Robust Atypical Mitosis Classification with DenseNet121: Stain-Aware Augmentation and Hybrid Loss for Domain Generalization

2025-10-26

AI Summary

Atypical mitosis is a critical histopathological marker for assessing tumor aggressiveness, yet its identification is hindered by inter-domain variations (e.g., staining and scanning equipment differences) and severe class imbalance. To address these challenges, we propose a domain-generalizable robust classification framework. First, we introduce stain-aware augmentation—combining Macenko stain normalization with joint geometric and intensity augmentation—to mitigate domain shift. Second, we design a weighted sampling strategy coupled with a hybrid loss function—integrating weighted binary cross-entropy and focal loss—to jointly tackle label imbalance. The model is built upon DenseNet121 and trained end-to-end. On the MIDOG 2025 test set, our method achieves 85.0% balanced accuracy, 0.927 AUROC, 89.2% sensitivity, and 80.9% specificity—substantially outperforming baseline methods. These results demonstrate strong generalization across heterogeneous imaging devices and underscore the clinical applicability of our approach.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Search and Optimization: Distributed Search

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Atypical mitotic figures are important biomarkers of tumor aggressiveness in histopathology, yet reliable recognition remains challenging due to severe class imbalance and variability across imaging domains. We present a DenseNet-121-based framework tailored for atypical mitosis classification in the MIDOG 2025 (Track 2) setting. Our method integrates stain-aware augmentation (Macenko), geometric and intensity transformations, and imbalance-aware learning via weighted sampling with a hybrid objective combining class-weighted binary cross-entropy and focal loss. Trained end-to-end with AdamW and evaluated across multiple independent domains, the model demonstrates strong generalization under scanner and staining shifts, achieving balanced accuracy 85.0%, AUROC 0.927, sensitivity 89.2%, and specificity 80.9% on the official test set. These results indicate that combining DenseNet-121 with stain-aware augmentation and imbalance-adaptive objectives yields a robust, domain-generalizable framework for atypical mitosis classification suitable for real-world computational pathology workflows.

Problem

• Classifying atypical mitotic figures despite severe class imbalance
• Addressing domain variability across different histopathology scanners
• Improving generalization under staining and scanner shifts

Innovation

• DenseNet-121 framework for atypical mitosis classification
• Stain-aware augmentation with geometric transformations
• Hybrid loss combining weighted cross-entropy and focal loss