Medizinische Universität Graz Austria/Österreich - Forschungsportal - Medical University of Graz
Gewählte Publikation:
Kreuzthaler, M.
Medical Information Search in Semi-Structured Data
Doktoratsstudium der Medizinischen Wissenschaft; Humanmedizin; [ Dissertation ] Graz Medical University; 2015. pp. 120
[OPEN ACCESS]
FullText
- Autor*innen der Med Uni Graz:
- Kreuzthaler Markus Eduard
- Betreuer*innen:
- Berghold Andrea
- Schulz Stefan
- Altmetrics:
- Abstract:
- Computer systems for clinical information management store large amounts of textual data in medical records. Finding reports and patient summaries use semi-structured document templates. Coded data, using controlled vocabularies, are mainly restricted to accounting, research, and quality assurance. Unstructured or semi-structured content is difficult to analyse, although there are multiple use cases for content retrieval from clinical texts which would benefit from semantically enhanced retrieval functionalities. This thesis focuses on the investigation of clinical narratives in combination with improved semantic indexing and extraction systems for patient-based decision making. It addresses the development and evaluation of technical solutions to support health professionals and researchers in retrieving targeted patient-related information in a timely and efficient way, according to their information needs. The information to be searched for is constituted within medical free text from various clinical domains in a hospital environment. Different state of the art approaches are explored to what extent they can be adapted to the domain and how they can be optimized to apply to clinical professionals' information needs. These approaches, applied to anonymized clinical textual data, show the potential of adapted solutions to medical domains and related sublanguages for enhanced information retrieval. The following search scenarios have been investigated: collection-based patient search and patient-based document search. Within these search scenarios, enhanced text processing methods have shown their applicability to support domain expert retrieval. The results show the trade-off of clinical information systems and the possibilities of novel frameworks and technologies for unstructured information processing. For selected clinical information system content, they can bridge the gap between patient-based storage systems and disease-related search systems.