Breast Cancer Detection from Mammography Images using Convolutional Neural Network (CNN) Method ResNet-50 Architecture

Abstract

Breast cancer is one of the most common types of cancer in the world and a serious health issue for women, making it crucial for every woman to undergo regular early screenings to minimize mortality rates. This screening can be performed using mammography and mammogram image analysis. This study proposes a breast cancer detection method based on Convolutional Neural Network (CNN), using texture and shape parameters from mammogram images to determine the class. There are two classes trained using CNN, namely cancer and non-cancer. This study compares the performance of the original ResNet-50 architecture model with the modified ResNet-50 architecture at the last layer through the addition of fully connected layers, batch normalization, and dropout. The results show that this model is capable of detecting breast cancer with high accuracy. The best model in this study uses a modified ResNet-50 architecture, with the Adam optimizer, batch size of 8, sigmoid activation function, learning rate of 0.0005, and 30 epochs. This model achieved an overall average with an accuracy of 97.00%, precision of 97.00%, recall of 97.00%, and f1-score of 97.00%.