Construcción de un modelo de procesamiento de historias clínicas electrónicas de pacientes con artritis reumatoide para la obtención de variables relevantes
Cargando...
Archivos
Autores
Cely Jimenez, Andrés
Director
Tipo de contenido
Document language:
Español
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Documentos PDF
Resumen
Este proyecto tuvo como fin desarrollar un modelo de procesamiento de lenguaje natural en el idioma español aplicado al dominio biomédico, con el fin de solucionar una necesidad desde del área de la medicina. Específicamente, en la atención de pacientes con artritis reumatoide, en donde se necesita extraer información de las historias clínicas. Fue necesario disponer de un corpus, o colección de documentos, tanto para el idioma español como para el dominio biomédico, porque se planteó el uso de un modelo basado en aprendizaje automático o de máquina usando un alto volumen de ejemplos, en este caso, un corpus anotado con etiquetas para la identificación de términos médicos.
Abstract: In this project a natural language processing model in Spanish applied to the biomedical domainwas developed, in order to fill a gap in the area of Medicine. Specifically, it was conceived as anaid for the care of patients with rheumatoid arthritis, where it is necessary to extract informationfrom the medical records. It was necessary to have a corpus, or collections of documents, in boththe Spanish language and the biomedical domain, since the use of a machine learning model withtraining using a high volume of examples was proposed, In this case, a corpus annotated withlabels for the identification of medical terms was used for training.A learning model was developed, based on recurrent neuronal networks using as training datasetan annotated corpus with UMLS semantic types. The model was trained to recognize (predict)medical entities. Once the model was trained, a software prototype was built under the Service-Oriented Architecture (SOA), in order to define and implement a web service to invoke the predic-tive model. A software client was developed, which in this case corresponds to a Web application,in Java, where the user can interact with the system to make annotations on unstructured text
Abstract: In this project a natural language processing model in Spanish applied to the biomedical domainwas developed, in order to fill a gap in the area of Medicine. Specifically, it was conceived as anaid for the care of patients with rheumatoid arthritis, where it is necessary to extract informationfrom the medical records. It was necessary to have a corpus, or collections of documents, in boththe Spanish language and the biomedical domain, since the use of a machine learning model withtraining using a high volume of examples was proposed, In this case, a corpus annotated withlabels for the identification of medical terms was used for training.A learning model was developed, based on recurrent neuronal networks using as training datasetan annotated corpus with UMLS semantic types. The model was trained to recognize (predict)medical entities. Once the model was trained, a software prototype was built under the Service-Oriented Architecture (SOA), in order to define and implement a web service to invoke the predic-tive model. A software client was developed, which in this case corresponds to a Web application,in Java, where the user can interact with the system to make annotations on unstructured text