Is Synthetic Image Augmentation Useful for Imbalanced Classification Problems? Case-Study on the MIDOG2025 Atypical Cell Detection Competition

2025-08-30

AI Summary

This work addresses the detection of rare, clinically critical asymmetric mitotic cells in histopathological images—challenged by severe class imbalance and cross-center domain shift. We systematically evaluate synthetic data augmentation and pretraining strategies, employing ConvNeXt-Small and Lunit’s self-supervised ViT as backbones, with five-fold cross-validation and joint training on real and synthetic samples. Results show that domain-specific pretraining enhances robustness, while ImageNet pretraining achieves higher performance ceilings; naive synthetic augmentation yields limited and inconsistent gains. The two models achieve a mean AUROC of 95%, with ConvNeXt attaining the highest AUROC (95.4%) on the held-out test set and Lunit demonstrating superior balanced accuracy. Our study uncovers the trade-off between pretraining source and data augmentation efficacy for rare pathological object recognition, offering a reproducible methodological framework for few-shot classification in medical imaging.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Learning to Search

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The MIDOG 2025 challenge extends prior work on mitotic figure detection by introducing a new Track 2 on atypical mitosis classification. This task aims to distinguish normal from atypical mitotic figures in histopathology images, a clinically relevant but highly imbalanced and cross-domain problem. We investigated two complementary backbones: (i) ConvNeXt-Small, pretrained on ImageNet, and (ii) a histopathology-specific ViT from Lunit trained via self-supervision. To address the strong prevalence imbalance (9408 normal vs. 1741 atypical), we synthesized additional atypical examples to approximate class balance and compared models trained with real-only vs. real+synthetic data. Using five-fold cross-validation, both backbones reached strong performance (mean AUROC approximately 95 percent), with ConvNeXt achieving slightly higher peaks while Lunit exhibited greater fold-to-fold stability. Synthetic balancing, however, did not lead to consistent improvements. On the organizers' preliminary hidden test set, explicitly designed as an out-of-distribution debug subset, ConvNeXt attained the highest AUROC (95.4 percent), whereas Lunit remained competitive on balanced accuracy. These findings suggest that both ImageNet and domain-pretrained backbones are viable for atypical mitosis classification, with domain-pretraining conferring robustness and ImageNet pretraining reaching higher peaks, while naive synthetic balancing has limited benefit. Full hidden test set results will be reported upon challenge completion.

Problem

• Distinguishing normal versus atypical mitotic figures in histopathology images
• Addressing highly imbalanced classification with cross-domain challenges
• Evaluating synthetic image augmentation for class imbalance mitigation

Innovation

• Used ConvNeXt and domain-specific ViT backbones
• Synthesized atypical examples for class balance
• Compared real-only versus real plus synthetic data