Abstract
A conditional generative adversarial network (cGAN) approach to optimize resolution or contrast and field-of-view (FOV) in beam-scanned light sheet microscopy by beam shape translation is demonstrated. Large FOV images acquired with collimated (pencil) beam (weakly-focused) illumination are used to predict large FOV, but higher contrast, images mimicking Gaussian-beam (highly focused) illumination. In this way, imaging throughput and resolution is improved.

This publication has 3 references indexed in Scilit: