Pneumoconiosis Staging Method Based on Uncertainty Estimation and Dynamic Sample Selection for Noisy Labels

doi:10.19678/j.issn.1000-3428.0070351

Computer Engineering ›› 2026, Vol. 52 ›› Issue (7): 411-420. doi: 10.19678/j.issn.1000-3428.0070351

• Interdisciplinary Integration and Engineering Applications • Previous Articles Next Articles

Pneumoconiosis Staging Method Based on Uncertainty Estimation and Dynamic Sample Selection for Noisy Labels

FENG Weiyi¹, REN Xueting¹, WANG Yueying¹, ZHAO Juanjuan¹^,²^,*(), QIANG Yan¹^,³, WANG Yingsen⁴, KONG Xiaomei⁵, ZHAO Huilan⁶

1. College of Computer Science and Technology (College of Data Science), Taiyuan University of Technology, Taiyuan 030024, Shanxi, China
2. School of Software, Taiyuan University of Technology, Taiyuan 030024, Shanxi, China
3. School of Software, North University of China, Taiyuan 030051, Shanxi, China
4. China Electric Power Research Institute, Beijing 100089, China
5. Department of Respiratory and Critical Care Medicine, The First Hospital of Shanxi Medical University, Taiyuan 030001, Shanxi, China
6. PET/CT Room, Shanxi Provincial Coal Central Hospital, Taiyuan 030002, Shanxi, China

Received:2024-09-10 Revised:2024-12-02 Online:2026-07-15 Published:2024-12-30
Contact: ZHAO Juanjuan

基于不确定性估计与动态样本选择的噪声标签尘肺分期方法

冯伟毅¹, 任雪婷¹, 王月莹¹, 赵涓涓¹^,²^,*(), 强彦¹^,³, 王盈森⁴, 孔晓梅⁵, 赵慧兰⁶

1. 太原理工大学计算机科学与技术学院(大数据学院), 山西太原 030024
2. 太原理工大学软件学院, 山西太原 030024
3. 中北大学软件学院, 山西太原 030051
4. 中国电力科学研究院有限公司, 北京 100089
5. 山西医科大学第一医院呼吸与危重症医学科, 山西太原 030001
6. 山西省煤炭中心医院PET/CT室, 山西太原 030002

通讯作者: 赵涓涓
作者简介:
冯伟毅(CCF学生会员), 男, 硕士研究生, 主研方向为医学图像处理
任雪婷, 博士研究生
王月莹, 博士研究生
赵涓涓(通信作者)
强彦, 教授、博士
王盈森, 博士研究生
孔晓梅, 主任医师、博士
赵慧兰, 主任医师
基金资助:
国家自然科学基金重点项目(U21A20469); 国家卫生健康委尘肺病重点实验室开放课题(YKFKT004); 国家自然科学基金(62376183); 山西省大健康产业高质量发展科研专项课题(DJKZXKT2023008); 山西省科技创新人才团队专项(202304051001009)

Abstract

Abstract:

During pneumoconiosis staging, the ambiguity in staging criteria leads to varying diagnostic results, causing instance label noise where label accuracy is affected by observer variability. Training deep learning models on such noisy labeled data often causes performance degradation. To address the challenges posed by noisy labels in pneumoconiosis staging, this study proposes a dual-branch dynamic sample selection strategy. In the key layer of the model, the method introduces Monte Carlo Dropout (MC-Dropout) and performs multiple forward propagation for uncertainty estimation of the prediction results, while combining the prediction probability of the model with the Gaussian Mixture Model (GMM) to gain insights into the feature spatial distribution of the samples. It ultimately realizes the branches of selective uncertainty estimation and Predicting the Feature Distributions of the GMM (FDP-GMM). Subsequently, a threshold value is set for each sample data in each branch, which is momentum updated with the model training. Finally, the dynamic sample selection module is used to divide the samples and differentiate the constraints to reduce the negative impact of noisy samples on the model. The method is evaluated on a clinically collected pneumoconiosis X-ray dataset, and the results show that it significantly outperforms similar methods at noise rates of 5%, 10%, 20%, and 40%, demonstrating its effectiveness and robustness in learning from noisy labels.

