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

Genomic Characterization of Lung Cancer in Never-Smokers Using Deep Learning.

Modern pathology : an official journal of the United States and Canadian Academy of Pathology, Inc·2026
Same author

Potential and limitations of computed tomography images as predictors of the outcome of ischemic stroke events: a review.

Frontiers in stroke·2026
Same author

Microbiome analysis of 940 lung cancers in never-smokers reveals lack of clinically relevant associations.

Nature communications·2025
Same author

Uncovering the role of LINE-1 in the evolution of lung adenocarcinoma.

Nature·2025
Same author

Leveraging large language models for structured information extraction from pathology reports.

Journal of pathology informatics·2025
Same author

Monitoring sleep duration, timing, and continuity among US youth and adults in NHANES using actigraphy.

Sleep health·2025

Related Experiment Video

Updated: May 8, 2025

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

2.4K

Multiple Instance Learning for WSI: A comparative analysis of attention-based approaches.

Martim Afonso1, Praphulla M S Bhawsar2, Monjoy Saha2

  • 1Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, Lisbon 1049-001, Portugal.

Journal of Pathology Informatics
|December 24, 2024
PubMed
Summary
This summary is machine-generated.

Weakly supervised Multiple Instance Learning (MIL) models show promise in digital pathology for predicting cancer phenotypes and detecting TP53 mutations from whole slide images (WSI). These AI approaches can identify specific cellular morphologies associated with cancer at the tile level, aiding diagnostics.

Keywords:
Attention mechanismCancerMultiple instance learningTP53Whole slide image

More Related Videos

Measuring Attention and Visual Processing Speed by Model-based Analysis of Temporal-order Judgments
13:00

Measuring Attention and Visual Processing Speed by Model-based Analysis of Temporal-order Judgments

Published on: January 23, 2017

9.8K
Methods to Test Visual Attention Online
09:44

Methods to Test Visual Attention Online

Published on: February 19, 2015

11.7K

Related Experiment Videos

Last Updated: May 8, 2025

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

2.4K
Measuring Attention and Visual Processing Speed by Model-based Analysis of Temporal-order Judgments
13:00

Measuring Attention and Visual Processing Speed by Model-based Analysis of Temporal-order Judgments

Published on: January 23, 2017

9.8K
Methods to Test Visual Attention Online
09:44

Methods to Test Visual Attention Online

Published on: February 19, 2015

11.7K

Area of Science:

  • Digital Pathology
  • Computational Pathology
  • Artificial Intelligence in Medicine

Background:

  • Whole slide images (WSI) are crucial in digital pathology but pose challenges for AI analysis due to slide-level annotations.
  • Accurate cancer phenotyping and identification of tile-level cellular morphologies linked to mutations are significant hurdles.
  • Existing weakly supervised Multiple Instance Learning (MIL) methods require further investigation for complex pathology tasks.

Purpose of the Study:

  • To compare the efficacy of Attention MIL (AMIL) and Additive MIL (AdMIL) for tumor detection and TP53 mutation prediction.
  • To evaluate the capability of MIL architectures in identifying tile-level morphological features associated with cancer and mutations.
  • To explore the potential of MIL models in aiding pathologists by highlighting regions of interest (ROIs) in high-dimensional WSIs.

Main Methods:

  • Two weakly supervised MIL approaches, AMIL and AdMIL, were implemented and compared.
  • Models were trained and tested on datasets derived from Lung Squamous Cell Carcinoma (TCGA-LUSC) and Invasive Breast Carcinoma (TCGA-BRCA) WSIs.
  • Analyses included tumor detection at 5× magnification and TP53 mutation detection across 5×, 10×, and 20× magnifications.

Main Results:

  • Modified additive MIL achieved performance comparable to reference implementations (AUC 0.96) for tumor detection, slightly below AMIL (AUC 0.97).
  • TP53 mutation detection was more sensitive to higher magnification features, indicating the importance of cellular morphology resolution.
  • MIL architectures demonstrated an ability to identify distinct sensitivities to morphological features at different magnifications, highlighting relevant ROIs.

Conclusions:

  • Weakly supervised MIL models are effective for both tumor detection and TP53 mutation prediction in digital pathology.
  • Higher magnifications are crucial for detecting subtle morphological features related to specific molecular alterations like TP53 mutations.
  • The ROI identification capability of MIL models offers significant potential for integration into digital pathology workflows, enhancing diagnostic efficiency.