Breast Cancer Detection Using Deep Multilayer Neural Networks

Abstract

Breast cancer is the most common cancer among women and is the second leading cause of death. There is currently no efficient way to prevent breast cancer, but its detection in early stages can increase the patient's chances of being cured and surviving. Computer-aided diagnosis (CAD) systems, based on image processing techniques, can provide a more reliable interpretation of mammographic images to detect microcalcifications and have been able to identify and classify benign and malignant tumors. If we are dealing with a massive number of images, this system increases the ability and accuracy of detection. Also, in cases where the number of images is not large, CAD systems can significantly improve the image quality. In addition, a CAD system can identify suspicious areas to provide radiologists with a visual aid to interpret mammograms. Deep learning and convolutional neural networks have recently shown significant performance for visual applications. Convolutional neural networks have also been used efficiently to analyze medical images and diagnose mammograms. In this paper, a CAD system based on convolutional neural networks (Mask R-CNN) with multi-task learning to detect breast cancer and segment mammogram images is proposed. The Mask-RCNN technique is one of the strongest and most flexible deep grids ever designed for machine vision. In this article, multitask learning with the integration of two tasks of classification and segmentation is used to diagnose breast cancer. R-CNN convolution neural network is used to diagnose cancerous mass. This system consists of two main stages, including the production of pseudo-color image and segmentation-detection based on convolutional neural networks (R-CNN Mask). The INbreast dataset is employed for evaluation of the proposed method.

Keywords

• multitask learning
• deep learning
• convolutional neural network
• breast cancer
• cad