Tongue Color Classification in TCM with Noisy Labels via Confident-Learning-Assisted Knowledge Distillation

LI Yanping; ZHUO Li; SUN Liangliang; ZHANG Hui; LI Xiaoguang; YANG Yang; WEI Wei

doi:10.23919/cje.2022.00.040

Volume 32 Issue 1

Jan. 2023

Turn off MathJax

Article Contents

Article Navigation > Chinese Journal of Electronics > 2023 > 32(1): 140-150

LI Yanping, ZHUO Li, SUN Liangliang, et al., “Tongue Color Classification in TCM with Noisy Labels via Confident-Learning-Assisted Knowledge Distillation,” Chinese Journal of Electronics, vol. 32, no. 1, pp. 140-150, 2023, doi: 10.23919/cje.2022.00.040

Citation:

LI Yanping, ZHUO Li, SUN Liangliang, et al., “Tongue Color Classification in TCM with Noisy Labels via Confident-Learning-Assisted Knowledge Distillation,” Chinese Journal of Electronics, vol. 32, no. 1, pp. 140-150, 2023, doi: 10.23919/cje.2022.00.040

Citation:

PDF( 4709 KB)

Tongue Color Classification in TCM with Noisy Labels via Confident-Learning-Assisted Knowledge Distillation

doi: 10.23919/cje.2022.00.040

LI Yanping^{1, 2
,},
ZHUO Li^{1, 2
,},
SUN Liangliang^{1, 2},
ZHANG Hui^{1, 2},
LI Xiaoguang^{1, 2},
YANG Yang^{3, 4},
WEI Wei^{3, 4}

1.
Beijing Key Laboratory of Computational Intelligence and Intelligent System, Beijing University of Technology, Beijing 100124, China
2.
Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China
3.
Department of Gastroenterology, Wangjing Hospital, China Academy of Chinese Medical Sciences, Beijing 100102, China
4.
Beijing Key Laboratory of Traditional Chinese Medicine Treatment of Functional Gastrointestinal Diseases, Beijing 100102, China

Funds: This work was supported by the National Natural Science Foundation of China (61871006), and Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine (ZYYCXTD-C-202210)

More Information

Author Bio:
Yanping LI is an M.S. degree candidate at the Faculty of Information Technology, Beijing University of Technology (BUT), Beijing, China. She is also with the Beijing Key Laboratory of Computational Intelligence and Intelligent System, BUT. Her research interest is medical image processing and analysis. (Email: ypinglee@163.com)

Li ZHUO (corresponding author) received the Ph.D. degree in pattern recognition and intelligent system from BUT in 2004. She is currently a Professor with the Faculty of Information Technology, BUT. She is also with the Beijing Key Laboratory of Computational Intelligence and Intelligent System, BUT. She authored or coauthored more than 180 refereed journal and conference papers and has written six book chapters. Her current research interests include image/video processing, image recognition, and medical image processing. (Email: zhuoli@bjut.edu.cn)

Liangliang SUN is an M.S. degree candidate at the Faculty of Information Technology, Beijing University of Technology, Beijing, China. He is also with the Beijing Key Laboratory of Computational Intelligence and Intelligent System, Beijing University of Technology. His research interest is medical image processing and analysis

Hui ZHANG was born in 1982. He received the Ph.D. degree in signal and information processing from Beijing Institute of Technology, China in 2010. He is a Lecturer of electronic science and technology in Beijing University of Technology. His research interests include motion analysis and video processing

Xiaoguang LI received the B.S. degree in electronic information engineering and the Ph.D. degree in circuits and systems from BUT, in 2003 and 2008, respectively. He is currently an Associate Professor with BUT. He was a Research Associate with The Hong Kong Polytechnic University in 2009. He was a Visiting Scholar with The University of Sydney in 2012, and with the University of Southern California from 2017 to 2018. His current research interests include image processing and medical image analysis
Received Date: 2022-03-10
Accepted Date: 2022-06-21

Available Online: 2022-10-09

Publish Date: 2023-01-05

Abstract

Abstract

Tongue color is an important tongue diagnostic index for traditional Chinese medicine (TCM). Due to the individual experience of TCM experts as well as ambiguous boundaries among the tongue color categories, there often exist noisy labels in annotated samples. Deep neural networks trained with the noisy labeled samples often have poor generalization capability because they easily overfit on noisy labels. A novel framework named confident-learning-assisted knowledge distillation (CLA-KD) is proposed for tongue color classification with noisy labels. In this framework, the teacher network plays two important roles. On the one hand, it performs confident learning to identify, cleanse and correct noisy labels. On the other hand, it learns the knowledge from the clean labels, which will then be transferred to the student network to guide its training. Moreover, we elaborately design a teacher network in an ensemble manner, named E-CA²-ResNet18, to solve the unreliability and instability problem resulted from the insufficient data samples. E-CA²-ResNet18 adopts ResNet18 as the backbone, and integrates channel attention (CA) mechanism and activate or not activation function together, which facilitates to yield a better performance. The experimental results on three self-established TCM tongue datasets demonstrate that, our proposed CLA-KD can obtain a superior classification accuracy and good robustness with a lower network model complexity, reaching 94.49%, 92.21%, 93.43% on the three tongue image datasets, respectively.
- Traditional Chinese medicine,
- Tongue color classification,
- Confident learning,
- Learning from noisy labels,
- Knowledge distillation,
- Channel attention mechanism,
- ResNet18

