Tesis Doctoral
DOI
10.11606/T.6.2014.tde-17072014-104356
Documento
Autor
Nombre completo
Bruna Bronhara Damiani
Dirección Electrónica
Área de Conocimiento
Fecha de Defensa
Publicación
São Paulo, 2014
Director
Tribunal
Antunes, Jose Leopoldo Ferreira (Presidente)
Barros, Laécio Carvalho de
Conde, Wolney Lisbôa
Marchioni, Dirce Maria Lobo
Ortega, Neli Regina de Siqueira
Título en portugués
Estimativa da composição corporal por impedância elétrica vetorial através de modelos fuzzy
Palabras clave en portugués
Câncer
Composição Corporal
Epidemiologia
Lógica Fuzzy
Resumen en portugués
Título en inglés
Body composition evaluation by Bioimpedance Vectorial Analysis using fuzzy models
Palabras clave en inglés
Body Composition
Cancer
Epidemiology
Fuzzy Logic
Nutritional Status
Resumen en inglés
Bioimpedance Vectorial Analysis (Biva) suggests seven patterns of body composition in a two-dimensional space composed by resistance and reactance. However, the boundaries defining those patterns are imprecise and Fuzzy Sets Theory has proved to be adequate to model the uncertainty inherent to the phenomenon. The aim was to build fuzzy models to estimate body composition according to Biva and apply them in differente populations. And italian database was used for refinement and testing (n = 20) and performance evaluation (n = 99) of the models. The consistency assessment was carried out in a hospitalized population (194 cancer patients) and non-hospitalized ( 1,716 U.S. individuals). Seven fuzzy models were build based on rules according to expert knowledge and scientific articles. The input variables were Z-score values of resistance/height and reactance/height. The output variable of each model was one of the seven categories of body composition: cachexia, obesity, dehydration, edema, athletic, thin or normal. Each model had as output a number ranged from 0 to 1 representing the degree of possibility to a given Biva pattern. The evaluation of performance was made using Kappa test. In the database of hospitalized and non-hospitalized patients the consistency of the models in identifying an given expected body composition was observed. An descriptive ans association analysis was held. The results showed good fit of the models and excellent agreement with the clinical condition of individuals (Kappa = 0.85, \ textit $<$ 0.001). The models were consistent with the positions of the impedance vector in the Biva graphy standardized in the two populations. The models were highly suggestive of possibility of cachexia in cancer patients. In North American patients, however, the higher possibility for edema suggested that in conditions such as cross-sectional area of the body should be considered in Biva approach. The fuzzy models proved to me promising in body composition according to Biva approach. Uncertainty was incorporated by using linguistic terms and a base-rule according to experts knowledge. By incorporating the uncertainty, the models provided more refined results than a classical approach.

ADVERTENCIA - La consulta de este documento queda condicionada a la aceptación de las siguientes condiciones de uso:
Este documento es únicamente para usos privados enmarcados en actividades de investigación y docencia. No se autoriza su reproducción con finalidades de lucro. Esta reserva de derechos afecta tanto los datos del documento como a sus contenidos. En la utilización o cita de partes del documento es obligado indicar el nombre de la persona autora.
BrunaBronhara.pdf (9.10 Mbytes)
Fecha de Publicación
2014-08-27