Determining breast cancer biomarker status and associated morphological features using deep learning.
Breast cancer
Pathology
Journal
Communications medicine
ISSN: 2730-664X
Titre abrégé: Commun Med (Lond)
Pays: England
ID NLM: 9918250414506676
Informations de publication
Date de publication:
2021
2021
Historique:
received:
30
01
2021
accepted:
18
06
2021
entrez:
23
5
2022
pubmed:
24
5
2022
medline:
24
5
2022
Statut:
epublish
Résumé
Breast cancer management depends on biomarkers including estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 (ER/PR/HER2). Though existing scoring systems are widely used and well-validated, they can involve costly preparation and variable interpretation. Additionally, discordances between histology and expected biomarker findings can prompt repeat testing to address biological, interpretative, or technical reasons for unexpected results. We developed three independent deep learning systems (DLS) to directly predict ER/PR/HER2 status for both focal tissue regions (patches) and slides using hematoxylin-and-eosin-stained (H&E) images as input. Models were trained and evaluated using pathologist annotated slides from three data sources. Areas under the receiver operator characteristic curve (AUCs) were calculated for test sets at both a patch-level (>135 million patches, 181 slides) and slide-level ( The patch-level AUCs are 0.939 (95%CI 0.936-0.941), 0.938 (0.936-0.940), and 0.808 (0.802-0.813) for ER/PR/HER2, respectively. At the slide level, AUCs are 0.86 (95%CI 0.84-0.87), 0.75 (0.73-0.77), and 0.60 (0.56-0.64) for ER/PR/HER2, respectively. Interpretability analyses show known biomarker-histomorphology associations including associations of low-grade and lobular histology with ER/PR positivity, and increased inflammatory infiltrates with triple-negative staining. This study presents rapid breast cancer biomarker estimation from routine H&E slides and builds on prior advances by prioritizing interpretability of computationally learned features in the context of existing pathological knowledge.
Sections du résumé
Background
UNASSIGNED
Breast cancer management depends on biomarkers including estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 (ER/PR/HER2). Though existing scoring systems are widely used and well-validated, they can involve costly preparation and variable interpretation. Additionally, discordances between histology and expected biomarker findings can prompt repeat testing to address biological, interpretative, or technical reasons for unexpected results.
Methods
UNASSIGNED
We developed three independent deep learning systems (DLS) to directly predict ER/PR/HER2 status for both focal tissue regions (patches) and slides using hematoxylin-and-eosin-stained (H&E) images as input. Models were trained and evaluated using pathologist annotated slides from three data sources. Areas under the receiver operator characteristic curve (AUCs) were calculated for test sets at both a patch-level (>135 million patches, 181 slides) and slide-level (
Results
UNASSIGNED
The patch-level AUCs are 0.939 (95%CI 0.936-0.941), 0.938 (0.936-0.940), and 0.808 (0.802-0.813) for ER/PR/HER2, respectively. At the slide level, AUCs are 0.86 (95%CI 0.84-0.87), 0.75 (0.73-0.77), and 0.60 (0.56-0.64) for ER/PR/HER2, respectively. Interpretability analyses show known biomarker-histomorphology associations including associations of low-grade and lobular histology with ER/PR positivity, and increased inflammatory infiltrates with triple-negative staining.
Conclusions
UNASSIGNED
This study presents rapid breast cancer biomarker estimation from routine H&E slides and builds on prior advances by prioritizing interpretability of computationally learned features in the context of existing pathological knowledge.
Identifiants
pubmed: 35602213
doi: 10.1038/s43856-021-00013-3
pii: 13
pmc: PMC9037318
doi:
Types de publication
Journal Article
Langues
eng
Pagination
14Informations de copyright
© The Author(s) 2021.
Déclaration de conflit d'intérêts
Competing interestsThis study was funded by Google LLC and Verily Life Sciences. P.G., R.J., H.W., F.T., M.M., G.S.C., L.H.P., Y.L., C.H.M., D.F.S., and P.-H.C.C. are employees of Google LLC and own Alphabet stock. I.F.-A. and T.B. are consultants of Google LLC. M.T., D.J.D., E.A.R., P.R., N.O., J.H.W., and C.R. declare no competing interests.
Références
Am J Surg. 2008 Oct;196(4):549-51
pubmed: 18809061
Nat Commun. 2020 Nov 16;11(1):5727
pubmed: 33199723
J Clin Oncol. 2020 Apr 20;38(12):1346-1366
pubmed: 31928404
Nature. 2012 Oct 4;490(7418):61-70
pubmed: 23000897
NPJ Digit Med. 2021 Apr 19;4(1):71
pubmed: 33875798
Histopathology. 2020 Oct;77(4):631-645
pubmed: 32618014
NPJ Digit Med. 2019 Jun 21;2:56
pubmed: 31304402
Mod Pathol. 2010 May;23 Suppl 2:S60-4
pubmed: 20436504
Nat Med. 2018 Oct;24(10):1559-1567
pubmed: 30224757
Nat Med. 2019 Jul;25(7):1054-1056
pubmed: 31160815
JAMA Netw Open. 2019 Jul 3;2(7):e197700
pubmed: 31348505
Am J Clin Pathol. 2012 Dec;138(6):796-802
pubmed: 23161712
Br J Cancer. 2005 Oct 31;93(9):1046-52
pubmed: 16175185
NPJ Breast Cancer. 2018 Sep 3;4:30
pubmed: 30182055
Eur J Cancer. 2006 Nov;42(17):2946-53
pubmed: 16989997
JAMA. 2019 Jan 22;321(3):288-300
pubmed: 30667505
Sci Rep. 2020 Apr 29;10(1):7275
pubmed: 32350370
Cell. 2015 Oct 8;163(2):506-19
pubmed: 26451490
Clin Med Res. 2009 Jun;7(1-2):4-13
pubmed: 19574486
Arch Pathol Lab Med. 2011 Feb;135(2):233-42
pubmed: 21284444