Publicação

Brain magnetic resonance spectroscopy classifiers

Ver documento

Detalhes bibliográficos
Resumo:During the last decade, the Magnetic Resonance Spectroscopy modality has become an integrant part of the diagnostic routine. However, the visual interpretation of these spectra is difficult and few clinicians are trained to use the technique. In this study, sixty-eight spectra obtained from twenty-two multi-voxel spectroscopies were classified using three well-known classification algorithms: K-Nearest Neighbors (KNN), Decision Trees and Naïve Bayes. The best results were obtained using NaïveBayes that presented an average balanced accuracy rate around 75%, although K-Nearest Neighbors presented very good results in some situations. The obtained results leads us to conclude that it is possible to classify magnetic resonance spectra with data mining techniques for further integration in a Clinical Decision Support System which may help in the diagnosis of new cases.
Autores principais:Alves, Victor
Outros Autores:Oliveira, Susana Marta Fonseca de; Rocha, Jaime
Assunto:Magnetic Resonance Spectroscopy Pattern Classification Decision Support Systems K-nearest Neighbor Decision Tree Naïve Bayes
Ano:2010
País:Portugal
Tipo de documento:comunicação em conferência
Tipo de acesso:acesso restrito
Instituição associada:Universidade do Minho
Idioma:inglês
Origem:RepositóriUM - Universidade do Minho
Descrição
Resumo:During the last decade, the Magnetic Resonance Spectroscopy modality has become an integrant part of the diagnostic routine. However, the visual interpretation of these spectra is difficult and few clinicians are trained to use the technique. In this study, sixty-eight spectra obtained from twenty-two multi-voxel spectroscopies were classified using three well-known classification algorithms: K-Nearest Neighbors (KNN), Decision Trees and Naïve Bayes. The best results were obtained using NaïveBayes that presented an average balanced accuracy rate around 75%, although K-Nearest Neighbors presented very good results in some situations. The obtained results leads us to conclude that it is possible to classify magnetic resonance spectra with data mining techniques for further integration in a Clinical Decision Support System which may help in the diagnosis of new cases.