Analyzing the impact of generative adversarial networks for augmenting imbalanced brain MRI datasets
Tatsat Bhatnagar, Sivran Kohli · 2025
In medical imaging, machine learning model training has difficulties when the dataset is not balanced, especially when classifying brain tumors from magnetic-resonance imaging (MRI) data. This study looked into using Deep Convolutional Generative Adversarial Networks (DCGANs) as an alternative means for tackling this problem of imbalanced classes in brain MRIs. Noticeable enhancements have been noted in the classifier&s;s capacity to identify examples that are not tumours post-augmentation with synthesized pictures, as well as its learning stability as depicted in our study.