Automatic data augmentation to improve generalization of deep learning in H&E stained histopathology
- PMID: 38281317
- DOI: 10.1016/j.compbiomed.2024.108018
Automatic data augmentation to improve generalization of deep learning in H&E stained histopathology
Abstract
In histopathology practice, scanners, tissue processing, staining, and image acquisition protocols vary from center to center, resulting in subtle variations in images. Vanilla convolutional neural networks are sensitive to such domain shifts. Data augmentation is a popular way to improve domain generalization. Currently, state-of-the-art domain generalization in computational pathology is achieved using a manually curated set of augmentation transforms. However, manual tuning of augmentation parameters is time-consuming and can lead to sub-optimal generalization performance. Meta-learning frameworks can provide efficient ways to find optimal training hyper-parameters, including data augmentation. In this study, we hypothesize that an automated search of augmentation hyper-parameters can provide superior generalization performance and reduce experimental optimization time. We select four state-of-the-art automatic augmentation methods from general computer vision and investigate their capacity to improve domain generalization in histopathology. We analyze their performance on data from 25 centers across two different tasks: tumor metastasis detection in lymph nodes and breast cancer tissue type classification. On tumor metastasis detection, most automatic augmentation methods achieve comparable performance to state-of-the-art manual augmentation. On breast cancer tissue type classification, the leading automatic augmentation method significantly outperforms state-of-the-art manual data augmentation.
Keywords: AutoML; Computational pathology; Data augmentation; Domain shift; Generalization; H&E.
Copyright © 2024 The Authors. Published by Elsevier Ltd.. All rights reserved.
Conflict of interest statement
Declaration of competing interest Geert Litjens reports financial support was provided by Dutch Research Council. Geert Litjens reports financial support was provided by HealthHolland. Geert Litjens reports a relationship with Canon Health Informatics that includes: consulting or advisory. Geert Litjens reports a relationship with Aiosyn that includes: equity or stocks. Jeroen van der Laak reports a relationship with Philips NV, Netherlands that includes: consulting or advisory and funding grants. Jeroen van der Laak reports a relationship with ContextVision, Sweden that includes: consulting or advisory and funding grants. Jeroen van der Laak reports a relationship with Aiosyn, Netherlands that includes: employment and equity or stocks. Jeroen van der Laak reports a relationship with Sectra, Sweden that includes: funding grants. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Similar articles
-
Clinical applications of graph neural networks in computational histopathology: A review.Comput Biol Med. 2023 Sep;164:107201. doi: 10.1016/j.compbiomed.2023.107201. Epub 2023 Jun 30. Comput Biol Med. 2023. PMID: 37517325 Review.
-
Domain generalization in deep learning based mass detection in mammography: A large-scale multi-center study.Artif Intell Med. 2022 Oct;132:102386. doi: 10.1016/j.artmed.2022.102386. Epub 2022 Aug 24. Artif Intell Med. 2022. PMID: 36207090
-
LiverNet: efficient and robust deep learning model for automatic diagnosis of sub-types of liver hepatocellular carcinoma cancer from H&E stained liver histopathology images.Int J Comput Assist Radiol Surg. 2021 Sep;16(9):1549-1563. doi: 10.1007/s11548-021-02410-4. Epub 2021 May 30. Int J Comput Assist Radiol Surg. 2021. PMID: 34053009
-
Deep computational pathology in breast cancer.Semin Cancer Biol. 2021 Jul;72:226-237. doi: 10.1016/j.semcancer.2020.08.006. Epub 2020 Aug 17. Semin Cancer Biol. 2021. PMID: 32818626 Review.
-
Staining Invariant Features for Improving Generalization of Deep Convolutional Neural Networks in Computational Pathology.Front Bioeng Biotechnol. 2019 Aug 23;7:198. doi: 10.3389/fbioe.2019.00198. eCollection 2019. Front Bioeng Biotechnol. 2019. PMID: 31508414 Free PMC article.
MeSH terms
LinkOut - more resources
Full Text Sources
Medical