Key words: noisy label learning, pneumoconiosis staging, uncertainty estimation, sample selection, medical image classification

摘要：

由于病变分期标准存在模糊性, 不同诊断专家可能会得出不同的诊断结果, 导致实例标签的噪声, 即标签准确性受到观察者差异的影响。深度学习依赖于数据的质量, 而这些带有噪声标签的数据往往会导致模型性能下降。为应对尘肺病分期中噪声标签带来的挑战, 提出一种双分支动态样本选择策略。该方法通过在模型的关键层引入Monte Carlo Dropout(MC-Dropout), 在关键层进行多次前向传播进行不确定性估计预测结果, 同时将模型的预测概率与高斯混合模型(GMM)相结合洞察样本的特征空间分布, 最终实现可选择的不确定性估计分支和高斯混合模型特征分布预测(FDP-GMM)分支。然后为每一个分支中每一个样本数据设立一个阈值, 阈值随着模型训练的推进动量更新, 最终依据该动态样本选择模块进行样本划分和差异化约束, 减小噪声样本对模型的负面影响。在临床收集的尘肺病X光数据集上进行评估, 实验结果表明, 该方法在5%、10%、20%以及40%实例噪声比下的性能显著优于同类方法, 证明其在噪声标签学习中的有效性和鲁棒性。

关键词: 噪声标签学习, 尘肺分期, 不确定性估计, 样本选择, 医学图像分类

FENG Weiyi, REN Xueting, WANG Yueying, ZHAO Juanjuan, QIANG Yan, WANG Yingsen, KONG Xiaomei, ZHAO Huilan. Pneumoconiosis Staging Method Based on Uncertainty Estimation and Dynamic Sample Selection for Noisy Labels[J]. Computer Engineering, 2026, 52(7): 411-420.

冯伟毅, 任雪婷, 王月莹, 赵涓涓, 强彦, 王盈森, 孔晓梅, 赵慧兰. 基于不确定性估计与动态样本选择的噪声标签尘肺分期方法[J]. 计算机工程, 2026, 52(7): 411-420.

/ Recommend / Download Citations

URL: https://www.ecice06.com/EN/10.19678/j.issn.1000-3428.0070351

https://www.ecice06.com/EN/Y2026/V52/I7/411

Figures/Tables 9

Fig.1 Framework of dual-branch dynamic adaptive threshold sample selection algorithm

Fig.2 Prediction of feature distribution in Gaussian mixture models

Fig.3 Accuracy of different methods on pneumoconiosis chest X-ray datasets with different noise rates

