The human brain is one of the most complex structures in existence, and Google, in cooperation with Harvard University, has released a reconstruction of 1.4 petabytes browsable for a very small part of the cerebral cortex, according to aitnews.
The dataset consists of imaging data covering approximately one cubic millimeter of brain tissue.
It includes tens of thousands of reconstructed neurons, millions of neuron segments, 130 million annotated synaptic points, and 104 corrected cells, along with many additional annotations and substructures.
The cerebral cortex is a thin, superficial layer of the brain, and this cortex plays an important role in thinking, memory, planning, cognition, language, and attention, as well as most other higher-level cognitive functions.
Although some progress has been made in understanding the visual organization of this highly complex tissue. However, its regulation at the level of individual neurons and their interconnected synapses is largely unknown.
This sample was donated anonymously from patients who underwent surgery for epilepsy at MGH, Massachusetts General Hospital, and then presented to researchers at Harvard’s Lichtman Laboratory.
Google and the cerebral cortex:
Using an automated tape that assembles a high-resolution microtome, Harvard researchers cut tissue into about 5,300 individual sections of 30 nanometers, then superimposed those sections on silicon wafers, and imaged brain tissue at a resolution of 4 nanometers with a 61-beam parallel-scanning electron microscope. Intended for a quick image.
The end result was 225 million individual 2D images, which Google then mathematically correlated and aligned in 3D volume with the thousands of Cloud TPUs utilized in the process.
This map of the human brain can now be accessed through Google’s web-based visualization tool Neuroglancer.
The future challenge for Google is storage, as the current 1.4 petabyte dataset is only one in a million of the size of an entire human brain.
For example, mapping the brain of a mouse could generate exabytes of data, and the company is looking into machine learning-based efforts to compress the data.
This is the largest sample of brain tissue ever imaged and reconstructed in this level of detail.
It also represents the first large-scale study of synaptic connectivity in the human cortex that extends across multiple cell types across all layers of the cortex.
The primary goals of this project are to produce a new resource for the study of the human brain and to improve and extend basic connectivity techniques.