ABSTRACT
The most common malignant type of breast tumor is Invasive Ductal Carcinoma (IDC). The IDC refers to cancer that has broken through the milk duct wall and invades the breast tissues. Over time, it can also spread to the surrounding breast tissue and cause cancer to other human tissue parts. According to the American Cancer Society, about 80% of all breast cancers are due to the IDC-type tumors. So, the early detection of IDC tumors is crucial. Nowadays, Machine Learning (ML) and Deep Learning (DL) techniques are used to identify the IDC tumor due to their success in the medical image processing field. The ML algorithms require a handcrafted feature extractor method, which is time-consuming and challenging, while deep learning algorithms are segmentation-free approaches. This study applied the three well-known CNN pre-trained models, including ResNet50, VGGNet19, and DenseNet201, to extract microscopic image features. The results of this study indicated that the three pre-trained models, including ResNet50, VGG19, and DenseNet201, achieved 79.61%, 87.27%, and 96.55% accuracy classifying IDC tumors where the DenseNet201 outperformed the other models. Therefore, the proposed approach would be an outstanding candidate for the early detection of IDC.
