Automated medical reporting that gives time back to care.
MediSpeech is an international research & innovation project building an open, AI-powered ecosystem that turns the clinical conversation into structured, interoperable medical records.
Up to a third of a clinician's working hours are spent on administrative documentation — time taken away from patients and a leading driver of physician burnout. MediSpeech targets that burden at its source.
From spoken consultation to structured clinical data.
MediSpeech captures the doctor–patient conversation and transforms it, in real time, into accurate, standardised and interoperable clinical documentation — ready for the electronic health record and for downstream decision support.
Speak
The consultation is captured through medical-grade automatic speech recognition, tuned for clinical language and multiple European languages.
Understand
NLP extracts symptoms, diagnoses and plans, mapping them to MeSH, SNOMED-CT, LOINC, ICPC-2 and ICD-10.
Report
Structured notes flow into the EHR over open, certified FHIR/OpenAPI interfaces, with explainable decision support.
AI-powered speech recognition
Medical-grade ASR that handles clinical terminology and the idiosyncrasies of individual speech, targeting a word error rate below 15% across multiple languages.
Data interoperability & harmonisation
Open, secure APIs over EHR standards (FHIR, OpenEHR), certified to ISO 27001 and NEN 7510, to move structured transcripts safely across heterogeneous systems.
Technology-advanced decision support
Explainable clinical decision support combining knowledge-based rules with reinforcement learning, keeping recommendations transparent and auditable for clinicians.
Automated medical reporting
Extracted entities and relations are assembled into structured, explainable medical notes and patient-journey artefacts — reducing admin time and improving completeness.
Nineteen partners, four countries, one mission.
MediSpeech brings together clinical, technology, insurance and academic partners across the Netherlands, Portugal, Türkiye and the United Kingdom, coordinated by MEDrecord BV.
Netherlands
- MEDrecord BV· project coordinator
- HealthTalk.ai
- Amsterdam UMC (AMC)
- Radboud University
- TNO
- University of Twente
Portugal
- MTG Research & Development Lab
- MedicineOne
- University of Porto, Faculty of Medicine
- he-ufp· Fernando Pessoa University Hospital
Türkiye
- Anadolu Sigorta
- SESTEK A.Ş.
- Turkcell Teknoloji
- Doğuş Bilgi İşlem ve Teknoloji Hizmetleri
- Experteam
- NP Istanbul Brain Hospital
United Kingdom
- Chilton Computing Ltd
- Bristol Heart Institute
- Digital Health & Care Alliance (DHACA)
Validated where it matters most.
Each national consortium develops and demonstrates MediSpeech in a real clinical setting, proving the technology against the languages, workflows and standards of its healthcare system.
Primary & specialist care documentation
Converting Dutch consultations into structured, OpenEHR-compliant records, delivering semantic interoperability to Dutch healthcare institutes through open, certified APIs.
Primary care follow-up
Four primary-care areas — type 2 diabetes, hypertension, child & adolescent health and low-risk pregnancy — with bilingual Portuguese/English processing and automatic coding into ICPC-2, ICD-10, LOINC and SNOMED-CT.
Neurodevelopmental care
Supporting diagnosis and management of autism spectrum disorder and attention deficit, with a speech-to-text intake and a smart engine for NLU and clinical decision support across on-site, online and telehealth channels.
Atrial fibrillation risk
Identifying and extracting atrial fibrillation risk factors from free-text cardiac notes — drawing on ~200K patient records — for earlier diagnosis and targeted treatment after cardiac surgery.
Six work packages across 42 months.
From January 2025 to June 2028, the consortium delivers MediSpeech through six interlocking work packages, each led by a partner with deep expertise in its field.
Consortium agreement
Governance and IP framework established.
Privacy guidelines
Data protection guidelines across all four countries.
Requirements set
Use cases, specifications and data protocols defined.
ASR & EHR APIs
First speech and interoperability APIs delivered.
AI models
Information-extraction and decision-support models.
National demonstrators
One clinical demonstrator per country.
Demonstration catalogue
Public catalogue of MediSpeech services.
Final evaluation
Usability, exploitation and project close-out.