Introduction

Patients who have been blinded by retinal degenerative diseases, such as age-related macular degeneration (AMD) or retinitis pigmentosa (RP), can regain functional vision with the help of retinal implants1,2,3. The principle stems from early research that discovered the survival of the majority of cells in the inner nuclear cell layer and the residual ganglion cells in the retina of patients with retinal degeneration such as RP. This finding illustrates the potential for vision restoration, bypassing damaged photoreceptors, by directly stimulating the surviving inner retinal neurons1,2,3,4,5. Clinical studies on retinal implants suggest that stimulating undamaged neurons in the retina could restore partial vision, even years after the onset of complete blindness in RP patients6,7. The previously marketed epiretinal prostheses, Argus II (Second Sight Medical Products Inc.) and IRIS II (Pixium Vision SA), as well as the subretinal prosthesis Alpha AMS (Retina Implant AG), have been discontinued2. Several technologically advanced retinal prostheses are currently being evaluated for clinical use. These include the NR600 epiretinal prosthesis by Nano Retina, Rainbow Medical Group, and the PRIMA subretinal prosthesis by Pixium Vision SA8,9,10. Additionally, there are ongoing clinical trials for two suprachoroidal prostheses in Australia by Bionic Vision Technologies11 and in Japan by Osaka University12indicating a possible turning point in the development of visual prostheses. Patients with retinal degeneration, aided by retinal prostheses, have been shown to perceive spots of light, recognize objects and letters, and even improve their orientation and mobility. Furthermore, long-term implantation and electrical stimulation of these prostheses have been demonstrated to be relatively safe2,13,14,15,16,17,18.

Despite showing promising preliminary clinical results, retinal prostheses are still limited by several factors that constrain their overall performance, including rudimentary resolution for critical vision2. These limitations are influenced by various factors, such as the number and size of electrodes in a multielectrode array (MEA), the distance between the electrode surface and the retina, and the type of stimulation used2,19. A clinical study focusing on retinal prostheses revealed that, in a sample of 30 patients equipped with these prostheses, only 55% of the electrodes could elicit visual perceptions when utilizing a stimulus fell below the charge density limit for the electrode18,20. Therefore, there remains a clinical need to optimize the stimulation strategy of retinal implants to enhance efficiency and improve the performance evoked in the visual cortex. Several preclinical studies indicate that neural response amplitude and power efficiency can be influenced by various stimulus parameters, such as phase duration and waveform shape19,21,22,23,24,25,26. However, only a few previous studies have examined the in vivo electrophysiological response to epiretinal retinal electrical stimulation at the cortical level25,27,28,29,30. Cortical activity maps demonstrate retinotopy in normal rats when epiretinal stimulation is applied to specific quadrants of the retina, but this phenomenon is not observed in rats with retinal degeneration29. An in vivo study reported that the retinotopy observed in cortical activity maps of rats which is likely due to a combination of factors, including the degeneration itself and the corresponding increase in stimulus thresholds29. The pathophysiology of retinal degeneration suggests that when assessing the efficiency of retinal stimulation, comparing the effect of stimulation parameters in both normal and retinal degeneration models would be more beneficial. Moreover, it has been reported that local field potential (LFP) can serve as a reliable signal for investigating the spread of cortical activity in response to retinal stimulation31.

In this study, the optimization of retinal stimulation efficiency is investigated by evaluating the LFPs in the visual cortex of both healthy Long Evans (LE) rats and retinal degenerated S334ter-line-3 (F1) rats in vivo, which is achieved through electrical stimulation of the retina using different parameters such as phase duration, stimulus frequency, and interphase-interval.

