Modelling digital health data: The ExaMode ontology for computational pathology

Area of Science:

• Digital Pathology
• Medical Informatics
• Ontology Engineering

Background:

• Computational pathology requires standardized nomenclature for knowledge extraction from annotated image datasets.
• Data annotation in digital pathology faces challenges like label heterogeneity, multilingualism, and diverse clinical practices, hindering dataset reuse.
• A shared model is needed to overcome data variability and integration problems in digital pathology.

Purpose of the Study:

• To present the ExaMode ontology, a novel semantic model for histopathology.
• To standardize nomenclature and facilitate knowledge extraction in digital pathology.
• To enable better data integration and analysis of multimodal histopathology data.

Main Methods:

• The ExaMode ontology was developed bottom-up with iterative feedback from pathologists and clinicians.
• It models the histopathology process for colon, cervical, and lung tumors, and celiac disease.
• The ontology is organized into 5 semantic areas, providing a template for various histopathology diseases.

Main Results:

• The ExaMode ontology serves as a common semantic layer for tools including entity linking, collaborative annotation, and multimodal data integration platforms.
• It enables data storage in graph databases using the RDF model, facilitating the development of advanced image analysis algorithms.
• Seamless data integration and unified query access via SPARQL enable extraction of clinical insights.

Conclusions:

• The ExaMode ontology addresses key challenges in digital pathology by providing a standardized semantic framework.
• Its application supports improved data annotation, integration, and analysis, leading to more accurate computational pathology algorithms.
• This ontology facilitates the extraction of valuable clinical insights from heterogeneous histopathology data.