Technology Category
Machine Learning: Transfer, Domain Adaptation, Multi-Task LearningComputer Vision: Object Detection & CategorizationSearch and Optimization: Learning to Search
Teacher-Student Model for Detecting and Classifying Mitosis in the MIDOG 2025 Challenge
🤖 AI Summary
Mitotic detection in histopathological images faces two major challenges: poor cross-domain generalization (due to variations across organs, species, and staining protocols) and severe class imbalance (mitoses are sparse). To address these, we propose a multi-task framework integrating domain generalization and teacher–student pseudo-supervision. Specifically, we design a domain generalization module leveraging contrastive representation learning and adversarial domain alignment to enhance cross-domain robustness. A teacher model generates multi-class pixel-level pseudo-masks to jointly optimize segmentation and classification in the student network. Additionally, a multi-scale CNN classifier is introduced to improve fine-grained discriminative capability. On the preliminary test set, our method achieves a detection F1-score of 0.7660 and a balanced classification accuracy of 84.14%, demonstrating significant improvements in small-object detection consistency and cross-domain stability.
Technology Category
Application Category
📝 Abstract
Counting mitotic figures is time-intensive for pathologists and leads to inter-observer variability. Artificial intelligence (AI) promises a solution by automatically detecting mitotic figures while maintaining decision consistency. However, AI tools are susceptible to domain shift, where a significant drop in performance can occur due to differences in the training and testing sets, including morphological diversity between organs, species, and variations in staining protocols. Furthermore, the number of mitoses is much less than the count of normal nuclei, which introduces severely imbalanced data for the detection task. In this work, we formulate mitosis detection as a pixel-level segmentation and propose a teacher-student model that simultaneously addresses mitosis detection (Track 1) and atypical mitosis classification (Track 2). Our method is based on a UNet segmentation backbone that integrates domain generalization modules, namely contrastive representation learning and domain-adversarial training. A teacher-student strategy is employed to generate pixel-level pseudo-masks not only for annotated mitoses and hard negatives but also for normal nuclei, thereby enhancing feature discrimination and improving robustness against domain shift. For the classification task, we introduce a multi-scale CNN classifier that leverages feature maps from the segmentation model within a multi-task learning paradigm. On the preliminary test set, the algorithm achieved an F1 score of 0.7660 in Track 1 and balanced accuracy of 0.8414 in Track 2, demonstrating the effectiveness of integrating segmentation-based detection and classification into a unified framework for robust mitosis analysis.
Problem
Research questions and friction points this paper is trying to address.
Addressing domain shift in mitosis detection across diverse tissue types
Handling severe data imbalance between mitotic and normal nuclei
Integrating segmentation and classification for robust mitosis analysis
Innovation
Methods, ideas, or system contributions that make the work stand out.
Teacher-student model with UNet segmentation backbone
Contrastive learning and adversarial training for domain generalization
Multi-scale classifier within multi-task learning framework
🔎 Similar Papers
2024-08-29Medical Imaging 2025: Digital and Computational PathologyCitations: 1
