Generation and Simulation of Yeast Microscopy Imagery with Deep Learning
This work addresses the problem of reducing experimental effort in synthetic biology by modeling microscopy imagery, but it is incremental as it builds on existing deep learning methods for image generation and simulation.
The study tackled modeling time-lapse fluorescence microscopy experiments by generating synthetic yeast cell images using a novel GAN and simulating brightfield microscopy sequences with a future frame prediction model, achieving results that indicate deep learning is a promising but incomplete approach for real-world experiments.
Time-lapse fluorescence microscopy (TLFM) is an important and powerful tool in synthetic biological research. Modeling TLFM experiments based on real data may enable researchers to repeat certain experiments with minor effort. This thesis is a study towards deep learning-based modeling of TLFM experiments on the image level. The modeling of TLFM experiments, by way of the example of trapped yeast cells, is split into two tasks. The first task is to generate synthetic image data based on real image data. To approach this problem, a novel generative adversarial network, for conditionalized and unconditionalized image generation, is proposed. The second task is the simulation of brightfield microscopy images over multiple discrete time-steps. To tackle this simulation task an advanced future frame prediction model is introduced. The proposed models are trained and tested on a novel dataset that is presented in this thesis. The obtained results showed that the modeling of TLFM experiments, with deep learning, is a proper approach, but requires future research to effectively model real-world experiments.