The big data challenges of connectomics

Abstract

The structure of the nervous system is extraordinarily complicated because individual neurons are interconnected to hundreds or even thousands of other cells in networks that can extend over large volumes. Mapping such networks at the level of synaptic connections, a field called connectomics, began in the 1970s with a the study of the small nervous system of a worm and has recently garnered general interest thanks to technical and computational advances that automate the collection of electron-microscopy data and offer the possibility of mapping even large mammalian brains. However, modern connectomics produces 'big data', unprecedented quantities of digital information at unprecedented rates, and will require, as with genomics at the time, breakthrough algorithmic and computational solutions. Here we describe some of the key difficulties that may arise and provide suggestions for managing them.

Figure 1

Segmenting brain images. (a) Shown is an electron micrograph of a small part of a 30-nm-thick section of mouse cerebral cortex. Even though the region shown is only 40 × 20 µm and less than 1,000th of the area of 1 mm², it contains the cross-sections of more objects than are practical to identify by the human eye. (b) Automatic computer-based methods, however, can attempt to completely segment such data (that is, saturate the segmentation). The result of the segmentation is shown as an image overlay with a distinct color (ID) for each cellular object. In subsequent sections (not shown), the same color is used for the same objects. (c,d) Higher magnification image of a small part of the automatically segmented image from a subsequent section (c) shows that the automatic method makes errors when compared to human tracing (d). The white arrow labeled S in c shows that a vesicle-filled axonal profile is split into several compartments, whereas the same axonal profile in d is labeled correctly as one object (compare the white arrow pairs in c and d). The black pair of arrows labeled M in c show two objects that have been merged into one by the automatic segmentation algorithm, whereas the objects are correctly labeled as separate in d (compare the black arrow pairs in c and d).

Figure 2

Transformation of segmented data into a connectivity graph. Left, in this schematic, the segmented axon of the dark gray neuron is found to establish four synapses on the dendritic spines of the light gray neuron (a–d). Middle, the segmented data is transformed into a layout graph that keeps notation of the location of every axonal and dendritic branch point and every synaptic junction. The layout graph has far less data than the segmented images that were used to generate it. Circles of different sizes and shading represent identification tags (IDs). Right, ultimately, the connectivity can be graphed quite simply, as shown here.