Búsqueda de historiales clínicos similares
Loading...
Official URL
Full text at PDC
Publication date
2020
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
Los historiales médicos contienen mucha información sobre un paciente. No solo síntomas, si no tratamientos aplicados y resultados de los mismos. En este trabajo crearemos una herramienta enfocada a los médicos en forma de aplicación, con interfaz web incluida, que analice y compare diferentes historiales médicos mediante diferentes técnicas de procesamiento de texto independientemente de su formato y devuelva el grado de similitud del mismo con otros para facilitar y agilizar la labor de los médicos a la hora de descubrir causas o asignar tratamientos a diferentes enfermedades. Los resultados iniciales, debido al formato simple, y cercano al lenguaje real de los historiales, reflejan bastante fielmente la similitud entre los historiales obteniéndose mejores resultados en historiales de la misma especialidad médica o de sintomática similar. Sin embargo, los segundos resultados, no fueron los esperados debido al formato de los archivos, mucho más técnico y difícil de manejar, denotando la importancia de mantener una cohesión estructural entre los archivos de una misma colección.
Medical reports contain several information about a patient. Not only Symptoms, but also applied treatments and their results, whether they are good or not. In the present work, we will create an application that analyzes and compares different medical reports by text analysis regardless of the format of the data-set applied which gives back the similarity percentage of that report with the rest, with the objective of simplify and hasten the doctor’s job when having to discover the cause or assign treatments to different illnesses.The initial results taking a collection with normalized data, are close to reality, showing a higher similarity between reports of the same medical branch or with similar symptoms thanks to their simplicity, and human-like language format. The second results, however, were not as good as we expected due to the a higher complexity and technical vocabulary. This shows the importance of keeping an internal structural cohesion between the data of the same collection.
Medical reports contain several information about a patient. Not only Symptoms, but also applied treatments and their results, whether they are good or not. In the present work, we will create an application that analyzes and compares different medical reports by text analysis regardless of the format of the data-set applied which gives back the similarity percentage of that report with the rest, with the objective of simplify and hasten the doctor’s job when having to discover the cause or assign treatments to different illnesses.The initial results taking a collection with normalized data, are close to reality, showing a higher similarity between reports of the same medical branch or with similar symptoms thanks to their simplicity, and human-like language format. The second results, however, were not as good as we expected due to the a higher complexity and technical vocabulary. This shows the importance of keeping an internal structural cohesion between the data of the same collection.
Description
Trabajo de fin de Grado, Facultad de Informática UCM, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2019/2020