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

Large language models enable prognostic stratification of cancer patients using real-world clinical notes.

PLOS digital health·2026
Same author

Single nucleus RNA profiling reveals potential therapeutic vulnerabilities in sinonasal carcinomas.

NPJ precision oncology·2026
Same author

Towards robust foundation models for digital pathology.

Nature communications·2026
Same author

Beyond attention heatmaps: How to get better explanations for multiple instance learning models in histopathology.

Medical image analysis·2026
Same author

Real-world data on the clinical impact of molecular tumor boards in high- and low-grade serous ovarian cancer.

Frontiers in oncology·2026
Same author

Modeling attention and binding in the brain through bidirectional recurrent gating.

Nature communications·2026

Related Experiment Video

Updated: Dec 24, 2025

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
05:33

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Published on: July 11, 2025

681

Resolving challenges in deep learning-based analyses of histopathological images using explanation methods.

Miriam Hägele1, Philipp Seegerer1, Sebastian Lapuschkin2

  • 1TU Berlin, Machine Learning Group, Berlin, 10587, Germany.

Scientific Reports
|April 15, 2020
PubMed
Summary

Deep learning in digital pathology benefits from explanation methods. Heatmaps reveal and mitigate biases in histopathology images, improving diagnostic accuracy and model generalization.

More Related Videos

Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps
08:59

Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps

Published on: October 28, 2018

7.5K
Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research
05:22

Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research

Published on: June 21, 2024

744

Related Experiment Videos

Last Updated: Dec 24, 2025

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
05:33

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Published on: July 11, 2025

681
Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps
08:59

Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps

Published on: October 28, 2018

7.5K
Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research
05:22

Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research

Published on: June 21, 2024

744

Area of Science:

  • Digital Pathology
  • Artificial Intelligence
  • Medical Image Analysis

Background:

  • Deep learning models achieve high prediction accuracy in digital pathology.
  • The medical field requires explainable AI (XAI) for trust and insight beyond quantitative metrics.
  • Current deep learning analyses in histopathology often overlook inherent data biases.

Purpose of the Study:

  • To investigate how explanation methods, specifically heatmaps, can address challenges in deep learning-based digital histopathology.
  • To identify and analyze dataset, class-correlated, and sampling biases in histopathological image data.
  • To demonstrate the utility of pixel-wise heatmaps over patch-level evaluation for bias detection and mitigation.

Main Methods:

  • Utilized heatmaps generated by explanation methods to analyze deep learning models.
  • Investigated three types of biases: dataset-wide, class-correlated, and sampling biases.
  • Performed binary classification of tumour tissue in H&E-stained histopathology images.

Main Results:

  • Heatmaps effectively detect and help remove hidden biases in histopathological data.
  • Pixel-wise heatmaps provide more precise diagnostic insights compared to patch-level analysis.
  • Reducing labelling bias led to a 5% improvement in the area under the receiver operating characteristic (ROC) curve.

Conclusions:

  • Explanation techniques, particularly heatmaps, are crucial for developing and deploying reliable AI in digital pathology.
  • Heatmaps enhance model generalization by identifying and correcting biases.
  • Pixel-wise heatmaps serve as a versatile diagnostic tool for improving deep learning applications in histopathology.