th International Conference, Strasbourg, France, 2021: 3-13.
[9]	Blum A, Mitchell T. Combining labeled and unlabeled data with co-training[C]∥Proceedings of the Eleventh Annual Conference on Computational Learning Theory, Madison, USA, 1998: 92-100.
[10]	Codella N, Rotemberg V, Tschandl P, et al. Skin lesion analysis toward melanoma detection 2018: a challenge hosted by the international skin imaging collaboration (isic)[J/OL]. [2023-07-23]. arXiv Preprint arXiv:.
[11]	Jaeger S, Karargyris A, Candemir S, et al. Automatic tuberculosis screening using chest radiographs[J]. IEEE Transactions on Medical Imaging, 2013, 33(2): 233-245.
[12]	Hssayeni M, Croock M, Salman A, et al. Intracranial Hemorrhage Segmentation Using a Deep Convolutional Model[J]. Data, 2020, 5(1): 1- 14.

相关文章 15

[1]	侯越,郭劲松,林伟,张迪,武月,张鑫. 分割可跨越车道分界线的多视角视频车速提取方法[J]. 吉林大学学报(工学版), 2025, 55(5): 1692-1704.
[2]	申自浩,高永生,王辉,刘沛骞,刘琨. 面向车联网隐私保护的深度确定性策略梯度缓存方法[J]. 吉林大学学报(工学版), 2025, 55(5): 1638-1647.
[3]	王友卫,刘奥,凤丽洲. 基于知识蒸馏和评论时间的文本情感分类新方法[J]. 吉林大学学报(工学版), 2025, 55(5): 1664-1674.
[4]	王军,司昌馥,王凯鹏,付强. 融合集成学习技术和PSO-GA算法的特征提取技术的入侵检测方法[J]. 吉林大学学报(工学版), 2025, 55(4): 1396-1405.
[5]	徐涛,孔帅迪,刘才华,李时. 异构机密计算综述[J]. 吉林大学学报(工学版), 2025, 55(3): 755-770.
[6]	赵孟雪,车翔玖,徐欢,刘全乐. 基于先验知识优化的医学图像候选区域生成方法[J]. 吉林大学学报(工学版), 2025, 55(2): 722-730.
[7]	车翔玖,武宇宁,刘全乐. 基于因果特征学习的有权同构图分类算法[J]. 吉林大学学报(工学版), 2025, 55(2): 681-686.
[8]	蔡晓东,周青松,张言言,雪韵. 基于动静态和关系特征全局捕获的社交推荐模型[J]. 吉林大学学报(工学版), 2025, 55(2): 700-708.
[9]	郭晓然,王铁君,闫悦. 基于局部注意力和本地远程监督的实体关系抽取方法[J]. 吉林大学学报(工学版), 2025, 55(1): 307-315.
[10]	汪豪,赵彬,刘国华. 基于时间和运动增强的视频动作识别[J]. 吉林大学学报(工学版), 2025, 55(1): 339-346.
[11]	刘元宁,臧子楠,张浩,刘震. 基于深度学习的核糖核酸二级结构预测方法[J]. 吉林大学学报(工学版), 2025, 55(1): 297-306.
[12]	李路,宋均琦,朱明,谭鹤群,周玉凡,孙超奇,周铖钰. 基于RGHS图像增强和改进YOLOv5网络的黄颡鱼目标提取[J]. 吉林大学学报(工学版), 2024, 54(9): 2638-2645.
[13]	赵宏伟,武鸿,马克,李海. 基于知识蒸馏的图像分类框架[J]. 吉林大学学报(工学版), 2024, 54(8): 2307-2312.
[14]	张云佐,郑宇鑫,武存宇,张天. 基于双特征提取网络的复杂环境车道线精准检测[J]. 吉林大学学报(工学版), 2024, 54(7): 1894-1902.
[15]	孙铭会,薛浩,金玉波,曲卫东,秦贵和. 联合时空注意力的视频显著性预测[J]. 吉林大学学报(工学版), 2024, 54(6): 1767-1776.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

k值	Dice	HD95
1：2	0.594 1	27.717 4
1：3	0.615 8	19.785 0
1：4	0.584 5	24.062 4

ω值	Dice	HD95
0	0.600 2	21.504 8
5	0.615 8	19.785 0
10	0.605 3	20.092 9

数据集	方法	Dice	HD95
皮肤病病灶分割	Full Decoupled MTCL	0.582 5 0.843 5 0.854 3	37.885 7 9.567 1 10.076 1
	CTCL	0.868 5	7.956 0
肺部X射线	Full Decoupled MTCL	0.863 9 0.892 9 0.881 3	6.001 2 3.059 5 4.069 2
	CTCL	0.904 9	2.407 3
颅内出血	Full Decoupled MTCL	0.419 9 0.585 2 0.578 1	22.213 0 24.473 3 20.539 0
	CTCL	0.615 8	19.785 0

U-Net	Transformer	协同训练	置信学习	Dice	HD95
√				0.419 9	22.213 0
	√			0.402 9	34.757 6
√	√	√		0.600 2	21.504 8
√	√	√	√	0.615 8	19.785 0

基于置信学习和协同训练的医学图像分割方法

Medical image segmentation based on confident learning and collaborative training

RICH HTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 12

相关文章 15

Metrics

本文评价

推荐阅读 10