Spike trains emitted by retinal ganglion cells are statistically structured into a sequence of discrete collective modes, a new study demonstrates. The findings suggest that neurons found in the human eye naturally display a form of error correction in the collective visual signals they send to the brain.
Earlier studies have shown that neuron groups in most parts of the nervous system represent and process information in a collective fashion.
These patterns carry different information than can be relayed by any single neuron on its own. However, the details of this collective signaling are poorly understood.
The fact that individual neurons’ signaling is prone to corruption by noise puts an important constraint on any such collective processing. Studies have shown extensive information redundancy among neurons in many brain regions, which pointing the way to a theory of error correction.
Retinal Ganglion Cells
The study, from Jason Prentice of Princeton University, New Jersey, and colleagues, focused on retinal ganglion cells. These neurons, found in the back of the eye, receive information from light-detecting cells and relay it to the brain. The cells provided an good model to investigate neural group error correction with.
Multiple retinal ganglion cells can monitor the same visual region, and the researchers hypothesized that this redundancy could allow for error correction.
Researchers used visual stimuli to activate groups of about 150 retinal ganglion cells with overlapping visual regions and recorded their response. They used the data to build a mathematical model of retinal ganglion activity and analyze its patterns and organization.
Collective Neuron Signaling
The resulting model showed that a visual stimulus more reliably activates collective retinal ganglion signals than signals from individual cells. This suggests that collective activity allows for error correction and results in the transmission of more precise visual information, meanwhile suppressing background noise introduced by the irregular activity of individual cells.
The team developed a unique statistical model that decomposed the population response into modes. It predicted the distribution of spiking activity in the ganglion cell population with high accuracy.
They found that the modes represent localized features of the visual stimulus that are distinct from the features represented by single neurons. Modes form clusters of activity states that are readily discriminated from one another.
The new model is more accurate than previous models developed to investigate this collective neuron signaling behavior. It not only reveals new insights about retinal ganglion activity, but could also be applied to explore neural codes in the rest of the human brain, says study co-author Michael Berry.
Image: Activity in a population of 152 retinal ganglion cells responding to a natural movie clip. Each row is the firing rate, depicted with a color scale, of one cell versus time. Credit: Prentice et al.