The project will strive to perform deep learning-based, ultrafast imaging of living samples with minimal photon budgets to reduce phototoxicity and preserve sample integrity. The simultaneous requirements of low-intensity excitation and high imaging speed necessitate the development of specialised post-processing methods for image restoration and denoising. We will develop new machine learning based algorithms to retrieve object data from low light level, noisy images. The proposed hardware and software will be used to investigate a novel phenomenon recently discovered in the host group, namely the peristaltic pumping of protein content within the luminal network of the endoplasmic reticulum. The research requires imaging at sub-diffraction resolution at the lowest possible light levels, since the ER network dynamics react sensitively to light. Furthermore, peristaltic movement in the ER and concomitant protein flows within it occur on fast temporal scales, and we need to acquire images at ca. 40 frames per second or higher – current imaging modalities are not capable of addressing this problem. PhD Supervisors: Prof. Clemens Kaminski- Department of Chemical Engineering and Biotechnology and  Prof.  Pietro Lió- Department of of Computer Science and Technology

Research Topic: Deep learning for reconstruction of nanoscopy image data