PhD

Phd Témánk:

• Adatgyűjtés, adathasznosítás új lehetőségeinek vizsgálata az egészségügyi informatikában
  Témavezető: dr. Bagyura Zsolt

Phd hallgatóink:

Bali Orsolya

Research topic:

This research project aims to improve the way we structure clinical information, which is often recorded as free-text notes by doctors and nurses. The central goal is to develop and test a step-by-step workflow that leverages state-of-the-art artificial intelligence techniques, like Large Language Models (LLM) and Retrieval-Augmented Generation (RAG). This approach aims to demonstrate the value of using standard semantic ontologies, like SNOMED for data encoding both at the data transformation and the data consumption part, focusing on diagnoses, medical procedures, or laboratory results.

To prove the effectiveness of this new workflow, we will compare its performance against a more straightforward method where the LLM assigns codes without the retrieval step, as well as against a “gold standard” dataset that has been manually coded by experts. The expectation is that the RAG-based workflow will prove to be more accurate, less prone to errors or hallucinations, and better equipped to manage complex medical terminology, ultimately offering a more robust and dependable guidance for structuring clinical data in everyday healthcare settings. The study also aims to illustrate the advantages of utilizing standard ontologies for medical data encoding, particularly within the context of cross-border data sharing initiatives such as the EHDS.

Székely Orsolya

Research topic:

Although data structuring has become a clear policy intention in EU recently, there are other challenges to be addressed in order to best serve clinical research with healthcare data. To date, most clinical data is still available in an unstructured format, even though data standards such as SNOMED CT, LOINC, FHIR or ICD-10 are actively being used, and it is difficult to retrieve clear, high-quality data from electronic health records. The Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) can serve as a common ground for all types of medical data. It allows researchers to organise and standardise medical terms and to systematically analyse medical observations from different research settings, often using different data formats or representations. In addition, the integration of personal health records into research could serve as another opportunity for adaptation, in recognition of more personalised medical care.