Materials and methods

Animals

Eight adult LE rats and five adult F1 rats were used in this study as wild-type and retinal degeneration models, respectively. All the rats in the two groups used in the experiments were between the ages of 7 and 10 weeks and weighed 300–500 g. The LE rats were obtained from the Laboratory Animal Services Centre of the Chinese University of Hong Kong, Hong Kong SAR, China, and the F1 rats were acquired from School of Medicine of University of California, San Francisco, USA. All animals were housed in the Laboratory Animal Research Unit at City University of Hong Kong, Hong Kong SAR, China. The animals were kept on a 12-h light/dark cycle, and food and water were available to them ad libitum. Upon completion of experiments, animals were humanely euthanized by overdose of anaesthetic (Pentobarbitone, 150 mg/kg, i.p.). All experimental procedures were approved by the Animal Subjects Ethics Sub-Committees of City University of Hong Kong (3 A-3-201202) and the Health Department of the Hong Kong Special Administrative Region (11–17 in DH/HA&P/8/2/5 Pt.1). Experiments in this study complied with the ARRIVE guidelines. All experiments were performed in accordance with relevant guidelines and regulations.

Epiretinal electrical stimulation Preparation

Rats were initially anesthetized by intraperitoneal injection with a mixture of ketamine and xylazine (Ketamine: 70 mg/kg, and Xylazine: 7 mg/kg; Alfasan International B.V., Holland) and then mounted in a stereotaxic device. Throughout the surgical procedure and subsequent neural recording, anesthesia was maintained with isoflurane gas, administered with an anesthesia machine (SurgiVet, Smiths Medical PM, Inc., USA). The subject’s body temperature was monitored and maintained at approximately 37.5 °C by a heating pad (Model # TP702; Gamry Industries, Inc., NY, USA). Prior to surgery, the stimulated eye of the pupil was dilated with several drops of tropicamide (Mydrin-P, Santen Pharmaceutical Co. Ltd., Osaka, Japan). During surgery and recording, the stimulated eye was periodically smeared with eye gel (Lubrithal, Dechra Veterinary Products A/S, Mekuvej, Denmark) to keep it clear and moist.

A rubber membrane, measuring 2 cm on each side and equipped with a small central hole was utilized to protrude the eyeball to facilitate the insertion of a stimulating electrode. The entire insertion procedure was operated under a microscope. A needle was employed to incise the upper sclera, thereby creating a pathway for the insertion of the bipolar concentric stimulating electrode (Pt/Ir, diameter of 75 μm, FHC Inc., Bowdoin, ME, USA). Following this, the stimulating electrode was inserted through the incision site along the established path, aided by an articulated holder (Noga, Noga Engineering Ltd., Israel) equipped with calipers. To ensure that the tip of the stimulating electrode was close enough to the epiretinal surface, a potentiostat (Reference 600, Gamry Instruments, Warminster, USA) was used to measure the electrode-retinal impedance in order to estimate the distance between the electrode and the retina surface. Similar operations of electrode implantation and impedance monitoring were introduced in previous studies32,33,34.

Electrophysiological signal recording

A craniotomy was performed on the occipital lobe (approximately 2.5–5.5 mm lateral to the midline and 4.5–7.5 mm posterior to the bregma). This site corresponds to the visual cortex contralateral to the stimulated eye. Following the procedure, the skull piece was carefully removed using a hand-held drill, after which the dura was also gently extracted with a needle. After craniotomy, a bone screw was implanted in the skull to serve as the ground electrode. Additionally, a 4 × 4 grid electrode array (Clunbury Scientific LLC., Bloomfield Hills, USA), comprising 16 electrodes each measuring 3 mm in length, with a 400 μm distance between each tip, was employed to record in vivo electrophysiological signals. This recording electrode array was inserted into the exposed primary visual cortex, approximately 800–950 μm deep, using a microdrive inserter (Micropositioner, David Kopf Instruments, Tujunga, USA) (Fig. 1a). Following craniotomy, the rats were given an hour to recover before being stimulated.

The response to retinal electrical stimulation, defined as electrically evoked potentials (EEPs), was recorded from the primary visual cortex (V1), contralateral to the stimulated eye (Fig. 1a). In this study, we employed LFPs to investigate the effect of stimulation parameters (i.e. frequency, phase duration, and interphase interval) on visual cortical responses. Using a data acquisition system (Micro1401, CED, UK) and an amplifier (A-M system, Inc. Model 3600, UK), the EEPs were recorded at a sampling frequency of 25 kHz with a band-passed filter of 0.3–300 Hz.

