Enhanced cell tracking using a GAN-based super-resolution video-to-video time-lapse microscopy generative model

Cells are among the most dynamic entities, constantly undergoing processes like growth, division, movement, and interaction with their environment and other cells. Time-lapse microscopy is central to capturing these dynamic behaviors, providing detailed spatiotemporal information at single-cell resolution in real time. Although deep learning has transformed cell segmentation, cell tracking remains challenging due to limited annotated time-lapse data. To address this, we introduce tGAN, a generative adversarial network (GAN)-based time-lapse microscopy generator that enhances the quality and diversity of synthetic annotated time-lapse microscopy data. Featuring a dual-resolution architecture, tGAN accurately captures both low- and high-resolution cellular details essential for accurate tracking. Our results show that tGAN generates high-quality, realistic annotated time-lapse videos with high temporal consistency and fine details. Importantly, annotated videos generated by tGAN enhance the performance of recent cell tracking models, reducing reliance on manual annotations. tGAN enhances deep learning's impact on bioimage analysis, enabling more generalizable cell tracking models.