Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Clinical Data Goes MEDS? Let's OWL make sense of it

Alberto Marfoglia, Jong Ho Jhee, Adrien Coulet

TL;DR

This work tackles the lack of semantic representation in structured EHR data by introducing MEDS-OWL, an OWL ontology that formalizes the MEDS event model, and meds2rdf, a converter that maps MEDS datasets to RDF graphs aligned with MEDS-OWL. The methodology integrates provenance and dataset metadata via DCAT and PROV-O, and enforces semantic constraints through SHACL validation, enabling reproducible, FAIR-aligned, graph-based analytics. A synthetic neurovasc dataset demonstrates the end-to-end pipeline, producing a compact RDF representation (over 1.3 million triples) with explicit event, subject, and code entities and rich provenance. The work positions MEDS-OWL as a reusable semantic layer that bridges pragmatic clinical event models with the Semantic Web, paving the way for interoperability across datasets and advanced graph-based reasoning, with open-source tooling for broader adoption.

Abstract

The application of machine learning on healthcare data is often hindered by the lack of standardized and semantically explicit representation, leading to limited interoperability and reproducibility across datasets and experiments. The Medical Event Data Standard (MEDS) addresses these issues by introducing a minimal, event-centric data model designed for reproducible machine-learning workflows from health data. However, MEDS is defined as a data-format specification and does not natively provide integration with the Semantic Web ecosystem. In this article, we introduce MEDS-OWL, a lightweight OWL ontology that provides formal concepts and relations to enable representing MEDS datasets as RDF graphs. Additionally, we implemented meds2rdf, a Python conversion library that transforms MEDS events into RDF graphs, ensuring conformance with the ontology. We demonstrate the approach on a synthetic clinical dataset that describes patient care pathways for ruptured intracranial aneurysms and validate the resulting graph using SHACL constraints. The first release of MEDS-OWL comprises 13 classes, 10 object properties, 20 data properties, and 24 OWL axioms. Combined with meds2rdf, it enables data transformation into FAIR-aligned datasets, provenance-aware publishing, and interoperability of event-based clinical data. By bridging MEDS with the Semantic Web, this work contributes a reusable semantic layer for event-based clinical data and establishes a robust foundation for subsequent graph-based analytics.

Clinical Data Goes MEDS? Let's OWL make sense of it

TL;DR

This work tackles the lack of semantic representation in structured EHR data by introducing MEDS-OWL, an OWL ontology that formalizes the MEDS event model, and meds2rdf, a converter that maps MEDS datasets to RDF graphs aligned with MEDS-OWL. The methodology integrates provenance and dataset metadata via DCAT and PROV-O, and enforces semantic constraints through SHACL validation, enabling reproducible, FAIR-aligned, graph-based analytics. A synthetic neurovasc dataset demonstrates the end-to-end pipeline, producing a compact RDF representation (over 1.3 million triples) with explicit event, subject, and code entities and rich provenance. The work positions MEDS-OWL as a reusable semantic layer that bridges pragmatic clinical event models with the Semantic Web, paving the way for interoperability across datasets and advanced graph-based reasoning, with open-source tooling for broader adoption.

Abstract

The application of machine learning on healthcare data is often hindered by the lack of standardized and semantically explicit representation, leading to limited interoperability and reproducibility across datasets and experiments. The Medical Event Data Standard (MEDS) addresses these issues by introducing a minimal, event-centric data model designed for reproducible machine-learning workflows from health data. However, MEDS is defined as a data-format specification and does not natively provide integration with the Semantic Web ecosystem. In this article, we introduce MEDS-OWL, a lightweight OWL ontology that provides formal concepts and relations to enable representing MEDS datasets as RDF graphs. Additionally, we implemented meds2rdf, a Python conversion library that transforms MEDS events into RDF graphs, ensuring conformance with the ontology. We demonstrate the approach on a synthetic clinical dataset that describes patient care pathways for ruptured intracranial aneurysms and validate the resulting graph using SHACL constraints. The first release of MEDS-OWL comprises 13 classes, 10 object properties, 20 data properties, and 24 OWL axioms. Combined with meds2rdf, it enables data transformation into FAIR-aligned datasets, provenance-aware publishing, and interoperability of event-based clinical data. By bridging MEDS with the Semantic Web, this work contributes a reusable semantic layer for event-based clinical data and establishes a robust foundation for subsequent graph-based analytics.
Paper Structure (5 sections, 4 figures, 5 tables)

This paper contains 5 sections, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Results and example usage
  4. Discussion
  5. Conclusion

Figures (4)

  • Figure 1: Overview of the end-to-end MEDS KG construction pipeline, illustrating (a) normalization and extraction of tabular data into MEDS event tuples (optional, if data are already in MEDS), (b) transformation into RDF KGs compliant with the MEDS-OWL ontology, and (c) SHACL-based validation and serialization into standard RDF formats.
  • Figure 2: Overview of the MEDS-OWL ontology. Azure nodes and edges denote locally defined classes and object properties, respectively. Black edges with green labels are datatype properties. Nodes and edges colored differently indicate imported entities from either the Data Catalogue Vocabulary (dcat prefix), Dublin Core Terms (dct prefix) or PROV-O (prefix prov). Dotted edges represent subClassOf predicate. The figure is drawn with WebVOWL wiens2018webvowl.
  • Figure 3: MEDS-OWL conversion workflow for the neurovasc dataset. (a) Initial preprocessing of three patient features: age, nimodipine (drug administration), and iot (tracheostomy procedure). (b) Conversion to MEDS tabular format via MEDS-extract, with each feature as a distinct event. (c) Transformation of MEDS output into RDF graphs using meds2rdf.
  • Figure 4: Provenance linkage of MEDS events, codes, and subjects using MEDS-OWL, PROV-O, and Dublin Core. Each event references its source meds:DatasetMetadata (here, neurovasc) via prov:wasDerivedFrom and is associated with the generating ETL activity (prov:Activity), here MEDS-extract. This approach provides a semantically rich, traceable representation of clinical data workflows.