MedalioNER

Local and ultra-lightweight AI model for extracting clinical information from medical documentation.

MedalionNER is a Named Entity Recognition (NER) system designed specifically for processing Polish-language medical documentation. Trained on 50,000 anonymized clinical notes, it accurately extracts key clinical information from unstructured texts — with performance on par with human experts.

What MedalioNER extracts

Diseases and Diagnoses

(e.g. myocardial infarction, type 2 diabetes

Test Results

(e.g. CRP: 42 mg/L, chest X-ray: left lower lobe opacity)

Medications

(both brand names and active substances, e.g. Paracetamol, Enalapril)

Anatomical Terms

(e.g. left kidney, thyroid gland)

Symptoms

(e.g. chest pain, fever, shortness of breath)

Departments & Specialties

(e.g. Emergency, Neurology, Cardiology Unit)

Medical Procedures

(e.g. abdominal ultrasound, gastroscopy, knee replacement)

Demographics & Identifiable Information

(e.g age, weights, name, surname)

Use Cases

Integration with HIS and EHR systems

Automatic extraction and population of structured fields from free-text notes
Filtering and retrieval of patient cases by diagnosis, procedure, or medication
Accelerated billing support via automatic code suggestions (ICD-10, ICD-9, ATC)

Scientific research & cohort studies

High-volume clinical data extraction for retrospective analysis
Efficient patient selection based on complex criteria (e.g. age > 65 + dyspnea + ACE inhibitors)
Anonymization of records for GDPR-compliant data sharing
Trend analysis of treatments, outcomes, and adherence to clinical guidelines

Regulatory compliance & data protection

Automatic detection and masking of sensitive information (names, IDs, dates)
Fully local processing — no data leaves your infrastructure
Audit support and quality assurance automation