Renan Caldeira Menechelli,  Patricia Bellin Ribeiro,  Homero Schiabel. 

Due to constant increase in the occurrence of breast cancer, many studies are intended to assist the diagnosis by computer. The experiments were performed with real images and the images that simulate mammographic nodules. About all of images are extracted shape descriptors to be submitted to classification as inputs to an artificial neural network MLP. The best values obtained with the generalization in the classification were of 72% to simulation images in 5 classes with Az = 0.75 and 82% to the real images in 2 classes with Az = 0.79. Then, we developed a software that implements the MLP to aid the classification of nodules present in mammographic images in 2 classes using shape descriptors.

http://www.lbd.dcc.ufmg.br/colecoes/wvc/2010/0032.pdf

Processando consulta.