The group of Florian Jug at the MPI-CBG in Dresden developed a new training scheme for deep neural networks to be able to denoise images without the need of image pairs. So all you need is your (noisy) experimental data!
For the full story read here.