| Abstract (ENG): |
The morphological classification of atypical mitotic figures (AMFs) is a
critical prognostic task in histopathology, but deep learning models often lack
generalization across diverse clinical settings. This study presents a robust and
reproducible pipeline for AMF detection. We compiled a large dataset from three
public sources and trained an ensemble of three ConvNeXt models using a 3-fold
cross-validation based bagging strategy. The pipeline achieved a balanced accuracy
of 89.18% on an internal hold-out set and demonstrated excellent generalization in
the MICCAI MIDOG2025 Challenge with a comparable 88.94% balanced accuracy,
securing rank #8 on the challenge leaderboard. The minimal performance drop
confirms the detection robustness against variations in tissue types, staining, and
scanners, providing a validated foundational tool for clinical AMF analysis. |
| Citation: |
Krauss, Sara and Spiess, Ellena and Hieber, Daniel and Schobel, Johannes and Kramer, Frank and Müller, Dominik
(2026)
Overcoming Domain Shift in Atypical Mitotic Figure Detection
with Deep Ensemble Learning.
In: (Proceedings of the) Medical Informatics Europe (MIE), May, 25-28, 2026, Genova, Italy, pp. 77-81, Paper / Research Article.
(Studies in Health Technology and Informatics; 336).
ISBN 9781643686615
|
