Benchmarking Pathology Feature Extractors for Whole Slide Image Classification

Weakly supervised whole slide image classification is a key task in computational pathology, which involves predicting a slide-level label from a set of image patches constituting the slide. Constructing models to solve this task involves multiple design choices, often made without robust empirical...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wölflein, Georg, Ferber, Dyke, Meneghetti, Asier R, Nahhas, Omar S. M. El, Truhn, Daniel, Carrero, Zunamys I, Harrison, David J, Arandjelović, Ognjen, Kather, Jakob Nikolas
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Wölflein, Georg
Ferber, Dyke
Meneghetti, Asier R
Nahhas, Omar S. M. El
Truhn, Daniel
Carrero, Zunamys I
Harrison, David J
Arandjelović, Ognjen
Kather, Jakob Nikolas
description Weakly supervised whole slide image classification is a key task in computational pathology, which involves predicting a slide-level label from a set of image patches constituting the slide. Constructing models to solve this task involves multiple design choices, often made without robust empirical or conclusive theoretical justification. To address this, we conduct a comprehensive benchmarking of feature extractors to answer three critical questions: 1) Is stain normalisation still a necessary preprocessing step? 2) Which feature extractors are best for downstream slide-level classification? 3) How does magnification affect downstream performance? Our study constitutes the most comprehensive evaluation of publicly available pathology feature extractors to date, involving more than 10,000 training runs across 14 feature extractors, 9 tasks, 5 datasets, 3 downstream architectures, 2 levels of magnification, and various preprocessing setups. Our findings challenge existing assumptions: 1) We observe empirically, and by analysing the latent space, that skipping stain normalisation and image augmentations does not degrade performance, while significantly reducing memory and computational demands. 2) We develop a novel evaluation metric to compare relative downstream performance, and show that the choice of feature extractor is the most consequential factor for downstream performance. 3) We find that lower-magnification slides are sufficient for accurate slide-level classification. Contrary to previous patch-level benchmarking studies, our approach emphasises clinical relevance by focusing on slide-level biomarker prediction tasks in a weakly supervised setting with external validation cohorts. Our findings stand to streamline digital pathology workflows by minimising preprocessing needs and informing the selection of feature extractors.
doi_str_mv 10.48550/arxiv.2311.11772
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2311_11772</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2311_11772</sourcerecordid><originalsourceid>FETCH-arxiv_primary_2311_117723</originalsourceid><addsrcrecordid>eNpjYJA0NNAzsTA1NdBPLKrILNMzMjY01DM0NDc34mTwc0rNS87ITSzKzsxLVwhILMnIz8lPr1RwS00sKS1KVXCtKClKTC7JLypWSMsvUggHSqcqBOdkpqQqeOYmpqcqOOckFhdnpmUmJ5Zk5ufxMLCmJeYUp_JCaW4GeTfXEGcPXbDN8QVFmUCrKuNBLogHu8CYsAoAlUE8Dg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Benchmarking Pathology Feature Extractors for Whole Slide Image Classification</title><source>arXiv.org</source><creator>Wölflein, Georg ; Ferber, Dyke ; Meneghetti, Asier R ; Nahhas, Omar S. M. El ; Truhn, Daniel ; Carrero, Zunamys I ; Harrison, David J ; Arandjelović, Ognjen ; Kather, Jakob Nikolas</creator><creatorcontrib>Wölflein, Georg ; Ferber, Dyke ; Meneghetti, Asier R ; Nahhas, Omar S. M. El ; Truhn, Daniel ; Carrero, Zunamys I ; Harrison, David J ; Arandjelović, Ognjen ; Kather, Jakob Nikolas</creatorcontrib><description>Weakly supervised whole slide image classification is a key task in computational pathology, which involves predicting a slide-level label from a set of image patches constituting the slide. Constructing models to solve this task involves multiple design choices, often made without robust empirical or conclusive theoretical justification. To address this, we conduct a comprehensive benchmarking of feature extractors to answer three critical questions: 1) Is stain normalisation still a necessary preprocessing step? 2) Which feature extractors are best for downstream slide-level classification? 