Predicting of Breast Cancer using Grad-CAM and Bounding Box

Abstract

Soft Computing, Machine Learning and Artificial Intelligence tools can save the lives through improved efficiency and diagnostic accuracy. For accurate diagnosis image analysis is the best approach. For this approach Feature extraction is a critical step in predicting any diagnosis. It involves identifying and extracting the most relevant features from the image which can then be used to train soft computing models for accurate prediction of diagnosis.   Soft Computing with processing the medical images is an effective method to knob suspicions inherent in attained image data. To lessen number of people lost to cancer, early diagnosis and treatment are paramount. In worldwide Breast Cancer is a common kind of cancer. As per WHO (World Health Organisation) 2.3 million new diagnoses and 685,000 demises in 2020 alone. The most cases of breast cancer are found in female. Uncontrolled progress of breast cells is the root cause of breast cancer. Breast cancer can take many forms. What form of breast cancer is in patient is determined by which breast cells become malignant. In this article, the breast cancer tumor is predicted with a feature extraction technique: Gradient-Weighted Class Activation Mapping (Grad-CAM) and Bounding Box (BB) along with Convolutional Neural Network (CNN) for accurate results in predicting the tumor and its location.