Functional Decomposition of Retinal Ganglion Cell Receptive Fields
Doctoral thesis
Date of Examination:2023-01-25
Date of issue:2023-02-23
Advisor:Prof. Dr. Tim Gollisch
Referee:Dr. Jan Clemens
Referee:Prof. Dr. Hansjörg Scherberger
Referee:Prof. Dr. Alexander Ecker
Referee:Dr. Viola Priesemann
Referee:Dr. Caspar Schwiedrzik
Files in this item
Name:SorenZappDissertation.pdf
Size:4.85Mb
Format:PDF
Abstract
English
The retina has the fascinating ability to extract various visual features from our surroundings. This feature detection is enabled by nonlinear operations in the retinal circuitry. The nonlinear processing is in part attributed to retinal ganglion cells and to their signal integration of the presynaptic circuitry. Of particular interest are nonlinear subunits in the receptive field of ganglion cells that are thought to correspond to the excitatory inputs of presynaptic bipolar cells. As the access to bipolar cells is experimentally limited, subunits are typically inferred computationally from ganglion cell responses. However, current computational methods are demanding in both time and computational hardware. This prevents them from being applied easily in routine analyses. Based on an extensive review of the current landscape of subunit analysis and its challenges, I developed computational methods that recover subunits without high demands for time or computational power. With the development of tools for hyperparameter selection, the methods identify subunits in different ganglion cell types from various species under variable experimental conditions. Furthermore, I conducted multielectrode-array recordings in the isolated marmoset retina to investigate the spatial arrangement of subunits using the developed methods. The population analyses of parasol and midget ganglion cells reveal that their subunits are arranged in distinct mosaics that tile the visual space. The subunit mosaics of ON and OFF midget cells align spatially, as do subunit mosaics of ON and OFF parasol cells. Facilitated by the methods, these findings suggest that the systematic spatial coordination of ON and OFF channels can be traced back computationally to the level of bipolar cells. Together, this thesis provides computational methods of subunit inference that overcome prevalent challenges and demonstrate new insights into the nonlinear signal integration in the primate retina.
Keywords: retina; receptive field; retinal ganglion cell; nonlinear spatial integration; multi-electrode arrays; matrix factorization