Prostate Cancer Histopathology Classification Using Multi-Instance Learning and Vision Transformers

• Felipe Navarro Balbino Alves Oncodata
• Matheus de Freitas Oliveira Baffa Oncodata
• Luiz Edmundo Lopes Mizutani Oncodata
• Adriano Brasileiro Silva Oncodata
• Fábio Rocha Fernandes Távora ARGOS Patologia
• Guilherme de Souza Velozo ARGOS Patologia
• Viviane Teixeira Loiola de Alencar Oncodata

Resumo

Prostate cancer is the most frequently diagnosed malignancy among men worldwide and remains a major public health concern. Although histopathological evaluation of prostate biopsies is the diagnostic gold standard, it faces limitations such as sampling errors, inter-observer variability, and limited access to specialized pathologists, underscoring the need for computational support. In this study, we propose a computer vision framework for automated classification of prostate histopathology using a multiple instance learning (MIL) approach built upon DINOv2-based foundation model embeddings. The system was trained and validated on a large dataset comprising prostate tissue whole-slide images from TCGA, GTEx, and a private Brazilian laboratory (ARGOS Patologia). Data were divided into training, testing, and an independent validation set, with patientlevel separation to prevent data leakage across subsets. The proposed model achieved an accuracy and sensitivity of 94% and an area under the receiver operating characteristic curve (AUC) of 98% on the independent validation set, demonstrating robust generalization across heterogeneous real-world samples. These results highlight the potential of MIL combined with foundation model representations to support reliable and scalable prostate cancer diagnosis.

Referências

Mayo Clinic, “Prostate cancer,” [link], 2025, accessed: 29 July 2025.

American Cancer Society, “What is prostate cancer?” [link], 2023, acessado em 13 de agosto de 2025.

International Agency for Research on Cancer (IARC), “Global cancer observatory: Cancer today,” [link], 2024, accessed: 29 July 2025.

Instituto Nacional do Câncer (INCA), “Estimativa de incidência de câncer no brasil, 2023–2025,” [link], 2025, accessed: 29 July 2025.

National Health Service (NHS), “Prostate cancer,” [link], 2025, accessed: 30 July 2025.

E. M. Schaeffer et al., “Prostate cancer, version 4.2023, nccn clinical practice guidelines in oncology,” Journal of the National Comprehensive Cancer Network, vol. 21, no. 10, pp. 1067–1096, Oct 2023.

N. Singhal et al., “A deep learning system for prostate cancer diagnosis and grading in whole slide images of core needle biopsies,” Scientific reports, vol. 12, no. 1, p. 3383, 2022.

N. M. Loorutu, H. Yazid, and K. S. Ab Rahman, “Prostate cancer classification based on histopathological images,” International Journal on Robotics, Automation and Sciences, vol. 5, no. 2, pp. 43–53, 2023.

M. Sarıateş and E. Özbay, “A classifier model using fine-tuned convolutional neural network and transfer learning approaches for prostate cancer detection,” Applied Sciences, vol. 15, no. 1, p. 225, 2024.

A. K. Chaurasia, H. C. Harris, P. W. Toohey, and A. W. Hewitt, “A generalised vision transformer-based self-supervised model for diagnosing and grading prostate cancer using histological images,” Prostate Cancer and Prostatic Diseases, pp. 1–9, 2025.

Z. Shao, H. Bian, Y. Chen, Y. Wang, J. Zhang, X. Ji et al., “Transmil: Transformer based correlated multiple instance learning for whole slide image classification,” Advances in neural information processing systems, vol. 34, pp. 2136–2147, 2021.

M. Y. Lu, D. F. Williamson, T. Y. Chen, R. J. Chen, M. Barbieri, and F. Mahmood, “Data-efficient and weakly supervised computational pathology on whole-slide images,” Nature biomedical engineering, vol. 5, no. 6, pp. 555–570, 2021.

X. Wang, S. Yang, J. Zhang, M. Wang, J. Zhang, J. Huang, W. Yang, and X. Han, “Transpath: Transformer-based self-supervised learning for histopathological image classification,” in International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2021, pp. 186–195.

H. Zhang, Y. Meng, Y. Zhao, Y. Qiao, X. Yang, S. E. Coupland, and Y. Zheng, “Dtfd-mil: Double-tier feature distillation multiple instance learning for histopathology whole slide image classification,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 18 802–18 812.