Ignifyr in Action
Ignifyr is the data integration engine of large European research projects in cardiology, oncology, paediatric palliative care and hospital care. In each project, the use cases define which data is needed, and these requirements are captured in a common data model. Most of these models are based on HL7 FHIR®, and some are based on the OMOP CDM. Ignifyr maps the data of every clinical site to the project’s model, so that AI and analytics tools work on the same harmonised data everywhere. In DataTools4Heart, AI4HF and AICCELERATE alone, this adds up to more than 300 million FHIR resources from more than 100,000 patients. We also mapped the public MIMIC-IV dataset to the same common data model. Ignifyr was previously released as toFHIR, the name used in some of the project publications.
DataTools4Heart
A European health data toolbox for cardiology
The challenge
Cardiology data that could advance research and care often stays unused in hospitals. Privacy rules prevent it from being pooled in one place, and every clinical site records it in its own formats, code systems and languages.
How Ignifyr is used
- Common data model: the data requirements of the use cases are captured in a FHIR-based common data model for heart-failure research, with 37 FHIR profiles, 6 code systems and 17 value sets built on ICD-10, ICD-10-PCS, LOINC, ATC and SNOMED CT.
- Mapping: each of the 8 clinical sites provides one dataset, and Ignifyr maps it locally, for example from CSV exports of the hospital information system, to the common data model. The resulting resources are stored in a local onFHIR repository, now available as Repofyr.
- Use: the feature extraction tool onfhir-feast turns the harmonised FHIR data into analysis-ready datasets for the federated learning platform. Patient data never leaves the hospital. The models travel to the data, and only model updates and aggregate results are shared.
Results
In September 2026, the DataTools4Heart platform entered clinical testing with real hospital data at the clinical sites. The same FHIR-based common data model is shared with the AI4HF project.
Project facts
- Programme
- Horizon Europe, grant agreement 101057849
- Duration
- October 2022 – September 2026
- Consortium
- 17 partners, coordinated by the University of Barcelona
- Clinical sites
-
- Region Stockholm (Karolinska University Hospital), Sweden
- Bucharest Emergency Clinical Hospital, Romania
- Fondazione Policlinico Universitario A. Gemelli IRCCS, Italy
- University College London, United Kingdom
- Amsterdam UMC, Netherlands
- International Clinical Research Center, St. Anne's University Hospital Brno, Czech Republic
- University Medical Center Utrecht, Netherlands
- Vall d'Hebron Research Institute, Spain
AI4HF
Trustworthy AI for personalised risk assessment in chronic heart failure
The challenge
Heart-failure risk models need large amounts of multi-source data, including cardiac imaging, biomarkers, ECGs and clinical reports, from diverse populations in both high-income and low- and middle-income countries. Every clinical site captures these data in different systems and formats.
How Ignifyr is used
- Common data model: the data requirements of the AI4HF use cases are captured in the same FHIR-based common data model as DataTools4Heart, so the same 37 profiles describe the data at every clinical site.
- Mapping: SRDC leads the project's data extraction, harmonisation and management tasks. At each clinical site, Ignifyr maps the local data to the common data model, combining HL7 FHIR and OMOP CDM to address both syntactic and semantic interoperability.
- Use: the harmonised data feeds a health data interoperability pipeline for AI model training and for running the trained models online.
Results
“We could quickly and effortlessly develop our next-generation healthcare applications using Ignifyr, integrating them with the existing healthcare systems in several European countries within the scope of AI4HF project.” — Gokce B. Laleci Erturkmen, AI4HF
Project facts
- Programme
- Horizon Europe, grant agreement 101080430
- Duration
- June 2023 – May 2027
- Budget
- €5.9 million
- Consortium
- 16 partners from Spain, Greece, Turkey, Czechia, Tanzania, Peru, Portugal, Germany, France, Belgium, the United Kingdom and the Netherlands, coordinated by the Netherlands Heart Institute
- Clinical sites
- 10 clinical sites in the Netherlands, Czechia, the United Kingdom, Tanzania, Peru, Spain and Portugal
AICCELERATE
AI for the smart hospital of the future
The challenge
AICCELERATE developed AI tools in three pilots: patient flow management for emergency and surgical units, a digital care pathway for Parkinson's disease, and paediatric care. Every AI model needed clean, consistent input data from hospital systems that were never designed to share it.
How Ignifyr is used
- Common data model: SRDC led the data infrastructure work package. The data requirements of the 7 use cases were captured in a FHIR-based common data model with 102 FHIR profiles, 10 code systems and 54 value sets.
- Mapping: Ignifyr, then called toFHIR, mapped the data of each clinical site to the common data model and loaded it into an onFHIR repository, now Repofyr.
- Use: onfhir-feast extracted analysis-ready datasets from the harmonised FHIR data for AI model training and online prediction.
Results
Example use case: predicting ICU admissions after cardiac surgery
In this Pilot 1 use case at Sant Joan de Déu Barcelona Children's Hospital, Ignifyr mapped raw CSV extracts to the common data model. The use case covered:
The complete pipeline is described in the peer-reviewed article “A scalable and transparent data pipeline for AI-enabled health data ecosystems”, Frontiers in Medicine, 2024.
“Ignifyr has greatly optimized our data transformation process. This tool has not only saved us time and effort but also enhanced the quality of our data, making it more valuable for machine learning in health research.” — A. Anil Sinaci, AICCELERATE
Project facts
- Programme
- Horizon 2020, grant agreement 101016902
- Duration
- January 2021 – April 2024
- Budget
- €10.9 million
- Consortium
- 16 partners from 8 countries, coordinated by Helsinki University Hospital
- Clinical sites
-
- Helsinki University Hospital, Finland
- Oulu University Hospital, Finland
- Ospedale Pediatrico Bambino Gesù, Italy
- Sant Joan de Déu Barcelona Children's Hospital, Spain
- University Hospital of Padova, Italy
MIMIC-IV
A public hospital dataset mapped to the heart-failure common data model
The challenge
MIMIC-IV is a large, publicly available database of de-identified health records from patients admitted to Beth Israel Deaconess Medical Center in Boston. Mapping it to the same FHIR-based common data model as the project data makes a widely used dataset available in the exact structure that the DataTools4Heart and AI4HF tools expect.
How Ignifyr is used
- Common data model: the target is the FHIR-based common data model of DataTools4Heart and AI4HF, with its 37 profiles, in FHIR R5.
- Mapping: Ignifyr mappings transform the MIMIC-IV hosp module, which holds hospital-wide records such as admissions, diagnoses, procedures, laboratory results and medications, into FHIR resources that conform to the common data model.
- Use: the resulting FHIR data has the same structure as the data at the clinical sites, so the same feature extraction and analysis tools run on it.
Results
Dataset facts
- Dataset
- MIMIC-IV, hosp module
- Source
- Beth Israel Deaconess Medical Center, Boston, United States
- Access
- Credentialed access through PhysioNet
- Target
- DataTools4Heart and AI4HF common data model, FHIR R5
Have a similar data integration challenge?
Tell us about your clinical sites, use cases and target data model, and we will show you how Ignifyr can help.
Contact Us