Machine learning for image restoration

Fluorescence microscopy usually involves a trade-off between producing a quality image and having a healthy sample. Illuminating the sample with higher laser power strengthens the fluorescent signal but risks damaging biological samples and photobleaching fluorescent dyes. Imaging at a slower frame rate with lower laser power often produces high-quality images but sacrifices information in samples that move. When such compromises hinder the recording of high-quality images, researchers often try to improve the images after the fact. To that end, Loïc Royer at the Chan Zuckerberg Biohub in San Francisco and Martin Weigert, Florian Jug, and Eugene Myers at the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden, Germany, have developed content-aware image restoration (CARE), a convolutional neural network trained on features specific to the system being observed.