(a) A 3D rendering of seven columns of the medulla of the Drosophila optic lobe from FIB-SEM showing reconstructed neurons from a ~ 30,000 µm3 volume. This reconstruction, which required over five man-years of effort, was ~3–5x faster than a comparable optic lobe reconstruction using an image stack from serial-section TEM, for which we have reconstructed a single medulla column (Takemura et al., 2013). Scale bar, 10 µm. (b) A cross-section of the neuropil of the medulla in the optic lobe of Drosophila, showing the high degree of reconstruction completeness that is possible with FIB-SEM data. The hexagonal periodicity reflects the hexagonal pattern of the ommatidia of the fly’s retina. The colors illustrate how all neural processes have been assigned. Green indicates various identified columnar input neurons contained within this volume, and yellow indicates axons and arbors of various medulla neurons that branch into or out of this volume. The small remainder (shown in red) highlights the ‘left over’ parts, including unidentified and orphaned fragments of neurons and glial processes. Well over 90% of the neuropil volume could be reconstructed and assigned to specific neurons. Scale bar, 1 µm. (c) Cross-section of the neuropil of the mushroom body of Drosophila. Notice that virtually all processes in this section have been identified and colorized green (to denote Kenyon cells) or yellow (for other identified mushroom body neurons). The only ‘left over’ uncoded processes are a few thin fragments dispersed within the mushroom body boundary that could not be confidently assigned to a specific cell. The mushroom body volume was comparable to the seven-column medulla volume and required a comparable reconstruction effort. Scale bar, 10 µm. Image process, segmentation, and 3D rendering provided by the Janelia FLYEM team, see Acknowledgements.