Solutions for Mitotic Figure Detection and Atypical Classification in MIDOG 2025

2025-08-29

AI Summary

This work addresses mitosis detection and atypia classification in computational pathology. We propose a two-stage detection–classification framework coupled with a multi-model ensemble strategy. In the first stage, a high-precision object detector localizes candidate mitotic regions; in the second stage, heterogeneous deep classifiers—including ViT and ResNeXt—are ensembled to jointly optimize feature discriminability and domain generalization. Our method integrates candidate region re-scoring, uncertainty-weighted ensemble aggregation, and cross-domain data augmentation to substantially enhance robustness against morphological variability and staining heterogeneity. Evaluated on the MIDOG 2025 challenge, our approach achieves top-ranked performance in both mitosis detection and atypia classification tasks, demonstrating superior accuracy, stability, and clinical applicability.

Technology Category

• Computer Vision: Object Detection & Categorization
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Deep learning has driven significant advances in mitotic figure analysis within computational pathology. In this paper, we present our approach to the Mitosis Domain Generalization (MIDOG) 2025 Challenge, which consists of two distinct tasks, i.e., mitotic figure detection and atypical mitosis classification. For the mitotic figure detection task, we propose a two-stage detection-classification framework that first localizes candidate mitotic figures and subsequently refines the predictions using a dedicated classification module. For the atypical mitosis classification task, we employ an ensemble strategy that integrates predictions from multiple state-of-the-art deep learning architectures to improve robustness and accuracy. Extensive experiments demonstrate the effectiveness of our proposed methods across both tasks.

Problem

• Detecting mitotic figures in computational pathology images
• Classifying atypical mitoses using deep learning
• Addressing domain generalization in mitosis analysis

Innovation

• Two-stage detection-classification framework for localization
• Ensemble strategy integrating multiple deep learning architectures
• Dedicated classification module for refining predictions