An international team of scientists has presented a new circuit with the functions of brain cells. It will help improve the performance of AI-based models.
Neurologists and computer scientists at Princeton University, the Allen Institute, and Baylor College of Medicine have published a three-dimensional electrical circuit that works with the functions of tens of thousands of neurons. This development makes it possible to study in detail the organ of most mammals.
The dataset, which is publicly available for viewing and use by everyone, shows the structures and connections of 200 thousand brain cells and almost 500 million synapses contained in a cubic millimeter of the mouse brain. They are from the neocortex, the part of the mammalian brain that processes what the eyes see.
“Our five-year mission was truly difficult, with an ambitious goal that many considered unattainable. In the first year of operation, one of the laboratory staff argued that even a pilot phase would not be possible. Today we are releasing a cubic millimeter of reconstructed mouse cerebral cortex, which is a thousand times larger than the goal of the first phase, ”the researchers noted.
The large-scale project “Machine Intelligence Based on Cortical Networks” (MICrONS) has been in development for five years and was funded by the Intelligence Advanced Research Projects Activity. The aim of the study is to obtain information about brain conductors to improve machine learning. However, the dataset is also valuable for neuroscience – both for scientists looking to understand how the brain transmits information along certain circuits and for biomedical researchers looking to treat brain diseases.
MICrONS data contains the largest number of cells and connections among all similar sets. It is so large that it can cover entire local circuits and almost complete 3D-shapes of individual mouse neurons. Some neurons make connections over incredibly long distances, sending signals throughout the brain. The scientists chose a cubic millimeter volume in order to cover circuits in several areas of the brain involved in vision, and at the same time convey the structure of as many whole neurons as possible.