Fig.4 Impact of Dropout Classifier forward propagation count on performance

Fig.5 Effect of different Dropout rates on performance

Fig.6 t-SNE feature space visualization under 20% noise rate

References 31

1	QI X M , LUO Y , SONG M Y , et al. Pneumoconiosis: current status and future prospects. Chinese Medical Journal, 2021, 134 (8): 898- 907. doi: 10.1097/CM9.0000000000001461
2	范存华, 周琅, 谢丽庄, 等. 江苏省非煤矿山行业尘肺病诊断调查. 中华劳动卫生职业病杂志, 2022, 40 (5): 358- 361.
	FAN C H , ZHOU L , XIE L Z , et al. Investigation on diagnosis of pneumoconiosis in non-coal mining industry in Jiangsu province. Chinese Journal of Industrial Hygiene and Occupational Diseases, 2022, 40 (5): 358- 361.
3	李忠学, 赵夏姿, 汤永中, 等. 国内尘肺病影像学诊断技术及临床应用价值的研究概况. 大众科技, 2022, 24 (8): 76- 79.
	LI Z X , ZHAO X Z , TANG Y Z , et al. Overview of research on imaging diagnostic technology and clinical application value of pneumoconiosis in China. Dazhong Keji, 2022, 24 (8): 76- 79.
4	SENER M U , SIMSEK C , ÖZKARA S , et al. Comparison of the international classification of high-resolution computed tomography for occupational and environmental respiratory diseases with the international labor organization international classification of radiographs of pneumoconiosis. Industrial Health, 2019, 57 (4): 495- 502. doi: 10.2486/indhealth.2018-0068
5	王峥, 钱青俊, 张建芳, 等. 计算机辅助诊断在尘肺病诊断中应用价值. 中国职业医学, 2020, 47 (4): 428- 431.
	WANG Z , QIAN Q J , ZHANG J F , et al. Application value of computer-aided diagnosis in diagnosing pneumoconiosis. China Occupational Medicine, 2020, 47 (4): 428- 431.
6	SUGANUMA N , YOSHIDA S , TAKEUCHI Y , et al. Artificial intelligence in quantitative chest imaging analysis for occupational lung disease. Seminars in Respiratory and Critical Care Medicine, 2023, 44 (3): 362- 369. doi: 10.1055/s-0043-1767760
7	DONG H T , ZHU B K , ZHANG X R , et al. Use data augmentation for a deep learning classification model with chest X-ray clinical imaging featuring coal workers' pneumoconiosis. BMC Pulmonary Medicine, 2022, 22 (1): 271. doi: 10.1186/s12890-022-02068-x
8	杭昊, 黄影平, 张栩瑞, 等. 面向道路场景语义分割的移动窗口变换神经网络设计. 光电工程, 2024, 51 (1): 230304.
	HANG H , HUANG Y P , ZHANG X R , et al. Design of swin transformer for semantic segmentation of road scenes. Opto-Electronic Engineering, 2024, 51 (1): 230304.
9	ARPIT D, JASTRZEBSKI S, BALLAS N, et al. A closer look at memorization in deep networks[C]//Proceedings of International Conference on Machine Learning. Sydney, Australia: International Machine Learning Society, 2017: 233-242.
10	崔锦莹, 梁立河, 任雪婷, 等. 基于噪声鲁棒学习的胸部X射线尘肺鉴别方法. 计算机工程, 2024, 50 (11): 350- 359.
	CUI J Y , LIANG L H , REN X T , et al. Method of noise robust learning based chest X-ray discrimination of pneumoconiosis. Computer Engineering, 2024, 50 (11): 350- 359.
11	SUN W J , WU D S , LUO Y , et al. ExpertNet: defeat noisy labels by deep expert consultation paradigm for pneumoconiosis staging on chest radiographs. Expert Systems with Applications, 2023, 232, 120710. doi: 10.1016/j.eswa.2023.120710
12	LU Y D, HE W B. SELC: self-ensemble label correction improves learning with noisy labels[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/2205.01156.
13	LI S K, XIA X B, GE S M, et al. Selective-supervised contrastive learning with noisy labels[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C., USA: IEEE Press, 2022: 316-325.
14	LI J N, SOCHER R, HOI S C H. DivideMix: learning with noisy labels as semi-supervised learning[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/2002.07394.
15	LIU Y. Understanding instance-level label noise: disparate impacts and treatments[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/2102.05336.
16	DAWSON G, POLIKAR R. Rethinking noisy label models: labeler-dependent noise with adversarial awareness[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/2105.14083.
17	ZHANG Y K, ZHENG S Z, WU P X, et al. Learning with feature-dependent label noise: a progressive approach[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/2103.07756.
18	ZHANG Z L, SABUNCU M R. Generalized cross entropy loss for training deep neural networks with noisy labels[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/1805.07836.
19	GAL Y, GHAHRAMANI Z. Dropout as a Bayesian approximation: representing model uncertainty in deep learning[C]//Proceeding of International Conference on Machine Learning. New York, USA: International Machine Learning Society, 2016: 1050-1059.
20	HAN B, YAO Q M, YU X R, et al. Co-teaching: robust training of deep neural networks with extremely noisy labels[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/1804.06872v3.
21	YU X R, HAN B, YAO J C, et al. How does disagreement help generalization against label corruption [EB/OL]. [2024-08-01]. https://arxiv.org/pdf/1901.04215.
22	TANAKA D, IKAMI D, YAMASAKI T, et al. Joint optimization framework for learning with noisy labels[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D.C., USA: IEEE Press, 2018: 5552-5560.
23	AYHAN M S , KVHLEWEIN L , ALIYEVA G , et al. Expert-validated estimation of diagnostic uncertainty for deep neural networks in diabetic retinopathy detection. Medical Image Analysis, 2020, 64, 101724. doi: 10.1016/j.media.2020.101724
24	RAGHU M, BLUMER K, SAYRES R, et al. Direct uncertainty prediction for medical second opinions[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/1807.01771.
25	JU L , WANG X , WANG L , et al. Improving medical images classification with label noise using dual-uncertainty estimation. IEEE Transactions on Medical Imaging, 2022, 41 (6): 1533- 1546. doi: 10.1109/TMI.2022.3141425
26	DEMPSTER A P , LAIRD N M , RUBIN D B . Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society Series B: Statistical Methodology, 1977, 39 (1): 1- 22. doi: 10.1111/j.2517-6161.1977.tb01600.x
27	HUANG Z Z, ZHANG J P, SHAN H M. Twin contrastive learning with noisy labels[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C., USA: IEEE Press, 2023: 11661-11670.
28	XUE C, DOU Q, SHI X Y, et al. Robust learning at noisy labeled medical images: applied to skin lesion classification[C]//Proceedings of the 16th International Symposium on Biomedical Imaging. Washington D.C., USA: IEEE Press, 2019: 1280-1283.
29	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Washington D.C., USA: IEEE Press, 2016: 770-778.
30	YI K, WU J X. Probabilistic end-to-end noise correction for learning with noisy labels[EB/OL]. [2024-08-01]. https://arxiv.org/pdf/1903.07788.
31	LI Y, HAN H, SHAN S, et al. DISC: learning from noisy labels via dynamic instance-specific selection and correction[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D.C., USA: IEEE Press, 2023: 24070-24079.

