First I want to explain fixational eye movements, then formulate my research question and research strategy. I will briefly sketch the experimental setup in case you don‘t remember all the details from Vidhya‘s Friday seminar talk.
First I want to briefly explain fixational eye movements. The picture to the right illustrates these eye movements as measured for a human eye. (Click) Fixational eye movements are a constant feature of normal vision, present even when „fixing“ the gaze on an object – hence the name “fixational eye movements“. Importantly, the visual perception fades away when the image of an object is stabilized artificially on the retina. Enhancement of spatial details and improvement of stimulus feature estimation: distinquish between grating orientation… So these movements are an integral part of normal vision.
To quantify the effects of these eye movements on neuronal coding, Martin Greschner and colleagues used video-oculography on turtles. They isolated a periodic component (Click) at approximately 5 Hz. The corresponding amplitude was relatively small, namely 5µm on the retina, which corresponds to the diameter of a photoreceptor. The authors concluded that ... This could potentially improve stimulus feature estimation by the brain.
First point: How should the brain interpret the responses? Greschner found synchronized firing -> population code
My research strategy derives from the fact that fixational eye movements result in oscillatory shifts of the image on the retina. Imitate these eye movements by a shifting black-white border. Green ellipses denote the receptive fields of 2 ganglion cells; blue arrow shows the shifting border orientation
Each dot represents a spike. First idea: work on single cell level, use e.g. spiking rate. But rate approximately constant for both stimuli. Have to use a much higher amplitude as compared to Martin Greschner to elicit any responses.
Quantify these correlations by a histogram plot of relative spike timings. Spike timing cross-correlations can provide information about the stimulus; relative spike timings are different for the two stimuli
So I created these histogram plots for measured spike timings, as shown in the figure; the 5 angles are color-coded. One observes that the correlations are periodic, showing the same period as the stimulus. Green curve: peak at zero: cells tend to spike synchronously. Red curve: delay.
Apply information theory... Upper right: „different patterns for different stimuli?“
Quantify population responses by information theory measures. Imutual: stimulus: 5 different angles, spike patterns; either of single cells or cell pairs. Synergy can be positive or negative.
Imutual (stimlocked) > Imutual (unlocked), and „additional information“ in stimulus locked case smaller
pptx - TAC Meeting
Neuronal Coding in the Retina
and Fixational Eye Movements
Friday Seminar Talk
November 6, 2009
Tim Gollisch Lab
• Experimental setup
• Review of fixational eye movements
• Research questions and strategy
• A look at the observed data
• Spike timing cross-correlations
• Information theory: entropy, mutual
information, synergy, ...
• Summary and outlook
Fixational Eye Movements
source: Martinez-Conde laboratory
• Constant feature of normal
• Visual perception fading
• Enhancement of spatial
Riggs LA and Ratliff F. The effects of
counteracting the normal movements
of the eye. Journal of the Optical
Society of America (1952)
Ditchburn RW and Ginsborg BL. Vision
with a stabilized retinal image.
Meister M, Lagnado L and Baylor DA.
Concerted signaling by retinal
ganglion cells. Science (1995)
Martinez-Conde S et al.
Microsaccades counteract visual
fading during fixation. Neuron (2006)
Fixational Eye Movements II
Eye movements of the turtle during fixation
• Periodic component at approximately 5 Hz
• Imitating fixational eye movements →
retina better encoder
• Neurons synchronize more
Greschner, Ammermüller et.al.
Nature Neuroscience (2002)
• How can the brain discriminate
between various stimuli in the context
of fixational eye movements? Optimal
• Synchronized responses of several
retinal ganglion cells → population
Concrete task: based on spike
responses, discriminate 5
Spike Timing Cross-Correlations II
Encoding the Spike Train
Encoding spike patterns
→ observer knows
the stimulus phase
Mutual information Imutual
→ How much information („bits“) do the
spikes contain about the stimulus
→ How much additional information is contained in
the simultaneous activity of two cells as compared to
the individual cells’ responses
Population Code: Synergy
Synergy versus mutual information
for several recordings
• Fixational eye movements provide
information about the stimulus
• If the brain uses individual cells, it needs
to know the phase of the fixational eye
• For multiple cells, the phase information
becomes less important since the cells
• Effect of shorter stimulus periods
and smaller amplitudes?
• Try different decoding stategies:
optimal patterns, bin sizes?
Tim Gollisch Lab
• Tim Gollisch
• Daniel Bölinger
• Vidhya Krishnamoorthy
Thesis Advisory Board
• Tim Gollisch
• Erwin Frey (LMU)
• Andreas Herz
• Günther Zeck
Boehringer Ingelheim Fonds
Foundation for Basic Research in Medicine