FullText(HTML)

References(38)

References

[1]	G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint, arXiv: 1503.02531, 2015.
[2]	C. H. Wu, T. C. Chen, Y. C. Hsieh, et al., “A hybrid rule mining approach for cardiovascular disease detection in traditional chinese medicine,” Journal of Intelligent & Fuzzy Systems, vol.36, no.2, pp.861–870, 2019. doi: 10.3233/JIFS-169864
[3]	Y. Wang, Y. Liu, L. Yu, et al., “Research methods about data mining technology in the study medication rule on famous veteran Teran doctors of TCM,” in Proc. of 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Madrid, Spain, pp.1948–1952, 2018.
[4]	J. Hou, H. Y. Su, B. Yan, et al., “Classification of tongue color based on CNN,” in Proc. of 2017 IEEE 2nd International Conference on Big Data Analysis (ICBDA), Beijing, China, pp.725–729, 2017.
[5]	L. Chen, B. Wang, Y. Ma, et al., “The retrieval of the medical tongue images based on color analysis,” in Proc. of 2016 11th International Conference on Computer Science & Education (ICCSE), Nagoya, Japan, pp.113–117, 2016.
[6]	Y. Lu, X. Li, L. Zhuo, et al., “Dccn: A deep-color correction network for traditional Chinese medicine tongue images,” in Proc. of 2018 IEEE International Conference on Multimedia & Expo Workshops (ICME Workshops), San Diego, USA, pp.1–6, 2018.
[7]	P. L. Qu, H. Zhang, L. Zhuo, et al., “Automatic analysis of tongue substance color and coating color using sparse representation-based classifier,” in Proc. of 2016 International Conference on Progress in Informatics and Computing (PIC), Shanghai, China, pp.289–294, 2016.
[8]	P. A. Gutiérrez, M. Perez-Ortiz, J. Sanchez-Monedero, et al., “Ordinal regression methods: Survey and experimental study,” IEEE Transactions on Knowledge and Data Engineering, vol.28, no.1, pp.127–146, 2015. doi: 10.1109/TKDE.2015.2457911
[9]	J. Sánchez-Monedero, P. A. Gutiérrez, and M. Pérez-Ortiz, “ORCA: A Matlab/Octave toolbox for ordinal regression,” Journal of Machine Learning Research, vol.20, no.125, pp.1–5, 2019.
[10]	G. Algan and I. Ulusoy, “Image classification with deep learning in the presence of noisy labels: A survey,” Knowledge-Based Systems, vol.215, article no.106771, 2021.
[11]	C. Northcutt, L. Jiang, and I. Chuang, “Confident learning: Estimating uncertainty in dataset labels,” Journal of Artificial Intelligence Research, vol.70, pp.1373–1411, 2021. doi: 10.1613/jair.1.12125
[12]	T. Xiao, T. Xia, Y. Yang, et al., “Learning from massive noisy labeled data for image classification,” in Proc. of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, pp.2691–2699, 2015.
[13]	A. Veit, N. Alldrin, G. Chechik, et al., “Learning from noisy large-scale datasets with minimal supervision,” in Proc. of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, USA, pp.6575–6583, 2017.
[14]	S. J. Delany, N. Segata, and B. M. Namee, “Profiling instances in noise reduction,” Knowledge-Based Systems, vol.31, pp.28–40, 2012. doi: 10.1016/j.knosys.2012.01.015
[15]	J. Luengo, S. O. Shim, S. Alshomrani, et al., “CNC-NOS: Class noise cleaning by ensemble filtering and noise scoring,” Knowledge-Based Systems, vol.140, pp.27–49, 2018. doi: 10.1016/j.knosys.2017.10.026
[16]	Y. Li, J. Yang, Y. Song, et al., “Learning from noisy labels with distillation,” in Proc. of IEEE International Conference on Computer Vision (ICCV), Venice, Italy, pp.1928–1936, 2017.
[17]	B. Sun, S. Chen, J. Wang, et al., “A robust multi-class AdaBoost algorithm for mislabeled noisy data,” Knowledge-Based Systems, vol.102, pp.87–102, 2016. doi: 10.1016/j.knosys.2016.03.024
[18]	H. Song, M. Kim, D. Park, et al., “Learning from noisy labels with deep neural networks: A survey,” arXiv preprint, arXiv: 2007.08199, 2020.
