File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: MuRCL: Multi-Instance Reinforcement Contrastive Learning for Whole Slide Image Classification

TitleMuRCL: Multi-Instance Reinforcement Contrastive Learning for Whole Slide Image Classification
Authors
Keywordscontrastive learning
multi-instance learning
reinforcement learning
Whole slide image analysis
Issue Date7-Dec-2022
PublisherInstitute of Electrical and Electronics Engineers
Citation
IEEE Transactions on Medical Imaging, 2023, v. 42, n. 5, p. 1337-1348 How to Cite?
AbstractMulti-instance learning (MIL) is widely adop- ted for automatic whole slide image (WSI) analysis and it usually consists of two stages, i.e., instance feature extraction and feature aggregation. However, due to the 'weak supervision' of slide-level labels, the feature aggregation stage would suffer from severe over-fitting in training an effective MIL model. In this case, mining more information from limited slide-level data is pivotal to WSI analysis. Different from previous works on improving instance feature extraction, this paper investigates how to exploit the latent relationship of different instances (patches) to combat overfitting in MIL for more generalizable WSI classification. In particular, we propose a novel Multi-instance Rein- forcement Contrastive Learning framework (MuRCL) to deeply mine the inherent semantic relationships of different patches to advance WSI classification. Specifically, the proposed framework is first trained in a self-supervised manner and then finetuned with WSI slide-level labels. We formulate the first stage as a contrastive learning (CL) process, where positive/negative discriminative feature sets are constructed from the same patch-level feature bags of WSIs. To facilitate the CL training, we design a novel reinforcement learning-based agent to progressively update the selection of discriminative feature sets according to an online reward for slide-level feature aggregation. Then, we further update the model with labeled WSI data to regularize the learned features for the final WSI classification. Experimental results on three public WSI classification datasets (Camelyon16, TCGA-Lung and TCGA-Kidney) demonstrate that the proposed MuRCL outperforms state-of-the-art MIL models. In addition, MuRCL can achieve comparable performance to other state-of-the-art MIL models on TCGA-Esca dataset.
Persistent Identifierhttp://hdl.handle.net/10722/331045
ISSN
2023 Impact Factor: 8.9
2023 SCImago Journal Rankings: 3.703
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorZhu, Z-
dc.contributor.authorYu, L-
dc.contributor.authorWu, W-
dc.contributor.authorYu, R-
dc.contributor.authorZhang, D-
dc.contributor.authorWang, L-
dc.date.accessioned2023-09-21T06:52:18Z-
dc.date.available2023-09-21T06:52:18Z-
dc.date.issued2022-12-07-
dc.identifier.citationIEEE Transactions on Medical Imaging, 2023, v. 42, n. 5, p. 1337-1348-
dc.identifier.issn0278-0062-
dc.identifier.urihttp://hdl.handle.net/10722/331045-
dc.description.abstractMulti-instance learning (MIL) is widely adop- ted for automatic whole slide image (WSI) analysis and it usually consists of two stages, i.e., instance feature extraction and feature aggregation. However, due to the 'weak supervision' of slide-level labels, the feature aggregation stage would suffer from severe over-fitting in training an effective MIL model. In this case, mining more information from limited slide-level data is pivotal to WSI analysis. Different from previous works on improving instance feature extraction, this paper investigates how to exploit the latent relationship of different instances (patches) to combat overfitting in MIL for more generalizable WSI classification. In particular, we propose a novel Multi-instance Rein- forcement Contrastive Learning framework (MuRCL) to deeply mine the inherent semantic relationships of different patches to advance WSI classification. Specifically, the proposed framework is first trained in a self-supervised manner and then finetuned with WSI slide-level labels. We formulate the first stage as a contrastive learning (CL) process, where positive/negative discriminative feature sets are constructed from the same patch-level feature bags of WSIs. To facilitate the CL training, we design a novel reinforcement learning-based agent to progressively update the selection of discriminative feature sets according to an online reward for slide-level feature aggregation. Then, we further update the model with labeled WSI data to regularize the learned features for the final WSI classification. Experimental results on three public WSI classification datasets (Camelyon16, TCGA-Lung and TCGA-Kidney) demonstrate that the proposed MuRCL outperforms state-of-the-art MIL models. In addition, MuRCL can achieve comparable performance to other state-of-the-art MIL models on TCGA-Esca dataset.-
dc.languageeng-
dc.publisherInstitute of Electrical and Electronics Engineers-
dc.relation.ispartofIEEE Transactions on Medical Imaging-
dc.rightsThis work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.-
dc.subjectcontrastive learning-
dc.subjectmulti-instance learning-
dc.subjectreinforcement learning-
dc.subjectWhole slide image analysis-
dc.titleMuRCL: Multi-Instance Reinforcement Contrastive Learning for Whole Slide Image Classification-
dc.typeArticle-
dc.identifier.doi10.1109/TMI.2022.3227066-
dc.identifier.scopuseid_2-s2.0-85144763677-
dc.identifier.volume42-
dc.identifier.issue5-
dc.identifier.spage1337-
dc.identifier.epage1348-
dc.identifier.eissn1558-254X-
dc.identifier.isiWOS:000982483400009-
dc.identifier.issnl0278-0062-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats