Abstract
Abstract: :
Purpose: Ganglion cells fire synchronously when responding to the same bar of light, and the degree of synchrony increases with the intensity of the stimulus. Here, we ask whether the intensity information encoded by degree of synchrony would be confounded by the stimulus coordination produced by a moving bar and whether information encoded by synchrony could contribute to feature detection. Method: A computer model of the mammalian inner plexiform layer was developed based on anatomical and physiological data. The model was stimulated with narrow moving bars at several different speeds and intensities. To evaluate the efficiency with which the stimulus intensity was encoded by either the firing synchrony or firing rate, event trains from a column of eight model ganglion cells activated by the moving bar were summed into a threshold detector with a short integration window. Results: Peri-stimulus time histograms (PSTHs) obtained from the model resembled published physiological data. Synchrony due to synaptic interactions produced narrow peaks in the correlation functions, but synchrony due to stimulus coordination produced much broader peaks. Input from the model ganglion cells resulted in a two-fold lower detection threshold than input from independent Possion event trains having identical PSTHs. The additional sensitivity to correlated input was present at most stimulus intensities, but it was particularly pronounced at higher intensities and for bars with moderate velocities. Conclusion: Correlations reduce the amount of information that can be extracted by averaging the firing rates of multiple ganglion cells, but the same information can be encoded by the degree of firing synchrony. In single trials, this information can be readily extracted by a threshold-based coincidence detector. We suggest that coding of information by firing synchrony may contribute to motion processing in the mammalian visual system.
Keywords: 364 computational modeling • 559 retinal connections, networks, circuitry • 429 image processing