Finding Regions of Interest in Whole Slide Images Using Multiple Instance Learning
See where this sits in the topic map →Summary AI-generated
- TL;DR
- Whole Slide Images (WSIs) from high-resolution digital microscope scans are foundational to modern digital pathology, but analyzing them with AI poses unique challenges.
- Problem
- Pathology labeling and molecular data, such as oncogene mutations from The Cancer Genome Atlas (TCGA), are typically recorded at the whole slide level rather than at the individual tile level. This creates a dual challenge: predicting overall cancer phenotypes while identifying the specific cellular morphologies associated with them.
- Method
- To address these challenges, a weakly supervised Multiple Instance Learning (MIL) approach was explored for Invasive Breast Carcinoma (TCGA-BRCA) and Lung Squamous Cell Carcinoma (TCGA-LUSC). The study investigated tumor detection at low magnification levels and TP53 mutations across various levels.
- Results
- A novel additive implementation of MIL matched the performance of a reference implementation (AUC 0.96) and was only slightly outperformed by Attention MIL (AUC 0.97).
- Contributions
- Not specified in the abstract.
- Limitations
- Not specified in the abstract.
- Takeaways
- Different AI architectures identified distinct sensitivities to morphological features via Regions of Interest (RoIs) at various amplification levels. Notably, TP53 mutation detection was most sensitive to features at higher magnifications where cellular morphology is resolved.
- Applications
- Tumor detection and the identification of molecular mutations, such as TP53, in digital pathology workflows.
- Topics
- Whole Slide Images; Multiple Instance Learning; Digital Pathology; Cancer Detection; Molecular Pathology
- For industry
- Healthcare and medical diagnostics
- Why it matters
- Advances AI-based digital pathology methods by bridging slide-level labels with tile-level morphological features, aiding molecular pathologists in identifying relevant regions of interest.
Abstract
Whole Slide Images (WSI), obtained by high-resolution digital scanning of microscope slides at multiple scales, are the cornerstone of modern Digital Pathology. However, they represent a particular challenge to AI-based/AI-mediated analysis because pathology labeling is typically done at slide-level, instead of tile-level. It is not just that medical diagnostics is recorded at the specimen level, the detection of oncogene mutation is also experimentally obtained, and recorded by initiatives like The Cancer Genome Atlas (TCGA), at the slide level. This configures a dual challenge: a) accurately predicting the overall cancer phenotype and b) finding out what cellular morphologies are associated with it at the tile level. To address these challenges, a weakly supervised Multiple Instance Learning (MIL) approach was explored for two prevalent cancer types, Invasive Breast Carcinoma (TCGA-BRCA) and Lung Squamous Cell Carcinoma (TCGA-LUSC). This approach was explored for tumor detection at low magnification levels and TP53 mutations at various levels. Our results show that a novel additive implementation of MIL matched the performance of reference implementation (AUC 0.96), and was only slightly outperformed by Attention MIL (AUC 0.97). More interestingly from the perspective of the molecular pathologist, these different AI architectures identify distinct sensitivities to morphological features (through the detection of Regions of Interest, RoI) at different amplification levels. Tellingly, TP53 mutation was most sensitive to features at the higher applications where cellular morphology is resolved.