Optimization of stimulation parameters for epi-retinal implant based on biosafety consideration

Optimizing stimulation protocol is essential for clinical application of retinal prosthesis. Elongating stimulation pulse width (~25ms /phase) has been proposed as an effective method to improve spatial resolution of epi-retinal implants. However, it is unknown whether longer stimulus pulse width will increase the risk of damaging the retina. In addition, with the advent of next generation retinal prosthesis featuring high-density microelectrode array, it is tempting to optimizing a single set of parameters for all electrodes instead of optimizing parameters of each electrode, but this approach raised biosafety concern. We sought to study the effect of stimulus pulse width on the response of retinal ganglion cells to electrical stimulation, and evaluate if the single parameter set approach was valid based on biosafety measures.

We stimulated mouse retina using biphasic pulse waveform generated by chosen electrodes (single or a 3x3 assembly) from multiple microelectrode arrays, recorded their action potentials and performed spike sorting. We tested various stimulus intensity with two fixed pulse width: a short one for 1 millisecond per phase, and a long one for 25 milliseconds per phase. All these assays were performed on two mouse models: the wildtype C57BL/6J mice and the photoreceptor degenerated rd10 mice. The action-potential-frequency vs stimulus amplitude profiles were plotted, and three parameters were extracted: the threshold (the lowest stimulus amplitude activating RGC units), safety-limit (stimulus amplitude that attenuated the firing rate to half of the maximum response), and the stimulation amplitude range (the difference between threshold and safety limit parameters).

In single-electrode stimulation experiment, we found that on average 85% of the recorded units showed attenuated response to extreme stimulation; among those units, an average of 51% stopped responding during stimulation ramping and failed to recover after one-hour post-stimulation, indicating extreme stimulation can damage RGC units. Twenty-five-millisecond pulse stimulation significantly reduced safety-limit and stimulation-amplitude-range parameters of recorded RGC units compared to 1ms pulse stimulation. During stimulus amplitude ramping, the maximum proportion of responsive healthy RGC units was 51% on average in 25ms pulse condition, and 76% on average in 1ms pulse condition, indicating long pulse may inflict more strain on RGCs, and a significant amount of inappropriately stimulated RGCs always exist. The contrast of these proportions could be explained by the tight correlation between the threshold and safety-limit parameter in 25ms pulse condition. These results were corroborated by those from 3x3 array stimulation experiments.

Base on a biosafety measure (RGCs' evoked firing rate in response to electrical stimulation), we proposed that longer stimulation pulse width could lead to reduced retinal response and thus highlighted the importance of carefully setting the stimulation amplitude in this case. Our results also suggested that optimizing a single set of parameters for all electrodes without individual tweaking always generated a significant amount of inappropriately stimulated RGCs, especially in the long pulse stimulation condition.