Technology Category
Computer Vision: Interpretability, Explainability, and TransparencyMachine Learning: Deep Neural Architectures and Foundation ModelsHumans and AI: Other Foundations of Human Computation & AI
Tissue Concepts v2: a Supervised Foundation Model for whole slide images
🤖 AI Summary
Current foundation models for whole-slide images (WSIs) suffer from heavy reliance on massive datasets and computational resources, alongside limited generalizability and interpretability. To address these challenges, this paper introduces Tissue Concepts v2 (TCv2), a supervised foundation model for histopathology. TCv2 employs an end-to-end multi-task learning framework trained efficiently using slide-level weak labels and incorporates a shared attention module to jointly encode tissue structural semantics while enhancing cross-task interpretability. Crucially, TCv2 is trained exclusively on publicly available WSI data and requires only modest computational resources. On multiple cancer subtype classification benchmarks, TCv2 consistently outperforms leading self-supervised methods, demonstrating superior generalization, high accuracy, and clinically meaningful interpretability. This work establishes a new paradigm for resource-efficient, interpretable, and scalable AI in digital pathology.
Technology Category
Application Category
📝 Abstract
Foundation models (FMs) are transforming the field of computational pathology by offering new approaches to analyzing histopathology images. Typically relying on weeks of training on large databases, the creation of FMs is a resource-intensive process in many ways. In this paper, we introduce the extension of our supervised foundation model, Tissue Concepts, to whole slide images, called Tissue Concepts v2 (TCv2), a supervised foundation model for whole slide images to address the issue above. TCv2 uses supervised, end-to-end multitask learning on slide-level labels. Training TCv2 uses a fraction of the training resources compared to self-supervised training. The presented model shows superior performance compared to SSL-trained models in cancer subtyping benchmarks and is fully trained on freely available data. Furthermore, a shared trained attention module provides an additional layer of explainability across different tasks.
Problem
Research questions and friction points this paper is trying to address.
Develops supervised foundation model for whole slide images
Reduces resource-intensive training for computational pathology
Improves performance and explainability in cancer subtyping
Innovation
Methods, ideas, or system contributions that make the work stand out.
Supervised foundation model for whole slide images
End-to-end multitask learning with slide labels
Shared attention module enhances explainability
🔎 Similar Papers
No similar papers found.
