| dc.description.abstracteng | The retina is a complex neural network, responsible for breaking down
the visual scene into its distinctive features such as local contrast,
motion and color. The retinal ganglion cells form the output layer of
this network, and a typical vertebrate retina may contain more than
10 different ganglion cell types. These cells can be separated based
on their anatomical or physiological properties, and each type is
expected to relay information about distinct visual features to
specific areas in the brain. Separating these channels of information
is crucial for understanding how the visual scene is encoded, and much
effort is put into classifying retinal ganglion cells. From the different
strategies used to classify a ganglion cell, the physiological one -- based
on the responses of the cell to light stimulation -- may be the most
challenging, because physiological properties do not always discriminate
between different cell types. For the salamander, previous attempts to
classify retinal ganglion cells were based on their temporal filtering
properties, and were successful in separating ganglion cells into coarse
temporal response types. But surprisingly, only one of the types showed
tiling (a mosaic arrangement) of its receptive fields. Because tiling
is considered a strong signature of single cell types, I ask here whether
a refined classification is possible -- and whether it yields tiling by
further ganglion cell types. Spiking activity was recorded from isolated
axolotl retinas using multi-electrode arrays, and more than 200 cells could
be simultaneously recorded in a typical experiment. The retina was stimulated
with an uncorrelated noise (white-noise) stimulus, which was used to estimate
via reverse correlation the receptive field properties of the ganglion
cells. Together with the autocorrelation of the spike-trains, the receptive
field extent and temporal filtering properties were used to characterize the
ganglion cells. While a single property did not easily distinguish between cell
types, a spectral clustering algorithm was able to classify the ganglion
cells into putative types based on a combination of their properties. The
identified types were then matched across retinas. At least two tiling types
were consistently observed across retinas, with the remaining types showing few
violations of tiling. Cell types with similar physiological properties, whose
distinction would be blurred if analyzed within a single property, could be distinguished
by a combination of properties. The results suggest that salamander ganglion cells
can be classified when their physiological features are taken in tandem, and that
tiling is a fundamental feature of ganglion cells types -- also in the salamander
retina. | de |