Deep learning-based histopathologic assessment of kidney tissue

Meyke Hermsen, Thomasde Bel, Marjolijn Den Boer, Eric J. Steenbergen, Jesper Kers, Sandrine Florquin, Joris J.T.H. Roelofs, Mark D. Stegall, Mariam P. Alexander, Byron H. Smith, Bart Smeets, Luuk B. Hilbrands, Jeroen A.W.M.Vander Laak

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Background The development of deep neural networks is facilitating more advanced digital analysis of histopathologic images. We trained a convolutional neural network for multiclass segmentation of digitized kidney tissue sections stained with periodic acid-Schiff (PAS). Methods We trained the network using multiclass annotations from 40 whole-slide images of stained kidney transplant biopsies and applied it to four independent data sets. We assessed multiclass segmentation performance by calculating Dice coefficients for ten tissue classes on ten transplant biopsies from the RadboudUniversityMedicalCenter inNijmegen, TheNetherlands, and on ten transplant biopsies from an external center for validation. We also fully segmented 15 nephrectomy samples and calculated the network's glomerular detection rates and compared network-based measures with visually scored histologic components (Banff classification) in 82 kidney transplant biopsies. Results The weighted mean Dice coefficients of all classes were 0.80 and 0.84 in ten kidney transplant biopsies from the Radboud center and the external center, respectively. The best segmented class was "glomeruli" in both data sets (Dice coefficients, 0.95 and 0.94, respectively), followed by "tubuli combined" and "interstitium." The network detected 92.7%of all glomeruli in nephrectomy samples, with 10.4% false positives. In whole transplant biopsies, the mean intraclass correlation coefficient for glomerular counting performed by pathologists versus the network was 0.94. We found significant correlations between visually scored histologic components and network-based measures. Conclusions This study presents the first convolutional neural network for multiclass segmentation of PAS-stained nephrectomy samples and transplant biopsies. Our network may have utility for quantitative studies involving kidney histopathology across centers and provide opportunities for deep learning applications in routine diagnostics.

Original languageEnglish (US)
Pages (from-to)1968-1979
Number of pages12
JournalJournal of the American Society of Nephrology
Volume30
Issue number10
DOIs
StatePublished - Jan 1 2019

Fingerprint

Learning
Transplants
Kidney
Biopsy
Nephrectomy
Periodic Acid
Datasets

ASJC Scopus subject areas

  • Nephrology

Cite this

Hermsen, M., Bel, T., Boer, M. D., Steenbergen, E. J., Kers, J., Florquin, S., ... Laak, J. A. W. M. V. (2019). Deep learning-based histopathologic assessment of kidney tissue. Journal of the American Society of Nephrology, 30(10), 1968-1979. https://doi.org/10.1681/ASN.2019020144

Deep learning-based histopathologic assessment of kidney tissue. / Hermsen, Meyke; Bel, Thomasde; Boer, Marjolijn Den; Steenbergen, Eric J.; Kers, Jesper; Florquin, Sandrine; Roelofs, Joris J.T.H.; Stegall, Mark D.; Alexander, Mariam P.; Smith, Byron H.; Smeets, Bart; Hilbrands, Luuk B.; Laak, Jeroen A.W.M.Vander.

In: Journal of the American Society of Nephrology, Vol. 30, No. 10, 01.01.2019, p. 1968-1979.

Research output: Contribution to journalArticle

Hermsen, M, Bel, T, Boer, MD, Steenbergen, EJ, Kers, J, Florquin, S, Roelofs, JJTH, Stegall, MD, Alexander, MP, Smith, BH, Smeets, B, Hilbrands, LB & Laak, JAWMV 2019, 'Deep learning-based histopathologic assessment of kidney tissue', Journal of the American Society of Nephrology, vol. 30, no. 10, pp. 1968-1979. https://doi.org/10.1681/ASN.2019020144
Hermsen M, Bel T, Boer MD, Steenbergen EJ, Kers J, Florquin S et al. Deep learning-based histopathologic assessment of kidney tissue. Journal of the American Society of Nephrology. 2019 Jan 1;30(10):1968-1979. https://doi.org/10.1681/ASN.2019020144
Hermsen, Meyke ; Bel, Thomasde ; Boer, Marjolijn Den ; Steenbergen, Eric J. ; Kers, Jesper ; Florquin, Sandrine ; Roelofs, Joris J.T.H. ; Stegall, Mark D. ; Alexander, Mariam P. ; Smith, Byron H. ; Smeets, Bart ; Hilbrands, Luuk B. ; Laak, Jeroen A.W.M.Vander. / Deep learning-based histopathologic assessment of kidney tissue. In: Journal of the American Society of Nephrology. 2019 ; Vol. 30, No. 10. pp. 1968-1979.
@article{3daffdd86a574b23a1ef720df31f69d0,
title = "Deep learning-based histopathologic assessment of kidney tissue",
abstract = "Background The development of deep neural networks is facilitating more advanced digital analysis of histopathologic images. We trained a convolutional neural network for multiclass segmentation of digitized kidney tissue sections stained with periodic acid-Schiff (PAS). Methods We trained the network using multiclass annotations from 40 whole-slide images of stained kidney transplant biopsies and applied it to four independent data sets. We assessed multiclass segmentation performance by calculating Dice coefficients for ten tissue classes on ten transplant biopsies from the RadboudUniversityMedicalCenter inNijmegen, TheNetherlands, and on ten transplant biopsies from an external center for validation. We also fully segmented 15 nephrectomy samples and calculated the network's glomerular detection rates and compared network-based measures with visually scored histologic components (Banff classification) in 82 kidney transplant biopsies. Results The weighted mean Dice coefficients of all classes were 0.80 and 0.84 in ten kidney transplant biopsies from the Radboud center and the external center, respectively. The best segmented class was {"}glomeruli{"} in both data sets (Dice coefficients, 0.95 and 0.94, respectively), followed by {"}tubuli combined{"} and {"}interstitium.{"} The network detected 92.7{\%}of all glomeruli in nephrectomy samples, with 10.4{\%} false positives. In whole transplant biopsies, the mean intraclass correlation coefficient for glomerular counting performed by pathologists versus the network was 0.94. We found significant correlations between visually scored histologic components and network-based measures. Conclusions This study presents the first convolutional neural network for multiclass segmentation of PAS-stained nephrectomy samples and transplant biopsies. Our network may have utility for quantitative studies involving kidney histopathology across centers and provide opportunities for deep learning applications in routine diagnostics.",
author = "Meyke Hermsen and Thomasde Bel and Boer, {Marjolijn Den} and Steenbergen, {Eric J.} and Jesper Kers and Sandrine Florquin and Roelofs, {Joris J.T.H.} and Stegall, {Mark D.} and Alexander, {Mariam P.} and Smith, {Byron H.} and Bart Smeets and Hilbrands, {Luuk B.} and Laak, {Jeroen A.W.M.Vander}",
year = "2019",
month = "1",
day = "1",
doi = "10.1681/ASN.2019020144",
language = "English (US)",
volume = "30",
pages = "1968--1979",
journal = "Journal of the American Society of Nephrology : JASN",
issn = "1046-6673",
publisher = "American Society of Nephrology",
number = "10",

