Enabling modern connectomics

Connectomics is giving scientists an unprecedented view into the brain. Building these detailed brain maps requires working with massive amounts of data. So Google Research has developed technologies to more efficiently process, analyze and share data. These technologies have enabled connectomics researchers to dramatically scale up progress on understanding the brain.

Here, you can read more about some of these technologies.

In September 2023, Google Research and collaborators embarked on a new project to map part of the mouse connectome. This animation shows the scale of this milestone by calculating the number of Pixel phones needed to store the data from past, present and future connectome projects. It would take just two Pixels, the height of an olive, to store the data from the first connectome – that of the C. elegans roundworm. In contrast, it would take a stack of Pixels as tall as a four-year-old girl to store the data from Google Research’s new partial mouse brain mapping project. And it would take a stack of Pixels the size of Mount Everest to achieve the next major goal in the field: mapping an entire mouse connectome.

Flood-filling networks

One of Google Research’s key technologies is the "flood-filling network", a recurrent neural network that traces the paths of neurons through three-dimensional representations of brain tissue. Flood-filling networks have already increased the accuracy of automated connectomics data interpretation by an order of magnitude over previous techniques. The figure below shows a flood-filling network in action as it traces a single neurite in 3D in a songbird brain.

A gif of the flood filling networks algorithm.


This image below shows all of the connections, or synapses, connecting one human brain cell to others. The cell is colored red, and synapses are shown with yellow or purple points. To build a human brain connectome, Google Research and collaborators mapped 130 million synapses between 50,000 cells. The project contains 1.4 petabytes of three-dimensional image data. In 2019, Google Research released a new, interactive tool called Neuroglancer for visualizing huge multidimensional datasets such as these. You can view images from the human brain connectome project in this Neuroglancer gallery.

Neuroglancer image


Google Research continues to use artificial intelligence to enable ever-larger connectome projects. In 2022, Google Research released a self-supervised learning technology, SegCLR, that can automatically extract key information from connectomics data. This work demonstrated that SegCLR can automatically identify the types of cells (e.g., pyramidal neuron, basket neuron, etc.) from surprisingly small pieces of those cells, as well as identify parts of each neuron (e.g., axondendrite, etc.), from the images and 3D reconstructions produced in connectomics projects.