To search, Click below search items.


All Published Papers Search Service


Localization-Based Methodology for a Fully Automatic Breast Cancer Image Diagnosis System Using Convolutional Neural network


Ali Fawzi Mohammed Ali and Mehdi G. duaimi


Vol. 17  No. 7  pp. 46-57


Digital histopathology images represent the major evolutions in a modern medicine diagnosis. Pathological examinations consider the standard medical diagnosis protocols and play a critical and important role in the diagnosis process. Early tumor detection step in the diagnosis stage is obtained by the cytological testing of the breast image mainly based on the cell morphology and architecture distribution. In this paper, we present a localization methodology to predict and diagnosis the breast cancer type. The proposal relies on a fully automatic analysis of the fine needle biopsies for cytological images. A fully automatic isolation (cells detection) and filtration steps of the nuclei cells images are designed and implemented as a fully automatic computer-aid system to classify the nuclei cells type of the breast cancer by using such a powerful classification approach. Instead of relying on supervised classifiers such as an SVM and Neural Network, we proposed such robustness framework for the breast cancer cell image diagnosis system using a Convolutional Neural Network (CNN). Our approach in this paper relies on design and implementation such a fully supervised learning system to diagnosis the nuclei cells images by two main steps. The first one is the cells detection and filtration approach using Circular Hough Transform (CHT) and Support Vector Machine (SVM). In this step, we rely on fully automatic cells images selection that we have designed and implemented to train the SVM classifier that is used later for the cells filtration. This approach allows to automatically selection of a set of nuclei images to train the SVM classifier which is proposed for cell images filtration and isolation. Secondly, the new set of nuclei cell images is used by the CNN to predict and classify the breast cancer types. The fully automatic diagnostic system achieves about (99.41%) diagnosis accuracy results which illustrate a higher performance diagnosis system based on our implementation which is effective, valuable by providing an accurate diagnostic accuracy result as either benign or malignant.


Breast cancer, histopathological images, SVM classification, Convolutional Neural Network (CNN).