MitoDetect++: A Domain-Robust Pipeline for Mitosis Detection and Atypical Subtyping

2025-08-28

AI Summary

Accurate detection and fine-grained classification of atypical mitoses in histopathological images remain critical yet challenging tasks for cancer diagnosis and grading. Method: We propose MitoDetect++, a unified deep learning framework comprising: (1) a U-Net architecture enhanced with an EfficientNetV2-L backbone and attention mechanisms for precise mitotic localization; (2) a Virchow2 vision transformer fine-tuned via Low-Rank Adaptation (LoRA) for binary (normal vs. atypical) mitosis classification; and (3) group-aware hierarchical cross-validation and test-time augmentation (TTA) to improve cross-domain robustness. Contribution/Results: Evaluated on a multi-center dataset, MitoDetect++ achieves a balanced accuracy of 0.892—surpassing state-of-the-art methods—and demonstrates strong clinical applicability and generalizability by winning both tasks in the MIDOG 2025 challenge.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Automated detection and classification of mitotic figures especially distinguishing atypical from normal remain critical challenges in computational pathology. We present MitoDetect++, a unified deep learning pipeline designed for the MIDOG 2025 challenge, addressing both mitosis detection and atypical mitosis classification. For detection (Track 1), we employ a U-Net-based encoder-decoder architecture with EfficientNetV2-L as the backbone, enhanced with attention modules, and trained via combined segmentation losses. For classification (Track 2), we leverage the Virchow2 vision transformer, fine-tuned efficiently using Low-Rank Adaptation (LoRA) to minimize resource consumption. To improve generalization and mitigate domain shifts, we integrate strong augmentations, focal loss, and group-aware stratified 5-fold cross-validation. At inference, we deploy test-time augmentation (TTA) to boost robustness. Our method achieves a balanced accuracy of 0.892 across validation domains, highlighting its clinical applicability and scalability across tasks.

Problem

• Automated detection and classification of mitotic figures
• Distinguishing atypical from normal mitosis in pathology
• Addressing domain shifts and improving generalization in detection

Innovation

• U-Net-based encoder-decoder with EfficientNetV2-L backbone
• Virchow2 vision transformer fine-tuned via LoRA
• Test-time augmentation and focal loss for robustness