The goal of the Mouse Brain Architecture (MBA) project is to generate brainwide maps of inter-regional neural connectivity. These maps will thus specify the inputs and outputs of every brain region, at a “mesoscopic” level of analysis corresponding to brain compartments defined in classical neuroanatomy.

To determine the outputs of a brain region (i.e., the other regions it projects to), anterograde tracers are used which are taken up by neurons locally (“the input”), then transported actively down the axons to the “output regions”. The whole brain is then sliced thinly, and each slice is digitally imaged. These 2-D images are reconstructed in 3D. The majority of the resulting 3-D brain image is unlabeled— only the injected region and its output regions have tracer in them, allowing for identification of this small fraction of the connectivity map. This procedure is repeated identically, to account for individual variability.

To determine the inputs to the same brain region as above, a retrograde tracer is injected in the same stereotaxic location (“the input”), and the process is repeated. In order to accumulate data from different mice (each of whom has a slightly different brain shape and size), 3-D spatial normalization is performed using registration algorithms.