Content-aware image restoration: pushing the limits of fluorescence microscopy

There is large consent that successful training of deep networks requires many thousand annotated training samples. In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. We show that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks. Using the same network trained on transmitted light microscopy images (phase contrast and DIC) we won the ISBI cell tracking challenge 2015 in these categories by a large margin. Moreover, the network is fast. Segmentation of a 512x512 image takes less than a second on a recent GPU. The full implementation (based on Caffe) and the trained networks are available at http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net .

We report a water-based optical clearing agent, SeeDB, which clears fixed brain samples in a few days without quenching many types of fluorescent dyes, including fluorescent proteins and lipophilic neuronal tracers. Our method maintained a constant sample volume during the clearing procedure, an important factor for keeping cellular morphology intact, and facilitated the quantitative reconstruction of neuronal circuits. Combined with two-photon microscopy and an optimized objective lens, we were able to image the mouse brain from the dorsal to the ventral side. We used SeeDB to describe the near-complete wiring diagram of sister mitral cells associated with a common glomerulus in the mouse olfactory bulb. We found the diversity of dendrite wiring patterns among sister mitral cells, and our results provide an anatomical basis for non-redundant odor coding by these neurons. Our simple and efficient method is useful for imaging intact morphological architecture at large scales in both the adult and developing brains.