CLICK ON weeks 0 - 40 and follow along every 2 weeks of fetal development
How the brain recognizes what the eye sees
Salk Institute researchers have now analyzed how neurons in a critical part of the brain, called V2, respond to natural scenes. They feel they have reached a better understanding of vision processing and describe their work in Nature Communications on June 8, 2017.
"Understanding how the brain recognizes visual objects is important not only for the sake of vision, but also as it provides a window on how the brain works in general.
Although we take our ability to see for granted, "seeing" is made up of complex mathematical formulations still being working out. Sharpee and team are not able to parse these formulations into actually viewed scenes as yet.
Human visual perception starts with the eye perceiving light and dark signals. These signals are bounced off the back of the brain in an area called V1. At the V1 or Primary Visual Cortex, light and dark signals reflect the depth of "edges" of whatever scene is being viewed. Multiple views of any one scene are received and then super imposed over each other in layers. Our brain then begins to distinguish faces from cars and other objects, as well as which of these layered objects is moving. How quickly and precisely this recognition happens is still being unraveled. Neuronal responses are encoding every edge of an object in every "scene" viewed, before reconstructing an image. All within seconds — it's complicated.
Now, Sharpee and Ryan Rowekamp, a postdoctoral research associate in her group, have developed a statistical method to descriminate between the signals being received and decoded — to establish something approaching computer-simulated vision. To develop their model, the team used publicly available data from primate brain scans. The scans were made while primates watched movies of nature, such as rain forest landscapes; all collected into a database stored at the Collaborative Research in Computational Neuroscience (CRCNS).
"We applied our new statistical technique to distinguish which features in the movie were causing V2 neurons to change. Interestingly, we found V2 neurons respond to combinations of edges."
The team found V2 neurons process visual information according to three principles:
First, V2 neurons combine edges with similar orientations, responding strongly to small changes in the position of curves at an object's boundaries.
Second, if a V2 neuron is activated by the edge of a particular orientation or position, then any orientation 90 degrees from that same location will be suppressed. This is called "cross-orientation suppression." Cross-oriented edge combinations allow for recognition of a variety of visual cues essential in shape detection.
Third, the "principle of pattern repeats" — or the increased perception of textures, as found on trees, water or the boundaries between the two, are deciphered much as one would when viewing an impressionist painting. Complex textural boundaries are being distinguished through changes in paint strokes, by volume and angle, as well as light and dark intensities.
"It was really satisfying when combining edge recognition with sensitivity to texture... as it started to pay off as a tool for analyzing and understanding complex visual data," explained Ryan Rowekamp.
An immediate application for this research might be the improvement of object-recognition algorithms for self-driving cars, or other robotic devices — or perhaps in the viewing of the galaxy. "Every time we add elements of computation that are found in the brain to computer-vision algorithms, algorithm performance is improved," adds Sharpee.
Discretization in neural circuits occurs on many levels, from generating action potentials and dendritic integration, to neuropeptide signaling and then the processing of signals from multiple neurons, to behavioral decisions. It is clear that discretization, when implemented properly, can convey many benefits. However, the optimal solutions depend on both the level of noise and how it impacts a particular computation. This Perspective discusses how current physiological data could potentially be integrated into one theoretical framework based on maximizing information. Key experiments for testing that framework are discussed.
Keywords: information theory, dendrites, neuromodulation, neural cell types, neuropeptide, ion channels, ionic currents, power law, scale-free dynamics, phase transitions
The work was funded by the National Science Foundation.
About the Salk Institute for Biological Studies:
Every cure has a starting point. The Salk Institute embodies Jonas Salk's mission to dare to make dreams into reality. Its internationally renowned and award-winning scientists explore the very foundations of life, seeking new understandings in neuroscience, genetics, immunology, plant biology and more. The Institute is an independent nonprofit organization and architectural landmark: small by choice, intimate by nature and fearless in the face of any challenge. Be it cancer or Alzheimer's, aging or diabetes, Salk is where cures begin. Learn more at: salk.edu.
Return to top of page
In a critical area of the brain – V2 – neurons respond to natural scenes,
giving researchers a better understanding of vision processing. The work
could improve self-driving cars and lead to therapies for sensory treatment.
Image Credit: Salk Institute.