Joint modelling histology and molecular markers for cancer classification DOI Creative Commons
Xiaofei Wang, Hanyu Liu, Yupei Zhang

et al.

Medical Image Analysis, Journal Year: 2025, Volume and Issue: 102, P. 103505 - 103505

Published: Feb. 22, 2025

Cancers are characterized by remarkable heterogeneity and diverse prognosis. Accurate cancer classification is essential for patient stratification clinical decision-making. Although digital pathology has been advancing diagnosis prognosis, the paradigm in shifted from purely relying on histology features to incorporating molecular markers. There an urgent need methods meet needs of new paradigm. We introduce a novel approach jointly predict markers model their interactions classification. Firstly, mitigate challenge cross-magnification information propagation, we propose multi-scale disentangling module, enabling extraction high-magnification (cellular-level) low-magnification (tissue-level) whole slide images. Further, based features, attention-based hierarchical multi-task multi-instance learning framework simultaneously Moreover, co-occurrence probability-based label correlation graph network Lastly, design cross-modal interaction module with dynamic confidence constrain loss gradient modulation strategy, Our experiments demonstrate that our method outperforms other state-of-the-art classifying glioma, promises promote precise oncology potential advance biomedical research applications. The code available at github.

Language: Английский

Joint modelling histology and molecular markers for cancer classification DOI Creative Commons
Xiaofei Wang, Hanyu Liu, Yupei Zhang

et al.

Medical Image Analysis, Journal Year: 2025, Volume and Issue: 102, P. 103505 - 103505

Published: Feb. 22, 2025

Cancers are characterized by remarkable heterogeneity and diverse prognosis. Accurate cancer classification is essential for patient stratification clinical decision-making. Although digital pathology has been advancing diagnosis prognosis, the paradigm in shifted from purely relying on histology features to incorporating molecular markers. There an urgent need methods meet needs of new paradigm. We introduce a novel approach jointly predict markers model their interactions classification. Firstly, mitigate challenge cross-magnification information propagation, we propose multi-scale disentangling module, enabling extraction high-magnification (cellular-level) low-magnification (tissue-level) whole slide images. Further, based features, attention-based hierarchical multi-task multi-instance learning framework simultaneously Moreover, co-occurrence probability-based label correlation graph network Lastly, design cross-modal interaction module with dynamic confidence constrain loss gradient modulation strategy, Our experiments demonstrate that our method outperforms other state-of-the-art classifying glioma, promises promote precise oncology potential advance biomedical research applications. The code available at github.

Language: Английский

Citations

0