Receptive Fields from Epi-retinal Recordings in Cats Give Hints for Optimizing Epi-retinal Implants for Blinds

Marcus Wilms, Thomas Schanze, Reinhard Eckhorn
Neurophysics Department, Philipps-University Marburg, Germany


Blinds with receptor degeneration can perceive localized phosphenes with epi-retinal focal electrical stimulation [1]. Recent work on the spatial resolution achievable with epi-retinal stimulation in anesthetized cats was done by assessing activity distributions in cortical areas 17 and 18 [2, 3]. They can be related to the position and form of phosphenes generated by epi-retinal implants. These cat experiments revealed a high variance in position and form of the cortical activation profiles that could not be explained by the retino-cortical mapping. Here we analyzed the reasons for this variance in order to improve the stimulation technique and hence, provide the chance for regular high-resolution perceptions with retina implants.


We used epi-retinal recordings because it is highly probable that the neurons whose signals are recorded by an electrode are mainly the same ones that can be activated by this electrode. We recorded local field potentials (LFP, 2-140 Hz) and spikes in anesthetized cats with multiple fiber-electrodes. The signals were separated by direct hardware- or offline-filtering from broad-band signals (2-10,000 Hz). Spikes were separated by offline threshold discrimination. Visual classical receptive fields (RFs) were analyzed using a multi-focal visual stimulation approach.


The positions and sizes of the visual RFs were generally distinct for LFP and spikes (N=15). While the center positions of the RFs based on LFP closely resembled the retinal electrode positions and were concentric, spike-RFs from the same electrode were often multimodal, i.e., different spike trains sorted from the same recording site had different RF-positions. They were often shifted distally but never proximally with respect to the optic disk. The connection lines between corresponding LFP- and spike-RFs were oriented parallel to the axonal fibers at this retinal recording location.


The well localized LFP-RFs probably reflect mainly postsynaptic dendritic potentials at the parallel dendrites of bipolar cells and the somata of ganglion cells. This is probable because the graded potentials are well localized and correspond to the actual tip position of the recording electrode. The dislocations of spike-RFs, however, are indicative of the recording of spikes from axons originating more distally from the recording site with respect to the optic disk. Distally generated axon-spikes are detected when they pass the retinal recording site en route to the optic disk. Multi-modal spike-RFs are therefore indicative of recordings from several axons at once.


Our study of simultaneously recorded LFP- and spike-RFs with epi-retinal electrodes indicates that one has to cope with ambiguous retinal and cortical activation, and hence with spatially ambiguous phosphenes with the presently developed epi-retinal implants. However, this may be avoided by a quasi-intra-retinal activation with protruding epi-retinal electrode tips that impinge on the retina. We will address this issue in our further investigations.

[1] Humayun, MS et al. (1999) Vis Res 39:2569-76
[2] Wilms, M (2001) PhD Thesis, University Marburg, Germany
[3] Schanze, T et al. (2002) GACEO 240:947-54
Support by BMBF grants 01 IN 501 F and 01 KP 0006 to RE is greatly acknowledged.