Semi-Supervised Learning for Glomerular Crescent Classification

• Joacy M. da Silva UEFS
• Washington L. C. dos Santos FIOCRUZ
• Angelo A. Duarte UEFS
• Luciano R. de Oliveira UFBA
• Michele F. Angelo UEFS
• Angelo C. Loula UEFS

Resumo

Glomeruli, the fundamental structures of the kidney responsible for blood filtration, are susceptible to various pathologies, including glomerular crescents. This work investigates the application of semi-supervised deep learning (SSL) for the classification of histopathological images containing these lesions. We evaluate several configurations of a deep learning model for classifying crescent versus normal glomeruli. Our investigation specifically compares the impact of varying the ratio of labeled to unlabeled data and the effect of different input perturbations (e.g., Random Flip + Random Crop) on model performance. Based on accuracy, precision, sensitivity, and F1-score, the model trained with four labeled folds and using Random Flip + Random Crop perturbations emerged as the most effective configuration. In a separate experiment, we expanded the classification task to distinguish crescent from non-crescent glomeruli by adding images of three other lesion types to the dataset. Our results show that a combined training approach using Mean-Teacher and Supervised Contrastive Learning (SCL) did not yield significant performance improvements compared to the model trained solely with the Mean-Teacher method.

Referências

P. Chagas, L. Souza, I. Araújo, N. Aldeman, A. Duarte, M. Angelo, W. L. dos Santos, and L. Oliveira, “Classification of glomerular hypercellularity using convolutional features and support vector machine,” Artificial Intelligence in Medicine, vol. 103, p. 101808, Mar. 2020. [Online]. DOI: 10.1016/j.artmed.2020.101808

L. Anguiano, R. Kain, and H.-J. Anders, “The glomerular crescent: triggers, evolution, resolution, and implications for therapy,” Current Opinion in Nephrology and Hypertension, vol. 29, no. 3, pp. 302–309, May 2020. [Online]. DOI: 10.1097/mnh.0000000000000596

E. Uchino, K. Suzuki, N. Sato, R. Kojima, Y. Tamada, S. Hiragi, H. Yokoi, N. Yugami, S. Minamiguchi, H. Haga, M. Yanagita, and Y. Okuno, “Classification of glomerular pathological findings using deep learning and nephrologist–AI collective intelligence approach,” International Journal of Medical Informatics, vol. 141, p. 104231, Sep. 2020. [Online]. DOI: 10.1016/j.ijmedinf.2020.104231

Y. Kawazoe, K. Shimamoto, R. Yamaguchi, I. Nakamura, K. Yoneda, E. Shinohara, Y. Shintani-Domoto, T. Ushiku, T. Tsukamoto, and K. Ohe, “Computational pipeline for glomerular segmentation and association of the quantified regions with prognosis of kidney function in IgA nephropathy,” Diagnostics, vol. 12, no. 12, p. 2955, Nov. 2022. [Online]. DOI: 10.3390/diagnostics12122955

J. M. da Silva, M. F. Angelo, W. L. C. dos Santos, and A. C. Loula, “Aprendizado profundo na classificação de lesões crescentes glomerulares: modelos e condições,” in CONFERENCE ON GRAPHICS, PATTERNS AND IMAGES, 34. (SIBGRAPI). Gramado (Virtual), Brasil: Proceedings... Los Alamitos: IEEE Computer Society, 2021, 2021. [Online]. Available: [link]

A. Akatsuka, Y. Horai, and A. Akatsuka, “Automated recognition of glomerular lesions in the kidneys of mice by using deep learning,” Journal of Pathology Informatics, vol. 13, p. 100129, 2022. [Online]. DOI: 10.1016/j.jpi.2022.100129

J. Li, Q. He, Y. Liu, Y. Wang, T. Guan, J. Ye, Y. He, and Z. Wang, “Glomerular lesion recognition based on pathology images with annotation noise via noisy label learning,” IEEE Access, vol. 11, pp. 41 325–41 336, 2023. [Online]. DOI: 10.1109/access.2023.3269792

J. U. Becker, D. Mayerich, M. Padmanabhan, J. Barratt, A. Ernst, P. Boor, P. A. Cicalese, C. Mohan, H. V. Nguyen, and B. Roysam, “Artificial intelligence and machine learning in nephropathology,” Kidney International, vol. 98, no. 1, pp. 65–75, Jul. 2020. [Online]. DOI: 10.1016/j.kint.2020.02.027

P.-C. Chung, W.-J. Yang, T.-H. Wu, C.-R. Huang, and Y.-Y. Hsu, “Emerging research directions of deep learning for pathology image analysis,” in 2022 IEEE Biomedical Circuits and Systems Conference (BioCAS). IEEE, Oct. 2022. [Online]. DOI: 10.1109/biocas54905.2022.9948651

Q. Gui, H. Zhou, N. Guo, and B. Niu, “A survey of class-imbalanced semi-supervised learning,” Machine Learning, vol. 113, no. 8, p. 5057–5086, May 2023. [Online]. DOI: 10.1007/s10994-023-06344-7

W. D. G. de Oliveira and L. Berton, “A systematic review for class-imbalance in semi-supervised learning,” Artificial Intelligence Review, vol. 56, no. S2, p. 2349–2382, Sep. 2023. [Online]. DOI: 10.1007/s10462-023-10579-0

A. Tarvainen and H. Valpola, “Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results,” in Advances in Neural Information Processing Systems, I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, Eds., vol. 30. Curran Associates, Inc., 2017. [Online]. Available: [link]

M. Hyun, J. Jeong, and N. Kwak, “Class-imbalanced semi-supervised learning,” 2020. [Online]. Available: [link]

A. Oliver, A. Odena, C. A. Raffel, E. D. Cubuk, and I. Goodfellow, “Realistic evaluation of deep semi-supervised learning algorithms,” in Advances in Neural Information Processing Systems, S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, Eds., vol. 31. Curran Associates, Inc., 2018. [Online]. Available: [link]

H.-Y. Zhou, A. Oliver, J. Wu, and Y. Zheng, “When semi-supervised learning meets transfer learning: Training strategies, models and datasets,” 2018. [Online]. Available: [link]

S. Zagoruyko and N. Komodakis, “Wide residual networks,” 2016. [Online]. Available: [link]

A. Tarvainen and H. Valpola, “Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results,” Advances in neural information processing systems, vol. 30, 2017.

K. Sohn, D. Berthelot, N. Carlini, Z. Zhang, H. Zhang, C. A. Raffel, E. D. Cubuk, A. Kurakin, and C.-L. Li, “Fixmatch: Simplifying semi-supervised learning with consistency and confidence,” in Advances in Neural Information Processing Systems, H. Larochelle, M. Ranzato, R. Hadsell, M. Balcan, and H. Lin, Eds., vol. 33. Curran Associates, Inc., 2020, pp. 596–608. [Online]. Available: [link]

C. Wei, K. Sohn, C. Mellina, A. Yuille, and F. Yang, “Crest: A class-rebalancing self-training framework for imbalanced semi-supervised learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2021, pp. 10 857–10 866.

Y. Oh, D.-J. Kim, and I. S. Kweon, “Daso: Distribution-aware semantics-oriented pseudo-label for imbalanced semi-supervised learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 9786–9796.