Publicação

RMID: a novel and efficient image descriptor for mammogram mass classification

Ver documento

Detalhes bibliográficos
Resumo:For mammogram image analysis, feature extraction is the most crucial step when machine learning techniques are applied. In this paper, we propose RMID (Radon-based Multi-resolution Image Descriptor), a novel image descriptor for mammogram mass classification, which perform efficiently without any clinical information. For the present experimental framework, we found that, in terms of area under the ROC curve (AUC), the proposed RMID outperforms, upto some extent, previous reported experiments using histogram based hand-crafted methods, namely Histogram of Oriented Gradient (HOG) and Histogram of Gradient Divergence (HGD) and also Convolution Neural Network (CNN). We also found that the highest AUC value (0.986) is obtained when using only the carniocaudal (CC) view compared to when using only the mediolateral oblique (MLO) (0.738) or combining both views (0.838). These results thus proves the effectiveness of CC view over MLO for better mammogram mass classification.
Autores principais:Obaidullah, Sk
Outros Autores:Ahmed, S.; Rato, Luis; Gonçalves, Teresa
Assunto:Image descriptor mammogram image breast cancer classification
Ano:2019
País:Portugal
Tipo de documento:artigo
Tipo de acesso:acesso restrito
Instituição associada:Universidade de Évora
Idioma:inglês
Origem:Repositório Científico da Universidade de Évora
Descrição
Resumo:For mammogram image analysis, feature extraction is the most crucial step when machine learning techniques are applied. In this paper, we propose RMID (Radon-based Multi-resolution Image Descriptor), a novel image descriptor for mammogram mass classification, which perform efficiently without any clinical information. For the present experimental framework, we found that, in terms of area under the ROC curve (AUC), the proposed RMID outperforms, upto some extent, previous reported experiments using histogram based hand-crafted methods, namely Histogram of Oriented Gradient (HOG) and Histogram of Gradient Divergence (HGD) and also Convolution Neural Network (CNN). We also found that the highest AUC value (0.986) is obtained when using only the carniocaudal (CC) view compared to when using only the mediolateral oblique (MLO) (0.738) or combining both views (0.838). These results thus proves the effectiveness of CC view over MLO for better mammogram mass classification.