Pix2Pix Generative adversarial Networks (GAN) for breast cancer detection
Zainab Aizaz, Kavita Khare, Afreen Khursheed, Aizaz Tirmizi · 2022
Breast cancer remains to be one of the most common cancers throughout the world. Detection of mammographically occult breast cancer is a trivial process due to hidden tumours in women with dense breasts. Data augmentation techniques are in demand due to existing paucity in the dash images that are labeled. In this brief, a model for the detection of mammographically occult breast cancer is developed with pix2pix Generative Adversarial Networks (GAN) as in imperative tool for data augmentation. In this work, GANs are used to construct transformations of original images in the dataset. The U-net based of CNN layers is used to preserve long-range connections in wide neighbourhoods to provide a balanced partitioning. For the cancer detection, this model can be used as an individual layer in a Convolutional neural network (CNN). Upon comparison with transfer learning approach, the proposed approach has higher accuracy for two VGG-16 and Resnet-50 CNNs.