A multi-task neural network for atypical mitosis recognition under domain shift

2025-08-28

AI Summary

Accurate identification of atypical mitoses is critical for assessing tumor aggressiveness, yet existing deep learning models suffer from poor generalization under domain shift. To address this, we propose a multi-task learning framework that jointly optimizes the primary classification task with semantically related auxiliary tasks—including mitotic localization, morphological segmentation, and background suppression—to explicitly guide the model toward discriminative cellular structures while mitigating domain-specific background variations. Our approach significantly enhances cross-domain robustness: it achieves consistent performance gains across three heterogeneous datasets—MIDOG 2025, Ami-Br, and MIDOG25—with an average F1-score improvement of 8.3% over single-task baselines. This demonstrates superior generalizability and strong potential for clinical deployment in diverse histopathological settings.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Multi-modal Vision
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recognizing atypical mitotic figures in histopathology images allows physicians to correctly assess tumor aggressiveness. Although machine learning models could be exploited for automatically performing such a task, under domain shift these models suffer from significative performance drops. In this work, an approach based on multi-task learning is proposed for addressing this problem. By exploiting auxiliary tasks, correlated to the main classification task, the proposed approach, submitted to the track 2 of the MItosis DOmain Generalization (MIDOG) challenge, aims to aid the model to focus only on the object to classify, ignoring the domain varying background of the image. The proposed approach shows promising performance in a preliminary evaluation conducted on three distinct datasets, i.e., the MIDOG 2025 Atypical Training Set, the Ami-Br dataset, as well as the preliminary test set of the MIDOG25 challenge.

Problem

• Recognizing atypical mitotic figures in histopathology images
• Addressing performance drops under domain shift conditions
• Ignoring domain-varying background through multi-task learning

Innovation

• Multi-task neural network for domain shift
• Auxiliary tasks to ignore varying backgrounds
• Improves atypical mitosis recognition performance