Impedance measurement and electrical stimulation of the retina

In this study, the stimulation electrode was positioned in the ventral-temporal region of the retina using an articulated holder equipped with calipers. Subsequently, the impedance between the electrode and the retinal surface was measured in real-time by a potentiostat deploying a 10-mV (r.m.s.) AC sinusoid signal at 100 kHz. This impedance served as a measure for evaluating the distance between the electrode and the retinal surface. Previous studies have established the relationship between impedance and the electrode-epiretinal distance, as well as EEPs, in rats32,33,34. These studies showed that the impedance monotonically increased as the electrode moved closer to the epiretinal surface. Moreover, when the impedance was valued between 5 and 8 kΩ, it correlated well with visual cortical responses, which however, showed a decrease when the impedance value exceeded 8 kΩ. In all experiments conducted in this study, the electrode was manipulated to alter the distance between it and the epiretinal surface until the impedance value consistently measured around 8 kΩ. Prior experiments have determined this to be the optimal impedance value to ensure safe and efficient stimulation while preserving the integrity of the implant. These studies also suggest that an impedance value of 8 kΩ signifies a distance of approximately 370 μm between the electrode and epiretinal surface34.

Fig. 1
figure 1

Experiment setup. (a) The recording array was positioned in the right cortex. The recording array consists of 16 recording channels whose distribution is shown in the figure. The arrows show the direction. (b) The stimulation pattern without interphase interval. The cathodic first, charge-balance and biphasic current stimulation is applied. The various current level range from 10 to 200 µA with a 10 µA step. (c) The stimulation pattern with interphase interval. It shows the stimulation pattern with interphase interval. (d) The top waveform shows the stimulation trial (charge-balanced biphasic pulse, cathodic first). The bottom stack waveform average across 10 trials and each represented the response to 20 different current level stimulation. It is shown that the response increases as the level of stimulation increases. Taking the response to 200 µA stimulation as an example, the baseline is denoted by a blue horizontal dashed line, and the first negative peak is marked by a blue circle. The response amplitude, which was equal to 356.0 µV, is indicated by a blue vertical dashed line. (e) Example for the fitted sigmoid curve of the response evoked by 5 Hz stimulation with 1500 µs phase duration. The fitted sigmoid curve of response evoked by 20 different stimulation levels is plotted as a function of varying current amplitude (blue for the LE group and orange for the F1 group). The circles represent the response level, which referred to the amplitude of the first negative peak to the baseline under each stimulus level. The threshold that elicited 50% (P50) of the saturated response is indicated by the dashed line.

Full size image

Charge-balanced, cathodic first, biphasic current stimulations, ranging from 10 to 200 µA in increments of 10 µA (Fig. 1b, each current repeated 10 times), were applied to the epiretinal surface using a stimulus generator (STG 2004, Multichannel Systems MCS GmBH, Germany). The reference electrode was inserted into the skin adjacent to the nose. Two stimulus patterns were used: one without an interphase interval (IPI) (Fig. 1b) and one with an IPI (Fig. 1c). The pattern without an IPI utilized four frequencies (1, 5, 10 and 20 Hz) and four phase durations (200, 500, 1000 and 1500 µs). In contrast, the pattern with an IPI used only a frequency of 5 Hz and a phase duration of 1000 µs. The IPI involved four different lengths: 0 (i.e., without an IPI), 500, 1000 and 1500 µs. Each parameter set underwent two retinal stimulation trials. Rats were allowed a 5-min recovery interval between trials to avoid retinal damage.

