Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Deep Learning–Based Retinoblastoma Protein Subtyping of Pulmonary Large-Cell Neuroendocrine Carcinoma on Small Hematoxylin and Eosin–Stained Specimens

Trandafir, Teodora E. ; Heijboer, Frank W.J. ; Akram, Farhan ; Derks, Jules L. ; Li, Yunlei ; Hillen, Lisa M. ; Speel, Ernst Jan M. ; Megyesfalvi, Zsolt ; Dome, Balazs LU and Stubbs, Andrew P. , et al. (2025) In Laboratory Investigation 105(9).
Abstract

Molecular subtyping of pulmonary large-cell neuroendocrine carcinoma (LCNEC) based on retinoblastoma protein (pRb) expression may influence systemic treatment decisions. Current histomorphologic assessments of hematoxylin and eosin–stained tissue samples cannot reliably differentiate LCNEC molecular subtypes. This study explores the potential of deep learning (DL) to identify histologic patterns that distinguish these subtypes, by developing a custom convolutional neural network to predict the binary expression of pRb in small LCNEC tissue samples. Our model was trained, cross-validated, and tested on tissue microarray cores from 143 resection specimens and biopsies from 21 additional patients, with corresponding immunohistochemical pRb... (More)

Molecular subtyping of pulmonary large-cell neuroendocrine carcinoma (LCNEC) based on retinoblastoma protein (pRb) expression may influence systemic treatment decisions. Current histomorphologic assessments of hematoxylin and eosin–stained tissue samples cannot reliably differentiate LCNEC molecular subtypes. This study explores the potential of deep learning (DL) to identify histologic patterns that distinguish these subtypes, by developing a custom convolutional neural network to predict the binary expression of pRb in small LCNEC tissue samples. Our model was trained, cross-validated, and tested on tissue microarray cores from 143 resection specimens and biopsies from 21 additional patients, with corresponding immunohistochemical pRb status. The best-performing DL model achieved a patient-wise balanced accuracy value of 0.75 and an area under the receiver operating characteristic curve value of 0.77 when tested on biopsies, significantly outperforming the hematoxylin and eosin–based subtype classification by lung pathologists. Explainable artificial intelligence techniques further highlighted coarse chromatin patterns and distinct nucleoli as distinguishing features for pRb retained status. Meanwhile, pRb lost cases were characterized by limited cytoplasm and morphologic similarities with small cell lung cancer. These findings suggest that DL analysis of small histopathology samples could ultimately replace immunohistochemical pRb testing. Such a development may assist in guiding chemotherapy decisions, particularly in metastatic cases.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; ; ; ; ; and , et al. (More)
; ; ; ; ; ; ; ; ; ; and (Less)
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
deep learning, hematoxylin and eosin, large-cell neuroendocrine carcinoma, retinoblastoma 1, subtyping
in
Laboratory Investigation
volume
105
issue
9
article number
104192
publisher
Nature Publishing Group
external identifiers
  • scopus:105005843483
  • pmid:40345665
ISSN
0023-6837
DOI
10.1016/j.labinv.2025.104192
language
English
LU publication?
yes
id
add0ceb7-d559-4262-adc1-d816e166bb39
date added to LUP
2025-07-18 09:28:41
date last changed
2025-07-18 09:29:55
@article{add0ceb7-d559-4262-adc1-d816e166bb39,
  abstract     = {{<p>Molecular subtyping of pulmonary large-cell neuroendocrine carcinoma (LCNEC) based on retinoblastoma protein (pRb) expression may influence systemic treatment decisions. Current histomorphologic assessments of hematoxylin and eosin–stained tissue samples cannot reliably differentiate LCNEC molecular subtypes. This study explores the potential of deep learning (DL) to identify histologic patterns that distinguish these subtypes, by developing a custom convolutional neural network to predict the binary expression of pRb in small LCNEC tissue samples. Our model was trained, cross-validated, and tested on tissue microarray cores from 143 resection specimens and biopsies from 21 additional patients, with corresponding immunohistochemical pRb status. The best-performing DL model achieved a patient-wise balanced accuracy value of 0.75 and an area under the receiver operating characteristic curve value of 0.77 when tested on biopsies, significantly outperforming the hematoxylin and eosin–based subtype classification by lung pathologists. Explainable artificial intelligence techniques further highlighted coarse chromatin patterns and distinct nucleoli as distinguishing features for pRb retained status. Meanwhile, pRb lost cases were characterized by limited cytoplasm and morphologic similarities with small cell lung cancer. These findings suggest that DL analysis of small histopathology samples could ultimately replace immunohistochemical pRb testing. Such a development may assist in guiding chemotherapy decisions, particularly in metastatic cases.</p>}},
  author       = {{Trandafir, Teodora E. and Heijboer, Frank W.J. and Akram, Farhan and Derks, Jules L. and Li, Yunlei and Hillen, Lisa M. and Speel, Ernst Jan M. and Megyesfalvi, Zsolt and Dome, Balazs and Stubbs, Andrew P. and Dingemans, Anne Marie C. and von der Thüsen, Jan H.}},
  issn         = {{0023-6837}},
  keywords     = {{deep learning; hematoxylin and eosin; large-cell neuroendocrine carcinoma; retinoblastoma 1; subtyping}},
  language     = {{eng}},
  number       = {{9}},
  publisher    = {{Nature Publishing Group}},
  series       = {{Laboratory Investigation}},
  title        = {{Deep Learning–Based Retinoblastoma Protein Subtyping of Pulmonary Large-Cell Neuroendocrine Carcinoma on Small Hematoxylin and Eosin–Stained Specimens}},
  url          = {{http://dx.doi.org/10.1016/j.labinv.2025.104192}},
  doi          = {{10.1016/j.labinv.2025.104192}},
  volume       = {{105}},
  year         = {{2025}},
}