Logo
DeutschClear Cookie - decide language by browser settings
Bölinger, Daniel (2012): Closed-loop experiments to investigate spatial contrast integration in the retina. Dissertation, LMU München: Faculty of Biology
[img]
Preview
PDF
Boelinger_Daniel.pdf

4Mb

Abstract

The fundamental goal of all neuronal processing is to make optimal decisions, and thereby to generate optimal behavior. To this end, the brain performs at each point in time millions of parallel computations. Right now, your brain might weigh thoroughly if it is worth to continue reading this thesis or not. In the end however, weighing is not enough: a decision has to be made. Such a decision is an intrinsically nonlinear process. If two possibilities are nearly equally evaluated, a small change in your considerations might lead to the acceptance of one alternative, and the rejection of the other. Furthermore, not all considerations have to contribute linearly to the decision. For example, a lack of time might certainly keep you from reading this thesis, while having excess spare time might still not make you read it. Such a nonlinear weighing of considerations does not only occur on the conscious level, but all the time in the individual neuronal circuits of the brain. Each neuron can be interpreted as a decision unit, computing whether to spike or not. It typically receives multiple parallel streams of information, and based on these generates its own neuronal output. The inputs resemble the considerations taken into account, while the output conveys the decision to subsequent circuits. How the decision is made is therefore determined by the way the inputs are combined into the neuronal output. In particular, individual inputs might contribute either linearly or nonlinearly to the decision. Thus, in order to understand which role a neuron plays in information processing, we have to assess the nonlinearities involved in the integration of different neuronal inputs. In this thesis, we study this ubiquitous signal integration in the output neurons of the amphibian retina, the retinal ganglion cells. Thereby we hope to gain a better understanding of the general mechanisms underlying signal integration in the circuits of the brain. This will also help us elucidate the functions of the retina in particular. Because of the high similarity of the retinas among all vertebrates, by studying the amphibian retina we also learn to better understand human vision. The amphibian retina is particularly suited to study the nonlinear integration of neuronal signals, because each single ganglion cell receives distinct inputs originating from tens to hundreds of photoreceptors. Indeed, ganglion cells do not just linearly average these inputs, but combine them in a nonlinear fashion. It turned out that it is precisely this nonlinearity which allows specific ganglion cells to decide whether particular features were present in their visual input. Hence, an understanding of how the retina encodes images into neuronal activity requires an understanding of how the spatially distinct light stimuli, that each cell experiences, are combined into the output of this very cell. This is the question of spatial integration which we address in the following. Many facts about this question are already available on a cellular level. Today we know which cell types mediate the signals from the photoreceptors to the ganglion cell, and we know much about the connections between the involved cells. Furthermore, in recent studies the transmission functions of some of the involved circuit elements were measured. In particular, it turned out that many of the processing steps are highly nonlinear. Although all these details are known, a detailed phenomenological description of spatial integration is still lacking. Most current models assume a linear integration, and thereby simply neglect the nonlinearities occurring on the cellular level. In this thesis, we attempt to fill the gap and strive for a functional characterization of spatial integration, and in particular of the involved nonlinearities. We pursued the investigation by performing electrophysiological experiments on retinal ganglion cells. In particular, we measured the neuronal output with an array of electrodes. While measuring, we presented videos containing well-defined light stimuli to the retina. We performed the experiments in a closed-loop approach which allowed us to assess the neuronal response online and use the results to determine the subsequently shown stimuli. The visual area, over which a ganglion cell pools its input, is called the receptive field of the cell. It has been known for almost 60 years that the receptive fields of many ganglion cells are organized in a center-surround structure. In the receptive field center, the cell is most sensitive to visual stimulation. Depending on the cell, it preferentially responds to either a brightening (ON cell) or a darkening (OFF cell) of the image. In contrast, the response in the receptive field surround is weaker, and it is of opposite sign than the center response. Taking this structural segregation of the receptive field into account, we divided our experiments into two parts. First, we determined how different stimuli are combined within the receptive field center. Afterwards, we focused on the integration of stimuli in the center and the surround. Throughout this thesis, we used a specific approach to study spatial integration. This approach is the measurement of so-called iso-response stimuli. Instead of showing predefined stimuli and measuring the neuronal outputs, we did the experiments the other way round: we predefined the output, and then searched for those stimuli which yielded the chosen response. The result of such a measurement was a set of stimuli which all triggered the same neuronal response in a given ganglion cell. Thereby, the cell’s response was either defined as the number of elicited spikes (iso-rate stimuli), or the first-spike latency (iso-latency stimuli). Iso-response stimuli allowed us to directly assess the nonlinearities involved in signal integration in retinal ganglion cells.