Automated classification of masses on mammography

Luciana Marinho SoaresAura ConciAlberto D. Vianna

A scheme for identification of breast cancer as benignan or malignant based on pattern recognition is presented. A database for use by the mammographic image analysis research community has been established. From these images, 52 cases with undoubted diagnosis have been used as input pattern for feacture extraction and classification training. After extensive experimentation a set of feactures is extracted using shape and contour characterization. Two classes of classifier are used: discriminant functions and nearest neighbor classifier.We implemented an automatic computer dignosis system that performs analysis capable of correct classification on all tested cases until now.

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

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: