Related Experiment Video
Updated: Aug 23, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.9K
Cross-scale Attention Guided Multi-instance Learning for Crohn's Disease Diagnosis with Pathological Images
Ruining Deng1, Can Cui1, Lucas W Remedios1
1Vanderbilt University, Nashville TN 37215, USA.
Multiscale Multimodal Medical Imaging : Third International Workshop, MMMI 2022, Held in Conjunction with MICCAI 2022, Singapore, September 22, 2022, Proceedings
|November 4, 2022
Summary
This study introduces a cross-scale attention mechanism for analyzing Whole Slide Images in Crohn's Disease detection. The novel approach improves multi-instance learning by integrating multi-scale features, achieving high diagnostic accuracy.
Area of Science:
- Computational pathology
- Digital pathology
- Machine learning in medicine
Background:
- Multi-instance learning (MIL) is crucial for Whole Slide Image (WSI) analysis due to limited annotations.
- Existing MIL methods often ignore the multi-scale nature of WSIs, unlike human pathologists.
- Pathologists integrate information across different magnifications for accurate diagnosis.
Purpose of the Study:
- To develop a novel cross-scale attention mechanism for MIL in WSI analysis.
- To explicitly aggregate inter-scale interactions within a single MIL network.
- To improve the computer-aided interpretation of pathological WSIs for Crohn's Disease detection.
Main Methods:
- Proposed a novel cross-scale attention mechanism to aggregate features from different resolutions.
- Integrated multi-scale interactions into a single MIL network.
- Generated differential multi-scale attention visualizations for explainable lesion localization.
Main Results:
- Achieved a superior Area under the Curve (AUC) score of 0.8924 on Crohn's Disease detection.
- Demonstrated superior performance compared to baseline MIL models.
- Trained on approximately 250,000 H&E-stained Ascending Colon patches from 50 samples.
Conclusions:
- The proposed cross-scale attention mechanism effectively integrates multi-scale information for WSI analysis.
- This approach enhances diagnostic accuracy for Crohn's Disease.
- The method provides explainable visualizations for lesion identification.
