Berning, M. et al. Neuron 87, 1193–1206 (2015).

The reconstruction of neural circuits from electron microscopy data sets is a notoriously slow and labor-intensive process. Berning et al. developed SegEM to speed up the segmentation of volumetric images. As a machine learning–based software tool, SegEM depends on a segmented training data set, which can either be generated manually or be reused from previous studies. In addition, SegEM requires a skeleton reconstruction of the data set to be segmented. Both prerequisites can be achieved with manageable manual input. Using SegEM, Berning et al. reconstructed mouse retina and cortex data sets 10–100-fold faster than with other methods, with high accuracy.