Colorectal Cancer Histopathological Grading using Multi-Scale Federated Learning

2025-11-05

AI Summary

To address inter-observer variability in colorectal cancer (CRC) histopathological grading, privacy-driven constraints on multi-institutional data sharing, and the neglect of multi-scale diagnostic features in centralized models, this paper proposes a federated learning–based multi-scale pathological grading framework. We design a novel dual-stream ResNetRS50 architecture to jointly model nuclear-level fine-grained features and tissue-level contextual information. To mitigate model drift induced by client-side data heterogeneity, we integrate the FedProx optimization strategy. The framework supports modular AI pipeline deployment, ensuring both privacy preservation and generalizability. Evaluated on the CRC-HGD dataset, our method achieves an overall accuracy of 83.5%, significantly outperforming centralized baselines. It attains an 87.5% recall for high-grade (Type III) tumors and improves accuracy to 88.0% at 40× magnification.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Consent frameworks and practices on the web

Abstract

Colorectal cancer (CRC) grading is a critical prognostic factor but remains hampered by inter-observer variability and the privacy constraints of multi-institutional data sharing. While deep learning offers a path to automation, centralized training models conflict with data governance regulations and neglect the diagnostic importance of multi-scale analysis. In this work, we propose a scalable, privacy-preserving federated learning (FL) framework for CRC histopathological grading that integrates multi-scale feature learning within a distributed training paradigm. Our approach employs a dual-stream ResNetRS50 backbone to concurrently capture fine-grained nuclear detail and broader tissue-level context. This architecture is integrated into a robust FL system stabilized using FedProx to mitigate client drift across heterogeneous data distributions from multiple hospitals. Extensive evaluation on the CRC-HGD dataset demonstrates that our framework achieves an overall accuracy of 83.5%, outperforming a comparable centralized model (81.6%). Crucially, the system excels in identifying the most aggressive Grade III tumors with a high recall of 87.5%, a key clinical priority to prevent dangerous false negatives. Performance further improves with higher magnification, reaching 88.0% accuracy at 40x. These results validate that our federated multi-scale approach not only preserves patient privacy but also enhances model performance and generalization. The proposed modular pipeline, with built-in preprocessing, checkpointing, and error handling, establishes a foundational step toward deployable, privacy-aware clinical AI for digital pathology.

Problem

• Automating colorectal cancer grading while preserving patient privacy
• Addressing inter-observer variability in histopathological cancer assessment
• Integrating multi-scale analysis within distributed learning frameworks

Innovation

• Federated learning framework for privacy-preserving distributed training
• Dual-stream ResNetRS50 backbone for multi-scale feature learning
• FedProx stabilization to mitigate client drift across hospitals