EvoPS: Evolutionary Patch Selection in the Training Embedding Space of Whole Slide Images.

Abstract

In computational pathology, the gigapixel scale of Whole-Slide Images (WSIs) requires their decomposition into thousands of patches, resulting in high-dimensional embeddings that are computationally costly to process and often dominated by uninformative regions. Existing patch selection methods typically rely on heuristic sampling and do not explicitly address the trade-off between representation compactness and diagnostic accuracy. To address this gap, we propose EvoPS (Evolutionary Patch Selection), a novel framework that formulates patch selection within the training embedding space as a multi-objective optimization problem and leverages an evolutionary search to simultaneously minimize the number of selected patch embeddings and maximize the performance of a downstream similarity search task, generating a Pareto front of optimal trade-off solutions. By identifying a compact and diagnostically informative subset of training patches, EvoPS produces higher-quality training representations that reduce memory requirements and improve the signal-to-noise ratio of the training set. We validated our framework across four major cancer cohorts from The Cancer Genome Atlas (TCGA) using five histopathology foundation models. The results demonstrate that EvoPS can reduce the required number of training patches by over 90% while consistently maintaining or even improving the final classification F₁-score compared to a state-of-the-art patch selection method. The EvoPS framework provides a robust and principled method for creating efficient, accurate, and interpretable WSI representations, empowering users to select an optimal balance between computational cost and diagnostic performance.