3) How does magnification affect downstream performance? Our study constitutes the most comprehensive evaluation of publicly available pathology feature extractors to date, involving more than 10,000 training runs across 14 feature extractors, 9 tasks, 5 datasets, 3 downstream architectures, 2 levels of magnification, and various preprocessing setups. Our findings challenge existing assumptions: 1) We observe empirically, and by analysing the latent space, that skipping stain normalisation and image augmentations does not degrade performance, while significantly reducing memory and computational demands. 2) We develop a novel evaluation metric to compare relative downstream performance, and show that the choice of feature extractor is the most consequential factor for downstream performance. 3) We find that lower-magnification slides are sufficient for accurate slide-level classification. Contrary to previous patch-level benchmarking studies, our approach emphasises clinical relevance by focusing on slide-level biomarker prediction tasks in a weakly supervised setting with external validation cohorts. Our findings stand to streamline digital pathology workflows by minimising preprocessing needs and informing the selection of feature extractors.</description><identifier>DOI: 10.48550/arxiv.2311.11772</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Learning</subject><creationdate>2023-11</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,782,887</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2311.11772$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2311.11772$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Wölflein, Georg</creatorcontrib><creatorcontrib>Ferber, Dyke</creatorcontrib><creatorcontrib>Meneghetti, Asier R</creatorcontrib><creatorcontrib>Nahhas, Omar S. M. El</creatorcontrib><creatorcontrib>Truhn, Daniel</creatorcontrib><creatorcontrib>Carrero, Zunamys I</creatorcontrib><creatorcontrib>Harrison, David J</creatorcontrib><creatorcontrib>Arandjelović, Ognjen</creatorcontrib><creatorcontrib>Kather, Jakob Nikolas</creatorcontrib><title>Benchmarking Pathology Feature Extractors for Whole Slide Image Classification</title><description>Weakly supervised whole slide image classification is a key task in computational pathology, which involves predicting a slide-level label from a set of image patches constituting the slide. Constructing models to solve this task involves multiple design choices, often made without robust empirical or conclusive theoretical justification. To address this, we conduct a comprehensive benchmarking of feature extractors to answer three critical questions: 1) Is stain normalisation still a necessary preprocessing step? 2) Which feature extractors are best for downstream slide-level classification? 3) How does magnification affect downstream performance? Our study constitutes the most comprehensive evaluation of publicly available pathology feature extractors to date, involving more than 10,000 training runs across 14 feature extractors, 9 tasks, 5 datasets, 3 downstream architectures, 2 levels of magnification, and various preprocessing setups. Our findings challenge existing assumptions: 1) We observe empirically, and by analysing the latent space, that skipping stain normalisation and image augmentations does not degrade performance, while significantly reducing memory and computational demands. 2) We develop a novel evaluation metric to compare relative downstream performance, and show that the choice of feature extractor is the most consequential factor for downstream performance. 3) We find that lower-magnification slides are sufficient for accurate slide-level classification. Contrary to previous patch-level benchmarking studies, our approach emphasises clinical relevance by focusing on slide-level biomarker prediction tasks in a weakly supervised setting with external validation cohorts. Our findings stand to streamline digital pathology workflows by minimising preprocessing needs and informing the selection of feature extractors.