Digital pathology is scaling fast, but without standardized and machine-readable tissue metadata, “find me all slides with X”, still means manual browsing, local naming conventions, and ad hoc filters. In order to assemble robust cohorts and run AI at scale, a standard that makes slides discoverable across archives is needed.
A Whole Slide Image (WSI) is a high-resolution digital scan of an entire glass slide containing a biological specimen (e.g., tissue sections or cell samples). WSIs are digitally viewable, analysable, and shareable, and they are widely used for artificial intelligence (AI) algorithm development. They are central in pathology and oncology, but also in neurology, hematology, microbiology, dermatology, pharmacology, toxicology, immunology, veterinary medicine, and forensic science.
The challenge here is that when assembling cohorts for AI training or validation, it is essential to know the actual morphological content of a WSI. Today, no interoperable standard exists for this type of structured metadata. As a result, slide selection often relies on manual inspection — a process that does not scale to archives containing millions of images.
It is this challenge that is addressed in the new paper: “From slides to AI-ready maps: Standardized multi-layer tissue maps as metadata for artificial intelligence in digital pathology.” Find it at doi.org/10.1016/j.artmed.2026.103368
In this publication, a general framework for generating standardized 2D tissue maps as WSI metadata is introduced. These maps describe morphological content using a shared syntax and semantics to enable interoperability between catalogues. The tissue maps are structured in three hierarchical layers:
-
Source
-
Tissue type
-
Pathological alterations
Each layer assigns WSI segments to defined classes, transforming slides into searchable, AI-ready objects. By integrating these standardized multi-layer tissue maps into WSI archives, we significantly enhance search functionality and enable the accelerated assembly of high-quality, balanced, and targeted datasets for AI development, validation, and cancer research.
The collaborators on this publication hail from the Medical University Graz, Masaryk University, Masaryk Memorial Cancer Institute, BBMRI-ERIC, and EMPAIA. The team includes Gernot Fiala, Markus Plass, Robert Harb, Peter Regitnig, Kristijan Skok, Wael Al Zoughbi, Carmen Zerner, Paul Torke, Michaela Kargl, Heimo Müller, Tomáš Brázdil, Matej Gallo, Jaroslav Kubín, Roman Stoklasa, Rudolf Nenutil, Norman Zerbe, Andreas Holzinger, and Petr Holub