Characterization of electrical evoked potentials of visual cortical neurons

The amplitude from the first negative peak to baseline was used as the response level to analyze the effects of varying stimulation parameters and stimulus patterns on cortical neural responses (Fig. 1d). Specifically, the first negative peak was defined as the minimum point in the timeframe from 10 ms to 50 ms post-stimulation. The response level was represented by a fitted sigmoid curve plotted as a function of varying current amplitude (ranging from 10 to 200 µA in increments of 10 µA). The activation threshold was defined as the current that evoked 50% (P50) of the saturated response (Fig. 1e). To quantify the response level and threshold, we selected the recording electrode among 16 that exhibited the highest amplitude of the first negative peak within the series, referred to as the most activated channel.

To evaluate the effectiveness of various electrical retinal stimulation parameters, the responses evoked by different parameter sets were compared. Additionally, a comparison of response levels between the LE and F1 groups was performed to further examine the impact of stimulation parameters on rats with retinal degeneration. To examine the spatially activated distribution, the activated area in the V1 was visualized using an interpolated heat map (Fig. 2a) and the activation ratio was calculated by converting the heat map into a binary map (Fig. 2b). Figure 2 represents heat maps and binary maps illustrating the response of a single rat to a range of stimulus currents (50, 100, 150, and 200 µA). The arrangement of the 16 recording electrodes was represented by the 4 × 4 array shown in the map. Responses were initially averaged across the ten trials, and subsequently, each rat’s response was normalized to the maximum response (at 200 µA). As indicated in Fig. 2a, regions on the heat map closer to the dark red range exhibited a stronger response as compared to those nearer to the dark blue range. Figure 2b illustrated the binary map where the region with a response level surpassing half of the maximal value is identified as the activated area, while the rest is categorized as the inactivated area. To evaluate the effectiveness of retinal stimulation with different stimulation parameters, the activation ratio was calculated. This ratio was defined as the proportion of the activated area to the entire region. For the statistical analysis of the activation ratio, self-normalization was implemented to mitigate variations among different animals.

Fig. 2
figure 2

Example of spatial maps. (a) Color-coded map of response amplitudes with 2-D interpolation, illustrating the spatial distribution maps across the 16 recording channels of one rat elicited by different stimulus currents (50, 100, 150, and 200 µA). Response amplitudes are normalized to the maximum (data from a single LE animal). The regions closer to the dark-red spectrum exhibit a higher response when averaged over 10 trials, while those closer to dark blue show a lower response. (b) In the binary map, the area with a response level surpassing half of the maximal values is identified as the activated area, while the remainder is categorized as the inactivated area. The red blocks represent the activated area under retinal stimulation, and the dark blue blocks represent the inactivated areas.

Full size image

Data processing and analysis

The Spike2 software (CED, UK) was employed to acquire the signal, and subsequently, the data was processed and analyzed offline using MATLAB. Comparison of the response levels across all 16 channels was conducted and the channel exhibiting the largest response was chosen for further analysis. The goodness of fit of the sigmoid function was calculated to exclude data with a fit lower than 60%, indicating no response. Statistical Package for Social Sciences (SPSS) software was used for statistical analysis. Given that the present study includes both the within-subjects factor and the between-subjects factor, a mixed analysis of variance (ANOVA) was conducted. This approach enables the evaluation of each factor’s individual effects as well as any interaction effects between them. To estimate the effect of frequency and phase duration on threshold and spatial activation ratio, a multi-factor mixed ANOVA, with frequency and phase duration as within-subjects factors and rat group as a between-subjects factor, was conducted. For analyzing the impact of frequency on response level and the impact of the IPI on threshold and spatial activation ratio, a two-way mixed ANOVA was performed. Tukey’s post-hoc comparison was employed to further examine the differences. One F1 rat did not yield a response signal due to the malfunction of the recording electrode array during the experiment. As a result, the analysis comprised of a total of eight LE rats (16 trials, 2 repeated trials for one experiment) and four F1 rats (8 trials).

The initial part of the result section focused on the effect of phase duration and frequency without IPI on threshold, activation ratio, and response level. The subsequent part focused on the impact of introducing IPI, analyzing its effect on the visual cortical response in terms of threshold and activation ratio. All subsections included a comparison between the LE and F1 groups.

