Understanding how visual information is processed in the brain

Understanding how visual information is processed in the brain

Information processing in the human brain involves the electrical firing of about 86 billion neurons that make trillions of connections among each other. The secrets of how the brain lets us think, feel, and act lie hidden in the complexity of this wiring diagram and the network of electrical signals that move across it in milliseconds.

Recent maps of neurons and their connections have offered new insights into the mysteries of the brain. Research funded by NIH's Brain Research Through Advancing Innovative Neurotechnologies® Initiative, or BRAIN Initiative®, have made great strides in mapping the mouse brain, common fruit fly brain, and human brain.

In a major step forward, researchers in the Machine Intelligence from Cortical Networks, or MICrONS, Consortium mapped the connections between hundreds of thousands of cells in a tiny region on the surface of the mouse brain involved in visual processing. The analyzed region is one cubic millimeter, about the size of a grain of sand. The scientists overlayed this map with neuron firing patterns in response to visual stimuli. The results appeared in a package of eight papers published in Nature and Nature Methods on April 9, 2025.

The researchers first showed video clips to a mouse whose neurons emitted light when firing. The team recorded these neuron firing patterns. Using this technique, they were able to measure the activity of nearly 76,000 neurons.

To map the connections between cells, the researchers cut nearly 28,000 ultra-thin slices of the brain tissue and imaged these using electron microscopy. Reconstruction required stitching together the images to align the connections across the entire volume. This was followed by months of tracing connections using deep learning artificial intelligence (AI) algorithms, followed by manual and automated proofreading.

This small area, the scientists found, holds four kilometers of axons, the long fibers that nerve cells use to communicate with each other. These are intertwined to make more than half a billion connections, called synapses, across more than 200,000 cells. The cells that were cataloged include different types of neurons as well as various support cells. The team showed that connectivity can be used to identify cell types that can be difficult to identify by shape alone.

Deep learning predictive models were constructed and validated to match neuronal activity to physical connectivity. The resulting map is the most comprehensive set of data ever assembled linking brain structure to neuron function in an active mammal. It provides new insights into how the brain processes visual information. Notably, the general blueprint of the brain cortex is largely the same across mammals, from mouse to human.

"While the current findings focus on a tiny fraction of the brain, they reveal the complex connections between the cells and show how those connections are wired to produce functional responses," explains Dr. John Ngai, director of the BRAIN Initiative. "This information, which was previously beyond our reach, could help us understand how the brain functions normally and offer a guide to what goes wrong as the result of various disorders or injuries."