EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Stress testing deep learning models for prostate cancer detection on biopsies and surgical specimens

Brennan T Flannery iD, Howard M Sandler, Priti Lal, Michael D Feldman, Juan C Santa-Rosario, Tilak Pathak, Tuomas Mirtti, Xavier Farre, Rohann Correa, Susan Chafe, Amit Shah, Jason A Efstathiou und 7 weitere

The Journal of Pathology · 2025-02 · 265 (2) · 146-157

DOI: 10.1002/path.6373 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of Pathology, online erschienen am 11.12.2024. Verfasst von Brennan T Flannery und 18 weiteren Autor:innen. Beteiligte Einrichtung: Case Western Reserve University, Cleveland, OH, USA. Hauptthemen laut MEDLINE: Prostatic Neoplasms, Deep Learning, Prostatectomy. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: NCI NIH HHS, Kidney Precision Medicine Project, United States Department of Veterans Affairs Biomedical Laboratory Research and Development Service.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The presence, location, and extent of prostate cancer is assessed by pathologists using H&E‐stained tissue slides. Machine learning approaches can accomplish these tasks for both biopsies and radical prostatectomies. Deep learning approaches using convolutional neural networks (CNNs) have been shown to identify cancer in pathologic slides, some securing regulatory approval for clinical use. However, differences in sample processing can subtly alter the morphology between sample types, making it unclear whether deep learning algorithms will consistently work on both types of slide images. Our goal was to investigate whether morphological differences between sample types affected the performance of biopsy‐trained cancer detection CNN models when applied to radical prostatectomies and vice versa using multiple cohorts ( N = 1,000). Radical prostatectomies ( N = 100) and biopsies ( N = 50) were acquired from The University of Pennsylvania to train (80%) and validate (20%) a DenseNet CNN for biopsies (M B ), radical prostatectomies (M R ), and a combined dataset (M B+R ). On a tile level, M B and M R achieved F1 scores greater than 0.88 when applied to their own sample type but less than 0.65 when applied across sample types. On a whole‐slide level, models achieved significantly better performance on their own sample type compared to the alternative model ( p < 0.05) for all metrics. This was confirmed by external validation using digitized biopsy slide images from a clinical trial [NRG Radiation Therapy Oncology Group (RTOG)] (NRG/RTOG 0521, N = 750) via both qualitative and quantitative analyses ( p < 0.05). A comprehensive review of model outputs revealed morphologically driven decision making that adversely affected model performance. M B appeared to be challenged with the analysis of open gland structures, whereas M R appeared to be challenged with closed gland structures, indicating potential morphological variation between the training sets. These findings suggest that differences in morphology and heterogeneity necessitate the need for more tailored, sample‐specific (i.e. biopsy and surgical) machine learning models. © 2024 The Author(s). The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Prostatic NeoplasmsDeep LearningProstatectomyHumansMaleBiopsyNeural Networks, Computer

Stichwörter der Autor:innen

biopsyconvolutional neural networksdeep learninggeneralizabilityinterpretabilitymachine learningmorphologyprostate cancerradical prostatectomy

Autor:innen und Institutionen

  • Case Western Reserve University, Cleveland, OH, USA.
  • Cedars-Sinai Medical Center, Los Angeles, CA, USA.
  • University of Pennsylvania, Philadelphia, PA, USA.
  • Indiana University, Indianapolis, IN, USA.
  • CorePlus, Carolina, Puerto Rico.
  • Emory Winship Cancer Institute, Atlanta, GA, USA.

ORCID-Profile: Brennan T Flannery

Förderung

  • NCI NIH HHS
  • Kidney Precision Medicine Project
  • United States Department of Veterans Affairs Biomedical Laboratory Research and Development Service
  • VA Merit Review Award
  • NIBIB NIH HHS
  • Peer Reviewed Cancer Research Program
  • The DoD Breast Cancer Research Program Breakthrough Level 1 Award
  • Prostate Cancer Research Program

Zitieren

APA

Flannery, B. T., Sandler, H. M., Lal, P., Feldman, M. D., Santa-Rosario, J. C., Pathak, T., Mirtti, T., Farre, X., Correa, R., Chafe, S., Shah, A., Efstathiou, J. A., Hoffman, K., Hallman, M. A., Straza, M., Jordan, R., Pugh, S. L., Feng, F., & Madabhushi, A. (2025). Stress testing deep learning models for prostate cancer detection on biopsies and surgical specimens. The Journal of Pathology, 265(2), 146-157. https://doi.org/10.1002/path.6373

Vancouver

Flannery BT, Sandler HM, Lal P, Feldman MD, Santa-Rosario JC, Pathak T, et al. Stress testing deep learning models for prostate cancer detection on biopsies and surgical specimens. The Journal of Pathology. 2025;265(2):146-157. doi:10.1002/path.6373

BibTeX
@article{Flannery2025,
  author = {Flannery, Brennan T and Sandler, Howard M and Lal, Priti and Feldman, Michael D and Santa-Rosario, Juan C and Pathak, Tilak and Mirtti, Tuomas and Farre, Xavier and Correa, Rohann and Chafe, Susan and Shah, Amit and Efstathiou, Jason A and Hoffman, Karen and Hallman, Mark A and Straza, Michael and Jordan, Richard and Pugh, Stephanie L and Feng, Felix and Madabhushi, Anant},
  title = {Stress testing deep learning models for prostate cancer detection on biopsies and surgical specimens},
  journal = {The Journal of Pathology},
  year = {2025},
  volume = {265},
  number = {2},
  pages = {146--157},
  doi = {10.1002/path.6373}
}
Kostenlos starten →