Results

Effect of stimulation phase duration and frequency on threshold

The effect of phase duration on the current and charge threshold was examined in this study using the stimulation pattern devoid of an interphase interval (Fig. 3). The charge threshold was obtained by multiplying the current stimulus threshold by the phase duration. This finding revealed that the F1 group exhibited a higher activation threshold compared to the LE group, in agreement with previous research33,35,36. The thresholds of the LE and the F1 groups were also compared at various stimulation frequencies, such as 1, 5, and 10 Hz. The threshold-duration curve was shown in Fig. 3. As the multi-factor mixed ANOVA showed significance in the rat group (**p < 0.01) and phase duration (*p < 0.05), a further post-hoc comparison was performed to examine the effect of the rat group and phase duration on the threshold. The interaction of frequency and rat group demonstrated a significant effect (*p < 0.05), which could be attributed to the dominant impact of the rat group. Despite Fig. 3a–c demonstrating that the current threshold overall decreased with an increase in the stimulation phase duration, the significant difference was only shown between 500 and 1500 µs phase duration in the F1 group at 1 Hz (*p < 0.05; Fig. 3a). In contrast to the current threshold, an increase in the phase duration resulted in an upward trend in the corresponding charge threshold (Fig. 3d–f). At 1 Hz stimulation, a significantly lower charge threshold was observed at 500 µs in both the LE and F1 groups (*p < 0.05), in comparison to at 1500 µs phase duration (Fig. 3d). At 5 Hz, the charge threshold increased significantly as the phase duration elevated from 500 to 1000 µs, and 1500 µs in the LE (*p < 0.05) and F1 groups (***p < 0.001) (Fig. 3e). Differentially, at 10 Hz, the statistical difference in charge threshold was only shown among various phase durations in the LE group, while there was no apparent difference in the F1 group (Fig. 3f).

At a stimulation frequency of 1 Hz, there was a significant difference observed between the LE and F1 groups in terms of both charge threshold and current threshold (Fig. 3a,d), which was not evident at 5 Hz (Fig. 3b,e) or 10 Hz (Fig. 3c,f). Figure 3a,d showed a noticeable difference in the current threshold and charge threshold between the LE and F1 group under 1 Hz stimulation with 1000 and 1500 µs phase duration (*p < 0.05). A more significant difference was observed in both the current threshold and charge threshold between the LE (76.94 ± 11.33 µA and 38.47 ± 5.67 nC, 16 trials, 8 rats) and F1 group (161.56 ± 17.00 µA and 80.78 ± 8.50 nC, 8 trials, 4 rats) (**p < 0.01) at 1 Hz stimulation with 500 µs phase duration (Fig. 3a,d). At 5 Hz (Fig. 3b,e), only the threshold of 1000 µs phase duration had apparent difference between the LE (94.85 ± 8.19 µA and 94.85 ± 8.19 nC, 16 trials, 8 rats) and F1 group (132.92 ± 14.35 µA and 132.92 ± 14.35 nC, 8 trials, 4 rats) (*p < 0.05). As the stimulation frequency increased to 10 Hz (Fig. 3c,f), there was no significant difference between the LE and the F1 groups. These results suggested that the degeneration of the retina may influence the sensitivity of retinal cells to low-frequency stimulation.

Fig. 3
figure 3

Current and charge thresholds of LE (16 trials, 8 rats) and F1 (8 trials, 4 rats) groups under 1, 5, and 10 Hz stimulation. The top row (a–c) displays the current threshold as a function of varying phase duration. The bottom row (d–f) shows the corresponding charge threshold as a function of varying phase duration. The charge threshold is calculated by multiplying the stimulus current threshold by the first phase duration. The statistical significance is assessed using multi-factor mixed ANOVA with Tukey post-hoc comparison; *p < 0.05, **p < 0.01, ***p < 0.001. The black, orange, and blue asterisk indicate the statistical difference between the LE and F1 groups, within the F1 group, and within the LE group, respectively. The error bars represent the standard error of the mean.

