GAN-Augmented Breast Cancer Classification with Histogram-Based Gradient Boosting

The main method for screening for breast cancer is mammography, which aims to lower the risk of breast cancer death by early detection. Deep learning techniques have been found to show good predictive results with many medical image datasets. Even so, extensive data augmentation approaches are necessary for accurate computer-assisted diagnosis due to the scarcity of annotated medical images. After being trained on photos, Generative Adversarial Networks (GANs) can produce new images that resemble real images and have a lot of genuine features. The realism and representativeness of the generated features can vary depending on the factors like quality of training data, training parameters, evaluation metrics, and application domain. We used a Histogram-Based Gradient Booster classifier to improve the traditional deep learning methods and encourage the use of GANs for data augmentation. When compared to the baseline, the results demonstrate that GANs increased recall without lowering precision and enhanced classification performance on the malignant class without sacrificing performance on the benign class.