Research Interests

retina circuit scheme

A simplified schematic drawing of the basic neural circuitry in the retina. Photoreceptors (red) take up light and pass on their signals to a network of horizontal cells (green), bipolar cells (blue), and amacrine cells (yellow). Finally, ganglion cells (purple) transmit the processed visual information through the optic nerve to different brain regions.

Our visual perception relies on sophisticated computations that are solved by our nervous system. We are usually unaware of the complexity of these computational challenges; they often only become apparent when the visual system is defective or — less severely — when it is fooled by a visual illusion. Among these challenges are: accounting for varying illumination conditions, partitioning an image into distinct objects, detecting motion signals, coping with eye movements, and constructing a three-dimensional representation of the environment.

Solving these tasks begins in the retina, an intricate network of nerve cells at the back of the eye. On one side of the retina, light is taken up by specialized receptor cells. On the other side, the nerve cells send out precisely timed electrical pulses ("action potentials"). All visual information that is available to the rest of the brain must be contained in this code that consists of patterns of action potentials.

We are interested in how the neural network of the retina processes visual information, represents it in its electrical activity and thus contributes to solving the various computational tasks of the visual system. To study the neural code of the retina and its circuit mechanisms, we combine electrophysiology and computational modeling.

photoreceptors

View through a microscope onto the photoreceptor layer of a salamander retina. The photoreceptors are densely packed and consist of the larger rods and smaller cones.

In the experiments, we project visual stimuli onto the photoreceptors of the retina while recording the action potentials of many retinal ganglion cells simultaneously with multi-electrode arrays. We also use whole-cell recordings with patch pipettes in order to assess the membrane potentials and currents inside individual neurons.

The computational models help us assess the relevant dynamics in the retinal network and provide a framework for the data analysis. They are combined with techniques from information theory and signal detection theory.

retina computation model

Schematic illustration of a computational model of signal processing in the retina. The retina in the eye converts visual images into spike trains in the membrane potential of ganglion cells. This conversion can be roughly modeled through a series of filters, transformations, and threshold crossing operations that contain both feedforward and feedback elements.

A particular interest of the group is the effect of natural eye movements on visual processing in the retina. Rapid shifts in the direction of gaze ("saccades") and small jitter and drift motion during fixation ("fixational eye movements") strongly affect the spatio-temporal stimulus patterns that fall onto the retina. We investigate how the retinal network copes with or makes use of these movement dynamics.