Full size image

The shorter phase duration evoked large Spatial activation in the primary visual cortex

The activation ratio in V1 was analyzed in this section to explore the effect of the phase duration (500, 1000, and 1500 µs) on the activation ratio of visual cortical response at different frequencies (1, 5, and 10 Hz). The binary map (Fig. 4a) displayed the activated and inactivated area and showed the activation ratio. The activation ratio used for statistical analysis was normalized to a value of 1.0 for the stimulation with 500 µs phase duration within each animal to reduce the variance among different animals.

Fig. 4
figure 4

Effects of phase duration of the electrical stimulation on the activation ratio. (a) An illustration of the activation ratio in a LE rat under 1 Hz stimulation with different phase durations (500, 1000, and 1500 µs) at 200 µA. The activation ratio was defined as the ratio of the activated area (red part) to the total area (including the activated and the inactivated area). (b) and (c) depict the activation ratio elicited by varying stimulation parameters (1, 5, and 10 Hz electrical stimulation with different phase durations) when receiving threshold and suprathreshold current, respectively, in the LE group (16 trials, 8 rats). (d) and (e) present similar data as in (b) and (c), respectively, for the F1 groups (8 trials, 4 rats). Multi-factor mixed ANOVA with Tukey post-hoc comparison was used for statistical analysis (*p < 0.05, **p < 0.01). The error bars represent the standard error of the mean.

Full size image

Figure 4a showed examples of binary maps with different stimulation phase durations at 200 µA. Figure 4b–e represented bar charts depicting the activation ratio evoked by threshold current and suprathreshold (i.e., 200 µA) for the LE and F1 groups, respectively. The threshold used to calculate the activation ratio referred to the threshold corresponding to each phase duration at each frequency, which could be seen in Fig. 3a–c. As the multi-factor mixed ANOVA revealed that phase duration (**p < 0.01) had a significant impact on the activation ratio, a subsequent post-hoc comparison was conducted. The multi-factor mixed ANOVA test showed no significance in the interaction effect and the frequency. However, a decreasing trend in activation ratios was observed at phase durations of 1000 µs and 1500 µs in the LE group. Although the activation ratio generally declined with phase duration increasing at threshold current in the LE group (Fig. 4b), there was no statistical difference at 1 Hz stimulation in response to the stimulation at threshold current. To ensure complete activation of the cortical neurons, we further computed the activation ratio of these neurons in response to suprathreshold (i.e., 200 µA) stimulation (Fig. 4c). The statistical analysis showed that, at 1 Hz stimulation, the activation ratio decreased significantly from 1 for a 500 µs phase duration to 0.74 ± 0.09 for a 1500 µs phase duration in LE group (*p < 0.05, 16 trials, 8 rats). Significant differences were also observed between the activation ratios elicited by 500 µs phase duration and 1500 µs phase duration (0.52 ± 0.10 for 5 Hz, 0.36 ± 0.15 for 10 Hz) (*p < 0.05, 16 trials, 8 rats).

For the F1 group, no significant difference was observed in the activation ratio under threshold current stimulation (Fig. 4d). When the stimulation current was increased to the suprathreshold level of 200 µA, the cortical activation ratio in the F1 group exhibited trends similar to those in the LE group. Specifically, the activation ratio decreased as the phase duration increased (Fig. 4(e)).

The cortical response level declined with increasing epiretinal stimulation frequency under equal charge injection

The effect of stimulation frequency on cortical response level was investigated using the two-way mixed ANOVA with frequency as within-subjects factors and rat group as between-subjects factors in this section, demonstrating the statistical difference in response to varying frequency (***p < 0.001). Separate plots of frequency versus charge contour figures (Fig. 5a,b) were generated to examine the overall trend of the impact of stimulation frequency on the response levels of EEPs in the LE and F1 groups, respectively. To determine the charge, the current was multiplied by the duration of a single phase. At 200 µA stimulation, four distinct charges (40, 100, 200 and 300 nC) were computed for varying phase durations. Under the same charge injection scale, it was observed that the response level was suppressed with stimulation frequency increasing, as shown in Fig. 5c–f.

