Attention Induction Based on Pathologist Annotations for Improving Whole Slide Pathology Image Classifier

Overview

Journal J Pathol Inform

Date 2025 Jan 23

PMID 39845979

Authors

Ryoichi Koga

Tatsuya Yokota

Koji Arihiro

Hidekata Hontani

Affiliations

Soon will be listed here.

Abstract

We propose a method of to improve an attention mechanism in a whole slide image (WSI) classifier. Generally, only some regions in a WSI are useful for lesion classification, and the WSI classifier is required to find and focus on such regions for the classification. Multiple instance learning and hierarchical representation learning are widely employed for WSI processing and both use attention mechanisms to automatically find the useful regions and then conduct the class prediction. Here, it is impractical to collect a large number of WSIs, and when the attention mechanism is trained with a small number of training WSIs, the resultant attention often fails to focus on the useful regions. To improve the attention mechanism without increasing the number of training WSIs, we propose a method of attention induction for a hierarchical representation of WSI that guides attention to focus on the regions useful for lesion classification based on pathologist's coarse annotations. Our experimental results demonstrate that the proposed method improves the attention mechanism, thereby enhancing the performance of WSI classification.

References

Takateyama Y, Haruishi T, Hashimoto M, Otake Y, Akashi T, Shimizu A . Attention induction for a CT volume classification of COVID-19. Int J Comput Assist Radiol Surg. 2022; 18(2):289-301. PMC: 9574825. DOI: 10.1007/s11548-022-02769-y. View

Koga R, Koide S, Tanaka H, Taguchi K, Kugler M, Yokota T . A study of criteria for grading follicular lymphoma using a cell type classifier from pathology images based on complementary-label learning. Micron. 2024; 184:103663. DOI: 10.1016/j.micron.2024.103663. View

Wilm F, Fragoso M, Marzahl C, Qiu J, Puget C, Diehl L . Pan-tumor CAnine cuTaneous Cancer Histology (CATCH) dataset. Sci Data. 2022; 9(1):588. PMC: 9515104. DOI: 10.1038/s41597-022-01692-w. View

Campanella G, Hanna M, Geneslaw L, Miraflor A, Silva V, Busam K . Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nat Med. 2019; 25(8):1301-1309. PMC: 7418463. DOI: 10.1038/s41591-019-0508-1. View

Abels E, Pantanowitz L, Aeffner F, Zarella M, van der Laak J, Bui M . Computational pathology definitions, best practices, and recommendations for regulatory guidance: a white paper from the Digital Pathology Association. J Pathol. 2019; 249(3):286-294. PMC: 6852275. DOI: 10.1002/path.5331. View

Hashimoto N, Ko K, Yokota T, Kohno K, Nakaguro M, Nakamura S . Subtype classification of malignant lymphoma using immunohistochemical staining pattern. Int J Comput Assist Radiol Surg. 2022; 17(7):1379-1389. PMC: 9206633. DOI: 10.1007/s11548-021-02549-0. View

Miyoshi H, Sato K, Kabeya Y, Yonezawa S, Nakano H, Takeuchi Y . Deep learning shows the capability of high-level computer-aided diagnosis in malignant lymphoma. Lab Invest. 2020; 100(10):1300-1310. DOI: 10.1038/s41374-020-0442-3. View

Hosseini M, Bejnordi B, Quoc-Huy Trinh V, Chan L, Hasan D, Li X . Computational pathology: A survey review and the way forward. J Pathol Inform. 2024; 15:100357. PMC: 10900832. DOI: 10.1016/j.jpi.2023.100357. View

Dooper S, Pinckaers H, Aswolinskiy W, Hebeda K, Jarkman S, van der Laak J . Gigapixel end-to-end training using streaming and attention. Med Image Anal. 2023; 88:102881. DOI: 10.1016/j.media.2023.102881. View

10.

Hou L, Samaras D, Kurc T, Gao Y, Davis J, Saltz J . Patch-based Convolutional Neural Network for Whole Slide Tissue Image Classification. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit. 2016; 2016:2424-2433. PMC: 5085270. DOI: 10.1109/CVPR.2016.266. View

11.

Hashimoto N, Hanada H, Miyoshi H, Nagaishi M, Sato K, Hontani H . Multimodal Gated Mixture of Experts Using Whole Slide Image and Flow Cytometry for Multiple Instance Learning Classification of Lymphoma. J Pathol Inform. 2024; 15:100359. PMC: 10844119. DOI: 10.1016/j.jpi.2023.100359. View

12.

Hashimoto N, Takagi Y, Masuda H, Miyoshi H, Kohno K, Nagaishi M . Case-based similar image retrieval for weakly annotated large histopathological images of malignant lymphoma using deep metric learning. Med Image Anal. 2023; 85:102752. DOI: 10.1016/j.media.2023.102752. View

13.

Deng R, Cui C, Remedios L, Bao S, Womick R, Chiron S . Cross-scale multi-instance learning for pathological image diagnosis. Med Image Anal. 2024; 94:103124. PMC: 11016375. DOI: 10.1016/j.media.2024.103124. View

14.

Xu H, Usuyama N, Bagga J, Zhang S, Rao R, Naumann T . A whole-slide foundation model for digital pathology from real-world data. Nature. 2024; 630(8015):181-188. PMC: 11153137. DOI: 10.1038/s41586-024-07441-w. View

15.

Veta M, Heng Y, Stathonikos N, Bejnordi B, Beca F, Wollmann T . Predicting breast tumor proliferation from whole-slide images: The TUPAC16 challenge. Med Image Anal. 2019; 54:111-121. DOI: 10.1016/j.media.2019.02.012. View