Search papers, labs, and topics across Lattice.
This paper introduces \textsc{$\mathrm{\Phi}$-Omni}, a novel framework for omni-modal self-supervised learning in computational pathology that leverages Partial Information Decomposition (PID) theory to disentangle synergistic information from histology, genomics, and clinical reports. By utilizing a Synergistic Information Bottleneck (SIB) and the $\mathrm{\Phi}\text{ID}$ objective, the model effectively minimizes redundancy while maximizing the extraction of unique diagnostic signals. The framework outperforms existing supervised and self-supervised learning baselines in few-shot performance across multiple independent datasets, highlighting its potential for enhancing diagnostic accuracy in pathology.
Synergistic Information Disentanglement can significantly boost few-shot learning performance in computational pathology by preserving unique diagnostic signals across modalities.
In computational pathology (CPath), developing omni-modal self-supervised learning (SSL) models that integrate histology, genomics, and clinical reports enables transferable representation learning for whole slide images (WSIs). Existing approaches implicitly force heterogeneous modalities into a uniform latent space by contrastive alignment, causing modality collapse where unique, synergistic diagnostic signals (termed as $\mathrm{\Phi}$) are discarded in favor of trivial redundancy. We hypothesize that the strongest task-agnostic SSL training signal stems from distilling the synergistic interactions over merely aligning shared redundancy. To this end, we introduce \textsc{$\mathrm{\Phi}$-Omni}, a synergistic information disentanglement framework grounded in Partial Information Decomposition (PID) theory for slide representation learning. Unlike standard contrastive approaches, \textsc{$\mathrm{\Phi}$-Omni} employs a Synergistic Information Bottleneck (SIB) regulated by the proposed $\mathrm{\Phi}\text{ID}$ objective, which explicitly suppresses marginal redundancy while maximizing irreducible synergy, thereby distilling high-order cross-modal interactions. Following pretraining on breast ($n$=1031) and lung ($n$=919) cohorts, \textsc{$\mathrm{\Phi}$-Omni} demonstrates superior few-shot performance across five independent external datasets spanning eight tasks compared to supervised and SSL baselines. Source code is available here.
