Segmentação de massas em mamografias digitalizadas

Detalhes bibliográficos
Ano de defesa: 2012
Autor(a) principal: Wirtti, Tiago Tadeu
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal do Espírito Santo
BR
Mestrado em Engenharia Elétrica
Centro Tecnológico
UFES
Programa de Pós-Graduação em Engenharia Elétrica
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://repositorio.ufes.br/handle/10/9628
Resumo: This work suggests a methodology for segmentation of masses in digital mammograms. The masses are distinguished from the other breast tissue by its homogeneous and differentiated density, and its peculiar shape: rounded, spiculated or undefined. The segmentation strategy is based on slicing the mammography by ranges of pixel intensity and on the assessment of each slice density using multiscale wavelet transform. The density data obtained from a wavelet transform are used to train a multilayer perceptron network. After the training phase, any mammography, except those used in the training phase, may be submitted to the trained neural network. Each image slice resulting from processing handled by the neural network has evidenced the relevant characteristics of the original image. The findings in each slice are evaluated by a gradient filter, generating slices containing relevant information on a gradient accumulated structure for each finding. The accumulated gradients that appear in the same position in subsequent slices are heuristically analyzed resulting in the selection of the masses. After processing 31 images from mini-MIAS database of mammograms (two images for training and the other for testing the classifier) it was obtained the following results: TPR (sensitivity) of 75.00%, 23.91% of FPR, and specificity of 76.09%.