Do Pathology Foundation Models Encode Disease Progression? A Pseudotime Analysis of Visual Representations

Vision foundation models trained on discretely sampled images achieve strong performance on classification benchmarks, yet whether their representations encode the continuous processes underlying their training data remains unclear. This question is especially pertinent in computational pathology, where we posit that models whose latent representations implicitly capture continuous disease progression may better reflect underlying biology, support more robust generalization, and enable quantitative analyses of features associated with disease transitions. Using diffusion pseudotime, a method developed to infer developmental trajectories from single-cell transcriptomics, we probe whether foundation models organize disease states along coherent progression directions in representation space. Across four cancer progressions and six models, we find that all pathology-specific models recover trajectory orderings significantly exceeding null baselines, with vision-only models achieving the highest fidelities $(τ> 0.78$ on CRC-Serrated). Model rankings by trajectory fidelity on reference diseases strongly predict few-shot classification performance on held-out diseases ($ρ= 0.92$), and exploratory analysis shows cell-type composition varies smoothly along inferred trajectories in patterns consistent with known stromal remodeling. Together, these results demonstrate that vision foundation models can implicitly learn to represent continuous processes from independent static observations, and that trajectory fidelity provides a complementary measure of representation quality beyond downstream performance. While demonstrated in pathology, this framework could be applied to other domains where continuous processes are observed through static snapshots.

💡 Research Summary

This paper investigates whether vision foundation models trained on static histopathology images implicitly capture the continuous biological processes that underlie disease progression. While recent computational pathology (CPath) models achieve state‑of‑the‑art performance on discrete classification benchmarks, it remains unclear if their learned embeddings preserve the temporal ordering inherent to tumor evolution. To address this gap, the authors adapt Diffusion Pseudotime (DPT)—a trajectory‑inference method originally developed for single‑cell transcriptomics—to the high‑dimensional embeddings produced by six foundation models across four well‑characterized cancer progression cohorts.

Methodology

1. Models evaluated:
   • A natural‑image baseline (DINOv2).
   • Three vision‑only pathology models (UNI‑2, Virchow‑2, Prov‑GigaPath) that adapt the DINOv2 self‑distillation objective to H&E tiles.
   • Two vision‑language models (CONCH and MUSK) that incorporate text supervision.
     For all models, the