[1]	CUI Jinying, LIANG Lihe, REN Xueting, QIANG Yan, ZHAO Juanjuan, KONG Xiaomei, YU Xiao, ZHANG Hua. Method of Noise Robust Learning Based Chest X-ray Discrimination of Pneumoconiosis [J]. Computer Engineering, 2024, 50(11): 350-359.
[2]	LIU Lixia, XUAN Shibin, LIU Chang, LI Jiaxiang. Quantification of Optic Cup and Optic Disc Uncertainty with Multi-Expert Annotations [J]. Computer Engineering, 2023, 49(1): 250-257,269.
[3]	ZHANG Chiming, WANG Qingfeng, LIU Zhiqin, HUANG Jun, CHEN Bo, FU Jie, ZHOU Ying. Diagnostic Method of Frequently Occurring Chest Diseases Based on Deep Learning [J]. Computer Engineering, 2020, 46(7): 306-311,320.
[4]	ZHANG Chiming, WANG Qingfeng, LIU Zhiqing, HUANG Jun, ZHOU Ying, LIU Qiyu, XU Weiyun. Pulmonary Nodule Auxiliary Diagnosis Method Based on Deep Transfer Learning [J]. Computer Engineering, 2020, 46(1): 271-278.
[5]	JIANG Wen-han; ZHOU Xiao-fei; YANG Jing-yu. Kernel Subclass Convex Hull Sample Selection Method and Its Application on SVM [J]. Computer Engineering, 2008, 34(16): 212-214.
[6]	ZHOU Xiao-fei; JIANG Wen-han; YANG Jing-yu. Nearest Neighbor Convex Hull Classifier Based on Subspace Sample Selection [J]. Computer Engineering, 2008, 34(12): 167-168.
[7]	WEN Yimin. Method for Flatted and Parallel Training Samples Selection Based on Support Vectors [J]. Computer Engineering, 2006, 32(21): 177-179，.

Please choose a citation manager

Content to export