[19]	N. Manwani and P. S. Sastry, “Noise tolerance under risk minimization,” IEEE Transactions on Cybernetics, vol.43, no.3, pp.1146–1151, 2013. doi: 10.1109/TSMCB.2012.2223460
[20]	A. Ghosh, H. Kumar, and P. S. Sastry, “Robust loss functions under label noise for deep neural networks,” in Proc. of the AAAI Conference on Artificial Intelligence (AAAI), AAAI Press, Palo Alto, CA, USA, vol.31, no.1, DOI: 10.1609/aaai.v31i1.10894, 2017.
[21]	X. Wang, E. Kodirov, Y. Hua, et al., “Improved mean absolute error for learning meaningful patterns from abnormal training data,” arXiv preprint, arXiv: 1903.12141v5, 2019.
[22]	L. P. F. Garcia, A. C. P. L. F. de Carvalho, and A. C. Lorena, “Noise detection in the meta-learning level,” Neurocomputing, vol.176, pp.14–25, 2016. doi: 10.1016/j.neucom.2014.12.100.
[23]	D. Angluin and P. Laird, “Learning from noisy examples,” Machine Learning, vol.2, no.4, pp.343–370, 1988.
[24]	L. P. F. Garcia, J. Lehmann, A. C. P. L. F. de Carvalho, et al., “New label noise injection methods for the evaluation of noise filters,” Knowledge-Based Systems, vol.163, pp.693–704, 2019. doi: 10.1016/j.knosys.2018.09.031.
[25]	G. Algan and I. Ulusoy. “Label noise types and their effects on deep learning,” arXiv preprint, arXiv: 2003.10471, 2020.
[26]	J. H. Cho and B. Hariharan, “On the efficacy of knowledge distillation,” in Proc. of the IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea (South), pp.4794–4802, 2019.
[27]	C. Yang, L. Xie, S. Qiao, et al., “Training deep neural networks in generations: A more tolerant teacher educates better students,” in Proc. of the AAAI Conference on Artificial Intelligence (AAAI), AAAI Press, Palo Alto, CA, USA, vol.33, no.01, DOI: 10.1609/aaai.v33i01.33015628, 2019.
[28]	L. Chen, H. Zhang, J. Xiao, et al., “SCA-CNN: Spatial and channel-wise attention in convolutional networks for image captioning,” in Proc. of Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, pp.6298–6306, 2017.
[29]	S. Woo, J. Park, J. Y. Lee, et al., “CBAM: Convolutional block attention module,” in Proc. of the European Conf. on Computer Vision, Munich, Germany, pp.3–19, 2018.
[30]	N. Ma, X. Zhang, M. Liu, et al., “Activate or not: Learning customized activation,” in Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, pp.8032–8042, 2021.
[31]	Z. Shen, Z. He, and X. Xue, “MEAL: Multi-model ensemble via adversarial learning,” in Proc. of the AAAI Conf. on Artificial Intelligence, AAAI Press, Palo Alto, CA, USA, vol.33, no.1, DOI: 10.1609/aaai.v33i01.33014886, 2019.
[32]	L. S. Shen, Y. H. Cai, and X. F. Zhang, Collection and Analysis of Tongue Picture in Traditional Chinese Medicine, Beijing University of Technology Press, Beijing, China, pp.26–30, 2007. (in Chinese)
[33]	C. Szegedy, V. Vanhoucke, S. Ioffe, et al., “Rethinking the inception architecture for computer vision,” in Proc. of 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, pp.2818–2826, 2016.
[34]	S. Reed, H. Lee, D. Anguelov, et al., “Training deep neural networks on noisy labels with bootstrapping,” arXiv preprint, arXiv: 1412.6596, 2014.
[35]	Z. Zhang and M. R. Sabuncu, “Generalized cross entropy loss for training deep neural networks with noisy labels,” in Proc. of 32nd Conference on Neural Information Processing Systems, Montréal, Canada, pp.8792–8802, 2018.
[36]	K. Yi and J. Wu, “Probabilistic end-to-end noise correction for learning with noisy labels,” in Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, pp.7017–7025, 2019.
[37]	B. Han, Q. Yao, X. Yu, et al., “Co-teaching: Robust training of deep neural networks with extremely noisy labels,” arXiv preprint, arXiv: 1804.06872, 2018.
[38]	X. Yu, B. Han, J. Yao, et al., “How does disagreement help generalization against label corruption?,” in Proc. of International Conference on Machine Learning (ICML), Long Beach, CA, USA, pp.7164–7173, 2019.