Comparison of Transfer Learning Model Performance for Breast Cancer Type Classification in Mammogram Images

Globally, breast cancer is the type of cancer that most women suffer from. Early detection of breast cancer is very important because there is a big chance of cure. Mammography screening makes it possible to detect breast cancer early. The study of computer-assisted breast cancer diagnosis is gaining increasing attention. Breast cancer comes in two forms: benign cancer and malignant cancer. advances in deep learning (DL) technology and its use to overcome obstacles in medical imaging, and classification using a number of transfer learning models to identify the type of breast cancer (malignant, benign, or normal). This work conducted a thorough comparison analysis of eight prevalent pre-trained CNN algorithms (VGG16, ResNet50, AlexNet, MobileNetV2, ShuffleNet, EfficientNet-b0, EfficientNet-b1, and EfficientNet-b2) for breast cancer classification. In this study, we permonData is divided into training, testing, and validation. Using the publicly accessible mini-DDSM dataset, we assess the proposed architecture. were used to measure the classification accuracy (Acc). For genBased on test results, the best accuracy was obtained using EfficientNetb2 with an accuracy value of 94% for training data and 98% for test data on mammogram images.