Loading...
Publication
ObsTrack: Geração Automática de Mapas Visuais Temporais de Acompanhamento Médico a partir de Registos Eletrónicos de Saúde com LLMs
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.99 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Esta dissertação apresenta o desenvolvimento do ObsTrack, um sistema inteligente capaz de
gerar mapas visuais temporais do acompanhamento clínico obstétrico a partir de Electronic
Health Records (EHRs) utilizando Large Language Models (LLMs). O trabalho parte da análise do
estado da arte sobre estruturação de dados médicos, Natural Language Processing (NLP) e
técnicas de visualização temporal, identificando limitações significativas, como a ausência de
integração entre dados estruturados e não estruturados e a falta de representações temporais
claras. A solução proposta combina processamento semântico de texto clínico com técnicas de
inferência temporal, permitindo extrair eventos relevantes e representá-los cronologicamente
numa interface gráfica interativa. A arquitetura do sistema integra-se com a plataforma
ObsCare, otimizando a comunicação médico-paciente e apoiando a tomada de decisão clínica.
Os resultados experimentais demonstram a capacidade do ObsTrack em gerar representações
coesas e interpretáveis da trajetória clínica, promovendo um acompanhamento mais eficiente,
personalizado e centrado na grávida. O trabalho conclui que a aplicação de LLMs em EHRs
representa um avanço significativo na forma como a informação clínica é explorada e
comunicada.
This dissertation presents the development of ObsTrack, an intelligent system capable of generating temporal visual maps of obstetric clinical monitoring from Electronic Health Records (EHRs) using Large Language Models (LLMs). The work is based on an analysis of the state -ofthe- art in medical data structuring, Natural Language Processing (NLP), and temporal visualization techniques, identifying significant limitations, such as the lack of integration between structured and unstructured data and the lack of clear temporal representations. The proposed solution combines semantic processing of clinical text with temporal inference techniques, allowing the extraction of relevant events and their chronological representation in an interactive graphical interface. The system architecture integrates with the ObsCare platform, optimizing doctor-patient communication and supporting clinical decision-making. Experimental results demonstrate ObsTrack's ability to generate cohesive and interpretable representations of the clinical trajectory, promoting more efficient, personalized, and pregnant woman-centered monitoring. The work concludes that the application of LLMs in EHRs represents a significant advance in the way clinical information is explored and communicated.
This dissertation presents the development of ObsTrack, an intelligent system capable of generating temporal visual maps of obstetric clinical monitoring from Electronic Health Records (EHRs) using Large Language Models (LLMs). The work is based on an analysis of the state -ofthe- art in medical data structuring, Natural Language Processing (NLP), and temporal visualization techniques, identifying significant limitations, such as the lack of integration between structured and unstructured data and the lack of clear temporal representations. The proposed solution combines semantic processing of clinical text with temporal inference techniques, allowing the extraction of relevant events and their chronological representation in an interactive graphical interface. The system architecture integrates with the ObsCare platform, optimizing doctor-patient communication and supporting clinical decision-making. Experimental results demonstrate ObsTrack's ability to generate cohesive and interpretable representations of the clinical trajectory, promoting more efficient, personalized, and pregnant woman-centered monitoring. The work concludes that the application of LLMs in EHRs represents a significant advance in the way clinical information is explored and communicated.
Description
Keywords
EHRs Knowledge Graph LLMs NLP Obstetrícia digital Visualização Temporal Digital Obstetrics Temporal Visualization