Colorectal Cancer Histopathological Grading using Multi-Scale Federated Learning

TL;DR

The paper addresses the challenge of colorectal cancer histopathological grading under privacy constraints by proposing a privacy-preserving federated learning framework that integrates multi-scale feature learning via a dual-stream ResNetRS50. The approach is evaluated on the CRC-HGD dataset, achieving 83.5% overall accuracy and 87.5% recall for Grade III, with magnification-dependent gains up to 88.0% at 40x. Compared to centralized training, federated learning offers competitive performance while preserving patient privacy, with strong generalization across heterogeneous hospital data. The work contributes a modular, scalable pipeline with automated checkpointing and error handling, advancing deployable privacy-aware AI for digital pathology.

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.

Figures (7)

• Figure 1: System architecture for multi-scale colorectal cancer grading using dual-stream processing.
• Figure 2: Representative histopathology samples showing morphological variations across tumor grades and magnifications
• Figure 3: Confusion matrix showing the classification performance of the model on the test set for tumor Grades I, II, and III. The diagonal elements (highlighted) represent the correct predictions
• Figure 4: Training progression comparison: Federated Learning versus centralized training across validation rounds
• Figure 5: Grade-specific performance metrics demonstrating balanced classification across all tumor grades