Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Prevalence and characteristics of NTRK fusions in 25,946 patients with non-small cell lung cancer.

Lung cancer (Amsterdam, Netherlands)·2026
Same author

An analysis of succinate dehydrogenase B in pleural mesothelioma.

Discover oncology·2026
Same author

The brain network underlying social participation: a multimodal, data-driven investigation.

Brain imaging and behavior·2026
Same author

The continuing importance of the IARC's international remit in cancer research.

Journal of the National Cancer Institute. Monographs·2026
Same author

An Analysis of G3BP2 in Non-Small Cell Lung Cancer.

Cancers·2026
Same author

Neonatal social communication and single genes predict the variability of post-pubertal social behavior in a mouse model of paternal 15q11-13 duplication.

Research square·2026
Same journal

An automated end-to-end pipeline for the management, de-identification, and distribution of whole-slide images using DICOM: An institutional implementation.

Journal of pathology informatics·2026
Same journal

Automatic framework for PD-L1 expression evaluation in Latino patients with non-small cell lung cancer.

Journal of pathology informatics·2026
Same journal

Erratum to "Pathologists in Venice - Real world cases for an immersive training experience": Education, gaming, and show. <i>Journal of Pathology Informatics</i>, Volume 17, 2025, 100418.

Journal of pathology informatics·2026
Same journal

Erratum to PIRO: A web-based search platform for pathology reports, leveraging large language models to generate discrete searchable insights. <i>Journal of Pathology Informatics</i>, Volume 17, 2025, 100436.

Journal of pathology informatics·2026
Same journal

Erratum regarding missing Declaration of Competing Interest statements in previously published articles.

Journal of pathology informatics·2026
Same journal

An integrated AI pipeline for automated cytogenetic analysis of bone marrow karyograms in hematological malignancies: A Pix2Pix enhancement and deep learning detection approach.

Journal of pathology informatics·2026
See all related articles

Related Experiment Video

Updated: Jun 10, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.7K

Deep feature batch correction using ComBat for machine learning applications in computational pathology.

Pierre Murchan1,2, Pilib Ó Broin2,3, Anne-Marie Baird4

  • 1Department of Histopathology and Morbid Anatomy, Trinity Translational Medicine Institute, Trinity College Dublin, Dublin D08 W9RT, Ireland.

Journal of Pathology Informatics
|October 14, 2024
PubMed
Summary
This summary is machine-generated.

ComBat harmonization effectively reduces batch effects in digital pathology AI models, preventing misdiagnosis. This method ensures reliable performance estimates and preserves true histological signals for better generalizability.

Keywords:
Artificial intelligenceBatch effectsComputational pathologyHistopathologyThe Cancer Genome Atlas (TCGA)

More Related Videos

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
04:09

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

8.2K
DNA-barcode-based Multiplex Immunofluorescence Imaging to Analyze FFPE Specimens from Genetically Reprogrammed Murine Melanoma
09:52

DNA-barcode-based Multiplex Immunofluorescence Imaging to Analyze FFPE Specimens from Genetically Reprogrammed Murine Melanoma

Published on: June 6, 2025

67

Related Experiment Videos

Last Updated: Jun 10, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.7K
Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
04:09

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

8.2K
DNA-barcode-based Multiplex Immunofluorescence Imaging to Analyze FFPE Specimens from Genetically Reprogrammed Murine Melanoma
09:52

DNA-barcode-based Multiplex Immunofluorescence Imaging to Analyze FFPE Specimens from Genetically Reprogrammed Murine Melanoma

Published on: June 6, 2025

67

Area of Science:

  • Digital pathology
  • Artificial intelligence (AI)
  • Computational pathology

Background:

  • AI model development for digital pathology necessitates large, multi-source datasets.
  • AI models risk learning confounding site-specific features, leading to overestimated performance and poor generalizability.
  • Site-specific features can cause potential misdiagnosis in AI-driven pathology analyses.

Purpose of the Study:

  • To evaluate the effectiveness of ComBat harmonization in mitigating batch effects for AI models in digital pathology.
  • To assess the impact of ComBat harmonization on the prediction of tissue-source site (TSS), clinical attributes, and genetic features.
  • To ensure reliable performance estimates and improve the generalizability of AI models trained on multi-institutional whole-slide images (WSIs).

Main Methods:

  • Whole-slide images (WSIs) from The Cancer Genome Atlas (TCGA) colon (COAD) and stomach adenocarcinoma datasets were utilized.
  • Patch embeddings were generated using three feature extraction models and subsequently harmonized with ComBat.
  • Attention-based multiple instance learning models were trained to predict tissue-source site (TSS), clinical, and genetic attributes using raw, normalized, and harmonized embeddings.

Main Results:

  • ComBat harmonization significantly reduced tissue-source site (TSS) prediction accuracy (AUROC dropped to ~0.5), indicating successful batch effect mitigation.
  • Predictability of clinical attributes associated with TSS, such as race and treatment response, decreased post-harmonization.
  • Prediction of genetic features, like MSI status, remained robust after ComBat harmonization, preserving true histological signals.

Conclusions:

  • ComBat harmonization effectively reduces the risk of AI models learning confounding features in WSIs.
  • This approach ensures more reliable performance estimates for AI models in digital pathology.
  • ComBat harmonization shows promise for integrating large-scale digital pathology datasets, enhancing model generalizability and reliability.