EvoPS: Evolutionary Patch Selection for Whole Slide Image Analysis in Computational Pathology

The introduction of EvoPS (Evolutionary Patch Selection) marks a significant advancement in computational pathology, particularly in the analysis of Whole-Slide Images (WSIs). Traditional methods often relied on random sampling or simple clustering, which could dilute critical diagnostic signals. EvoPS addresses this by framing patch selection as a multi-objective optimization problem, allowing for a simultaneous reduction in the number of selected patches and an enhancement in performance for similarity search tasks. Validation across four major cancer cohorts from The Cancer Genome Atlas (TCGA) demonstrated that EvoPS can reduce the number of required training patch embeddings by over 90%, while consistently maintaining or even improving the final classification F1-score. This capability is crucial for researchers dealing with large-scale data, as it not only streamlines the analysis process but also enhances the potential for accurate diagnostics, thereby contributing to improved ou…