SIGNIFICANTLY ACCURATE SYSTEM FOR BREAST CANCER MALIGNANCY OR BENIGN CLASSIFICATION

Abstract

Breast cancer happens to one out of eight females worldwide. It is the most elevated reason for cancer malignancy deadliness among ladies. It is identified by finding the cancerous cells in breast tissue. Novel techniques in medical image processing utilized histopathology dataset images taken by an advanced microscope, and then disintegrate the images by applying various algorithms and techniques. Artificial Intelligence methods are presently being applied for processing pathological imagery and tools. Here in the project work, we concentrate on building up the capability of computer-aided diagnosis (CAD) to anticipate the severity of cancerous cells. Common cancerous cell detecting is a tedious process and involves the fault of physicians, to this end we can use computer-aided detection (CAD) system to reduce the fault and obtain the more acceptable outcome in comparison to a common pathological detection system. Here we are comparing, our framework with the other three machine learning frameworks in breast image segmentation and classification on a well-known dataset (BreakHis) trial arrangement. Classification in deep neural network mainly utilize feature extraction by the means of convolutional neural network and then by embedding a fully connected network, the result would be an acceptable output. Deep learning has a vast amount of functionality in medical image processing without any need for supervision of any professional person during the process and the procedure can be done automatically. Here in our project we train a bunch of histopathology images through a convolutional neural network and obtain accuracy in prediction more than 92%.