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNpjYJA0NNAzsTA1NdBPLKrILNMzMjY01DM0NDc34mTwc0rNS87ITSzKzsxLVwhILMnIz8lPr1RwS00sKS1KVXCtKClKTC7JLypWSMsvUggHSqcqBOdkpqQqeOYmpqcqOOckFhdnpmUmJ5Zk5ufxMLCmJeYUp_JCaW4GeTfXEGcPXbDN8QVFmUCrKuNBLogHu8CYsAoAlUE8Dg</recordid><startdate>20231120</startdate><enddate>20231120</enddate><creator>Wölflein, Georg</creator><creator>Ferber, Dyke</creator><creator>Meneghetti, Asier R</creator><creator>Nahhas, Omar S. M. El</creator><creator>Truhn, Daniel</creator><creator>Carrero, Zunamys I</creator><creator>Harrison, David J</creator><creator>Arandjelović, Ognjen</creator><creator>Kather, Jakob Nikolas</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20231120</creationdate><title>Benchmarking Pathology Feature Extractors for Whole Slide Image Classification</title><author>Wölflein, Georg ; Ferber, Dyke ; Meneghetti, Asier R ; Nahhas, Omar S. M. El ; Truhn, Daniel ; Carrero, Zunamys I ; Harrison, David J ; Arandjelović, Ognjen ; Kather, Jakob Nikolas</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-arxiv_primary_2311_117723</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Wölflein, Georg</creatorcontrib><creatorcontrib>Ferber, Dyke</creatorcontrib><creatorcontrib>Meneghetti, Asier R</creatorcontrib><creatorcontrib>Nahhas, Omar S. M. El</creatorcontrib><creatorcontrib>Truhn, Daniel</creatorcontrib><creatorcontrib>Carrero, Zunamys I</creatorcontrib><creatorcontrib>Harrison, David J</creatorcontrib><creatorcontrib>Arandjelović, Ognjen</creatorcontrib><creatorcontrib>Kather, Jakob Nikolas</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Wölflein, Georg</au><au>Ferber, Dyke</au><au>Meneghetti, Asier R</au><au>Nahhas, Omar S. M. El</au><au>Truhn, Daniel</au><au>Carrero, Zunamys I</au><au>Harrison, David J</au><au>Arandjelović, Ognjen</au><au>Kather, Jakob Nikolas</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Benchmarking Pathology Feature Extractors for Whole Slide Image Classification</atitle><date>2023-11-20</date><risdate>2023</risdate><abstract>Weakly supervised whole slide image classification is a key task in computational pathology, which involves predicting a slide-level label from a set of image patches constituting the slide. Constructing models to solve this task involves multiple design choices, often made without robust empirical or conclusive theoretical justification. To address this, we conduct a comprehensive benchmarking of feature extractors to answer three critical questions: 1) Is stain normalisation still a necessary preprocessing step? 2) Which feature extractors are best for downstream slide-level classification? 3) How does magnification affect downstream performance? Our study constitutes the most comprehensive evaluation of publicly available pathology feature extractors to date, involving more than 10,000 training runs across 14 feature extractors, 9 tasks, 5 datasets, 3 downstream architectures, 2 levels of magnification, and various preprocessing setups. Our findings challenge existing assumptions: 1) We observe empirically, and by analysing the latent space, that skipping stain normalisation and image augmentations does not degrade performance, while significantly reducing memory and computational demands. 2) We develop a novel evaluation metric to compare relative downstream performance, and show that the choice of feature extractor is the most consequential factor for downstream performance. 3) We find that lower-magnification slides are sufficient for accurate slide-level classification. Contrary to previous patch-level benchmarking studies, our approach emphasises clinical relevance by focusing on slide-level biomarker prediction tasks in a weakly supervised setting with external validation cohorts. Our findings stand to streamline digital pathology workflows by minimising preprocessing needs and informing the selection of feature extractors.</abstract><doi>10.48550/arxiv.2311.11772</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2311.11772
ispartof
issn
language eng
recordid cdi_arxiv_primary_2311_11772
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
Computer Science - Learning
title Benchmarking Pathology Feature Extractors for Whole Slide Image Classification
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-11-30T04%3A51%3A15IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Benchmarking%20Pathology%20Feature%20Extractors%20for%20Whole%20Slide%20Image%20Classification&rft.au=W%C3%B6lflein,%20Georg&rft.date=2023-11-20&rft_id=info:doi/10.48550/arxiv.2311.11772&rft_dat=%3Carxiv_GOX%3E2311_11772%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true