Knowledge-driven Subspace Fusion and Gradient Coordination for Multi-modal Learning

TL;DR

This paper introduces a biologically interpretable multi-modal learning framework that effectively integrates histology images and genomics data for cancer diagnosis, leveraging subspace fusion and gradient coordination to improve performance.

Contribution

It proposes a novel knowledge-driven subspace fusion scheme and a gradient coordination strategy to enhance multi-modal learning in cancer analysis.

Findings

Outperforms state-of-the-art methods in glioma diagnosis, grading, and survival analysis.

Demonstrates robustness and interpretability in integrating histology and genomics data.

Effective in modeling complex tumor and microenvironment interactions.

Abstract

Multi-modal learning plays a crucial role in cancer diagnosis and prognosis. Current deep learning based multi-modal approaches are often limited by their abilities to model the complex correlations between genomics and histology data, addressing the intrinsic complexity of tumour ecosystem where both tumour and microenvironment contribute to malignancy. We propose a biologically interpretative and robust multi-modal learning framework to efficiently integrate histology images and genomics by decomposing the feature subspace of histology images and genomics, reflecting distinct tumour and microenvironment features. To enhance cross-modal interactions, we design a knowledge-driven subspace fusion scheme, consisting of a cross-modal deformable attention module and a gene-guided consistency strategy. Additionally, in pursuit of dynamically optimizing the subspace knowledge, we further propose a novel gradient coordination learning strategy. Extensive experiments demonstrate the effectiveness of the proposed method, outperforming state-of-the-art techniques in three downstream tasks of glioma diagnosis, tumour grading, and survival analysis. Our code is available at https://github.com/helenypzhang/Subspace-Multimodal-Learning.