Dense Tissue Pattern Characterization Using Deep Neural Network

Abstract Breast tumors are from the common infections among women around world. Classifying various types of breast contribute to treating more efficiently. However, this classification task is often hindered by dense tissue patterns captured in mammograms. The present study has been proposed a pattern characterization framework using deep neural network. A total 322 mammograms belonging mini-MIAS dataset and 4880 DDSM have taken, an ROI fixed size 224 × pixels each mammogram extracted. In work, tedious experimentation executed different combinations training testing sets activation function with AlexNet , ResNet-18 model. Data augmentation used create similar type virtual image for proper DL After that, set applied on trained model validate During experiments, four functions ‘ sigmoid ’, tanh ReLu leakyReLu ’ used, outcome reported. It found that perform always outstanding respect others. For experiment, accuracy kappa coefficient computed. obtained value MIAS 91.3% 0.803, respectively. dataset, 92.3% 0.846 achieved. combination both images, achieved 91.9%, 0.839 Finally, it concluded yield performance task.