Altered Neural Net for Breast Histological Image Categorization
Rajyalakshmi Uppada, Nainavarapu Radha · 2023
Manual diagnosis of histopathological breast images is both time-taking and inaccurate process. Breast-cancer Computerized Diagnosis (BCD) using the Hematoxylin-Eosin (HE) tissue attained from biopsy relies on the efficiency and accuracy of the process. Accuracy of the conventional classification methods depend on extracted features. To conquer the difficulties of traditional feature-based techniques, deep-learning approaches are preferred as an alternative in attaining accurate classification. Deep Altered-Convolutional Neural Network (A-CNN) for categorization of HE breast tissue images as Benign/Malignant is stated. Network architecture of A-CNN is intended to gain information at nuclei and overall tissue levels. Extracted features from A-CNN are used to train the Multi-Class SVM (MC-SVM) classifier. The cascaded categorization method accomplished better accuracy of 94.16 % and a best AUC of 0.913 than the individual ones.