BDBComp
Parceria:
SBC
Seleção de Atributos Relevantes para Busca por Similaridade e Classificação de Imagens Médicas

Marcela X. RibeiroJoaquim C. FelipeAgma J. M. Traina

This paper presents a new method to identify relevant features to discriminate images into categories.Shape features of a set of medical images are represented by Zernike moments. MinERA (MineradorEstátistico de Regras de Associação) is a new algorithm that identifies the most relevant features todiscriminate images into categories (benign and malignant tumours, for instance), through miningassociation rules. The algorithm uses statistical measures that describe the behaviour of the attributes, considering the image categories, in order to find rules of interest. An experiment with images of tumoral masses of mammograms compares the accuracy of MinERA and C4.5 algorithms to determine a set of relevant Zernike moments, capable to classify the images into benign and malignant. The results show that MinERA reaches a higher precision to perform this task.

http://www.lbd.dcc.ufmg.br/colecoes/wvc/2005/0011.pdf

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato: bdbcomp@lbd.dcc.ufmg.br
     Mantida por:
LBD