}

TY - JOUR

T1 - Deep learning-based histopathologic assessment of kidney tissue

AU - Hermsen, Meyke

AU - Bel, Thomasde

AU - Boer, Marjolijn Den

AU - Steenbergen, Eric J.

AU - Kers, Jesper

AU - Florquin, Sandrine

AU - Roelofs, Joris J.T.H.

AU - Stegall, Mark D.

AU - Alexander, Mariam P.

AU - Smith, Byron H.

AU - Smeets, Bart

AU - Hilbrands, Luuk B.

AU - Laak, Jeroen A.W.M.Vander

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Background The development of deep neural networks is facilitating more advanced digital analysis of histopathologic images. We trained a convolutional neural network for multiclass segmentation of digitized kidney tissue sections stained with periodic acid-Schiff (PAS). Methods We trained the network using multiclass annotations from 40 whole-slide images of stained kidney transplant biopsies and applied it to four independent data sets. We assessed multiclass segmentation performance by calculating Dice coefficients for ten tissue classes on ten transplant biopsies from the RadboudUniversityMedicalCenter inNijmegen, TheNetherlands, and on ten transplant biopsies from an external center for validation. We also fully segmented 15 nephrectomy samples and calculated the network's glomerular detection rates and compared network-based measures with visually scored histologic components (Banff classification) in 82 kidney transplant biopsies. Results The weighted mean Dice coefficients of all classes were 0.80 and 0.84 in ten kidney transplant biopsies from the Radboud center and the external center, respectively. The best segmented class was "glomeruli" in both data sets (Dice coefficients, 0.95 and 0.94, respectively), followed by "tubuli combined" and "interstitium." The network detected 92.7%of all glomeruli in nephrectomy samples, with 10.4% false positives. In whole transplant biopsies, the mean intraclass correlation coefficient for glomerular counting performed by pathologists versus the network was 0.94. We found significant correlations between visually scored histologic components and network-based measures. Conclusions This study presents the first convolutional neural network for multiclass segmentation of PAS-stained nephrectomy samples and transplant biopsies. Our network may have utility for quantitative studies involving kidney histopathology across centers and provide opportunities for deep learning applications in routine diagnostics.

AB - Background The development of deep neural networks is facilitating more advanced digital analysis of histopathologic images. We trained a convolutional neural network for multiclass segmentation of digitized kidney tissue sections stained with periodic acid-Schiff (PAS). Methods We trained the network using multiclass annotations from 40 whole-slide images of stained kidney transplant biopsies and applied it to four independent data sets. We assessed multiclass segmentation performance by calculating Dice coefficients for ten tissue classes on ten transplant biopsies from the RadboudUniversityMedicalCenter inNijmegen, TheNetherlands, and on ten transplant biopsies from an external center for validation. We also fully segmented 15 nephrectomy samples and calculated the network's glomerular detection rates and compared network-based measures with visually scored histologic components (Banff classification) in 82 kidney transplant biopsies. Results The weighted mean Dice coefficients of all classes were 0.80 and 0.84 in ten kidney transplant biopsies from the Radboud center and the external center, respectively. The best segmented class was "glomeruli" in both data sets (Dice coefficients, 0.95 and 0.94, respectively), followed by "tubuli combined" and "interstitium." The network detected 92.7%of all glomeruli in nephrectomy samples, with 10.4% false positives. In whole transplant biopsies, the mean intraclass correlation coefficient for glomerular counting performed by pathologists versus the network was 0.94. We found significant correlations between visually scored histologic components and network-based measures. Conclusions This study presents the first convolutional neural network for multiclass segmentation of PAS-stained nephrectomy samples and transplant biopsies. Our network may have utility for quantitative studies involving kidney histopathology across centers and provide opportunities for deep learning applications in routine diagnostics.

UR - http://www.scopus.com/inward/record.url?scp=85072791755&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072791755&partnerID=8YFLogxK

U2 - 10.1681/ASN.2019020144

DO - 10.1681/ASN.2019020144

M3 - Article

C2 - 31488607

AN - SCOPUS:85072791755

VL - 30

SP - 1968

EP - 1979

JO - Journal of the American Society of Nephrology : JASN

JF - Journal of the American Society of Nephrology : JASN

SN - 1046-6673

IS - 10

ER -