Imbalanced Cell-Cycle Classification Using Wgan-Div And Mixup

Classification of cell-cycle phases is required to determine the cellular changes and corresponding behaviours for diagnostic and prognostic research studies. One of the main challenges in cell-cycle classification is data imbalance caused by the different duration of each phase. In this paper, we present an imbalanced cell-cycle classification method that utilises Wasserstein divergence GAN and mixup for data augmentation to achieve over-sampling of the minority classes. We also optimised the standard random sampling strategy that allows the finetuning of the target distribution to improve the classification performance. Experiments on a public dataset of Jurkat cells captured by imaging flow cytometry show that our method achieves state-of-the-art performance.