Computational Pathology Foundations era
During 1990–1996, representative figures in computational pathology advanced boundary delineation in histology with active-contour models from Kass, Witkin, and Terzopoulos, enabling automated morphometric measurements. Haralick and colleagues' texture descriptors based on gray-level co-occurrence matrices provided robust quantifications of tissue patterns, underpinning data-driven biomarkers. Thresholding and color-segmentation techniques, including Otsu’s method and the practical frameworks popularized in Gonzalez and Woods, offered repeatable segmentation for stained slides and high-resolution scans. Fuzzy-clustering approaches such as fuzzy c-means from Bezdek and the fuzzy-connectedness ideas advanced by Udupa and colleagues contributed robust tissue classification and paved the way for automated analysis pipelines.
Open-Source Digital Pathology era
Michael D. Abràmoff [1], affiliated with Radboud University Nijmegen [2] and the University of Iowa [3] during this era, emerges as a leading proponent of open-source digital pathology. His notable contribution, evidenced by the 2004 paper Image processing with ImageJ [4], was to promote accessible, extensible image-analysis workflows that empowered pathologists to perform quantitative analyses with a shared, interoperable tool. Through active work at Radboud University Nijmegen [2] and the University of Iowa [3], Abràmoff helped catalyze community-driven plugin ecosystems and benchmarks that underpinned reproducible biomarker extraction in open platforms [1]. These efforts, led by Abràmoff [1], collectively advanced the era by reducing methodological heterogeneity and enabling scalable, cross-institutional digital pathology workflows. Foundation-Model Pathology era
In the Foundation-Model Pathology era (2018–2024), key voices include Anant Madabhushi, Geert Litjens, and Katharina N. Kather, whose work helped define transferable, representation-focused pathology AI. Madabhushi and colleagues pioneered multi-instance learning and patch-based deep architectures for whole-slide image analysis, enabling weak supervision and localization of cancer regions across large datasets. Litjens contributed to foundational surveys and benchmarking efforts that consolidated CNN-based histopathology methods and highlighted the need for reproducibility across institutions. Kather advanced large-scale, multi-cohort cancer classification from histology images, demonstrating cross-domain transferability of learned features and informing subsequent self-supervised and transformer-driven approaches.