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

TL;DR

This paper introduces a DenseNet-121-based method with stain-aware augmentation and hybrid loss to improve the robustness and domain generalization of atypical mitosis classification in histopathology images, addressing class imbalance and variability.

Contribution

It presents a novel framework combining stain-aware augmentation, hybrid loss, and DenseNet-121 for robust atypical mitosis classification across diverse imaging domains.

Findings

Achieved 85.0% balanced accuracy on test set

Demonstrated strong generalization across multiple domains

Improved sensitivity and AUROC compared to baseline methods

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.