Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

MeCaMIL: Causality-Aware Multiple Instance Learning for Fair and Interpretable Whole Slide Image Diagnosis.

Yiran Song, Yikai Zhang, Shuang Zhou

IEEE Transactions on Medical Imaging

|July 13, 2026

Summary

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

High-resolution mri guided whole mouse brain neuronal cell type atlas using deep learning.

Communications biology·2026

Same author

CXR-LT 2026 Challenge: Multi-Center Long-Tailed and Zero Shot Chest X-ray Classification.

ArXiv·2026

Same author

Deep learning enabled decision support systems in epilepsy surgery: a scoping review.

npj health systems·2026

Same author

Two-Stage Decoupling Framework for Variable-Length Glaucoma Prognosis.

Learning with longitudinal medical images and data : first International Workshop, LMID 2025, held in conjunction with MICCAI 2025, Daejeon, South Korea, September 27, 2025, Proceedings. International Workshop on Learning with Longitudi...·2026

Same author

OVERVIEW OF THE CXR-LT 2026 CHALLENGE: MULTI-CENTER LONG-TAILED AND ZERO SHOT CHEST X-RAY CLASSIFICATION.

Proceedings. IEEE International Symposium on Biomedical Imaging·2026

Same author

Detecting early brain susceptibility changes before demyelination in cuprizone mouse model using quantitative susceptibility mapping (QSM).

NeuroImage·2026

MeCaMIL is a new AI framework for pathology images that uses causal graphs to improve fairness and interpretability. It addresses limitations in current methods by integrating patient demographics, reducing bias, and enhancing diagnostic accuracy.

Area of Science:

Computational pathology
Artificial intelligence in medicine
Causal inference

Background:

Multiple instance learning (MIL) is standard for whole slide image (WSI) analysis but lacks causal interpretability and fails to integrate patient demographics, raising fairness concerns.
Existing MIL methods can perpetuate health disparities due to algorithmic bias, hindering clinical translation.

Purpose of the Study:

To introduce MeCaMIL, a causality-aware MIL framework designed to explicitly model demographic confounders using structured causal graphs.
To disentangle disease-relevant signals from spurious demographic correlations using principled causal inference.

Main Methods:

MeCaMIL integrates patient demographics (age, gender, race) within a causal graph structure, unlike prior methods that treat them as auxiliary features.

Related Experiment Videos

Employs causal inference with collider structures to address confounding variables.

Evaluated on three benchmarks: CAMELYON16, TCGA-Lung, and TCGA-Multi.

Main Results:

MeCaMIL achieves state-of-the-art performance across benchmarks (e.g., CAMELYON16: ACC/AUC/F1: 0.939/0.983/0.946).
Demonstrates superior fairness, with over 65% average relative reduction in demographic disparity variance.
Generalizes to survival prediction (mean C-index: 0.653) and shows the causal graph is essential for performance and fairness.

Conclusions:

MeCaMIL provides a principled framework for fair, causally interpretable, and clinically actionable AI in digital pathology.
Structural causal modeling offers superior interpretability compared to post-hoc attention visualization.
The framework has the potential to reduce health disparities in diverse patient populations.