Team Westwood Solution for MIDOG 2025 Challenge: An Ensemble-CNN-Based Approach For Mitosis Detection And Classification

TL;DR

The paper tackles generalizable mitosis detection and atypical mitosis classification across diverse pathology datasets. It presents a two-track approach: track 1 uses a two-stage pipeline with nnUNetV2 for candidate localization and a CNN ensemble refined by a random forest, while track 2 employs a CNN ensemble for atypical mitosis classification with TTA-enhanced inferences. Results show competitive performance in preliminary and final tests, with notable robustness gains from ensembling and data-driven tuning, and indicate data quality and expansion as key avenues for improvement. The work highlights practical strategies for improving generalization in digital pathology through staged detection, diverse CNN ensembles, and test-time augmentation.

Abstract

This abstract presents our solution (Team Westwood) for mitosis detection and atypical mitosis classification in the MItosis DOmain Generalization (MIDOG) 2025 challenge. For mitosis detection, we trained an nnUNetV2 for initial mitosis candidate screening with high sensitivity, followed by a random forest classifier ensembling predictions of three convolutional neural networks (CNNs): EfficientNet-b3, EfficientNet-b5, and EfficientNetV2-s. For the atypical mitosis classification, we trained another random forest classifier ensembling the predictions of three CNNs: EfficientNet-b3, EfficientNet-b5, and InceptionV3. On the preliminary test set, our solution achieved an F1 score of 0.7450 for track 1 mitosis detection, and a balanced accuracy of 0.8722 for track 2 atypical mitosis classification. On the final test set, our solution achieved an F1 score of 0.6972 for track 1 mitosis detection, and a balanced accuracy of 0.8242 for track 2 atypical mitosis classification.

Figures (2)

• Figure 1: Illustration of our mitosis detection pipeline for track 1 challenge. ROI: region of interest, TTA: test-time augmentation, CNN: convolution neural network.
• Figure 2: Illustration of track 2 atypical mitosis classification challenge.