Ensemble learning of pathology foundation models for precision oncology.

Ensemble learning of five pathology foundation models improves performance across multiple oncology tasks using 53,699 WSIs.

• Why it matters: Inconsistent performance and limited generalizability of existing models hinder reliable cancer diagnosis and treatment prediction, especially with limited data.
• What they did: Researchers developed ELF, which combines five pretrained pathology models into a unified slide-level representation, trained on a large dataset of 53,699 WSIs from 20 anatomical sites.
• The result: ELF outperforms individual and slide-level models in disease classification, biomarker detection, and therapy response prediction, supporting further use of ensemble learning in pathology-based oncology applications.