The cortical response level induced by electrical stimulation exhibited a decreasing trend with increasing stimulation frequency (Fig. 5c–f), regardless of the charge amount, as evidenced by the results of each of the four phase durations. With a charge injection of 40 nC, as the stimulation frequency increased from 1 to 20 Hz (Fig. 5c), the response level overall decreased from 195.39 ± 74.56 µV and 71.87 ± 55.39 µV to 49.41 ± 14.38 µV and 16.21 ± 2.90 µV for the LE (16 trials, 8 rats) and F1 (8 trials, 4 rats) groups, respectively. With a charge injection of 300 nC (Fig. 5f), the response level of the LE group (16 trials, 8 rats) reduced significantly from 599.37 ± 111.50 µV at 1 Hz stimulation to 63.28 ± 17.25 µV at 20 Hz stimulation (***p < 0.001). Similarly, in the F1 group (8 trials, 4 rats), the response level declined from 426.91 ± 184.47 µV to 102.29 ± 44.41 µV, but not significantly. This finding was consistent with the results of previous studies that investigated the optic nerves of rabbits37,38. Although the two-way mixed ANOVA did not indicate a significant difference between the LE and F1 groups, there was a noticeable distinction between the two groups. Specifically, the LE group exhibited a more pronounced difference in response level than the F1 group for the four typical injecting charges (40, 100, 200 and 300 nC) at 1 Hz stimulation (Fig. 5g), while responses of both groups were similar at high stimulation frequencies (10 and 20 Hz) (Fig. 5i,j). Notably, with a charge injection of 300 nC, the response level of the LE group even dropped below that of the F1 group at 10 and 20 Hz (Fig. 5f).

Fig. 5
figure 5

The effect of the stimulation on the response level of the cortical neurons. (a) and (b) display the frequency vs. charge 2D interpolated contour plots for LE and F1 groups, respectively. The color bar visualizes the response level (in µV). (c), (d), (e), and (f) illustrate response level changes elicited by 200 µA stimulation as a function of stimulation frequency under four phase durations (200, 500, 1000, and 1500 µs) represented by charges (40, 100, 200 and 300 nC). (g), (h), (i), and (j) show the response level evoked by 200 µA stimulation at 1, 5, 10, and 20 Hz as a function of charges (40, 100, 200, and 300 nC). In these figures, the blue circle and the purple triangle represent the LE (16 trials, 8 rats) and the F1 (8 trials, 4 rats) group, respectively. Two-way mixed ANOVA with Tukey post-hoc comparison is used for statistical analysis: *p < 0.05, **p < 0.01, ***p < 0.001. No significant difference is showed between LE and F1 groups. The blue asterisk represents the statistical significance within the LE group. The error bars represent the standard error of the mean.

Full size image

Effect of interphase interval on visual cortical response

The effect of interphase intervals of electrical retinal stimulation on the cortical response was investigated in this section using the two-way mixed ANOVA with IPI as within-subjects factors and rat group as between-subjects factors. Our results showed that thresholds in response to the stimulation with different IPIs in the F1 group were higher than that in the LE group. Yet no significant difference was found between the two groups (Fig. 6a).

To examine the effect of stimulation IPI on cortical EEPs, we analyzed the spatial responses by determining the activation ratio of cortical response elicited by this particular form of stimulation. Quantitative analysis for spatial characteristics was performed using the responses elicited by threshold current and 200 µA in the LE (Fig. 6b,c) and F1 groups (Fig. 6d,e). The threshold used to calculate the activation ratio referred to the threshold corresponding to each IPI, as shown in Fig. 6a. The activation ratio used for statistical analysis was normalized to a value of 1.0 for stimulation without IPI within each animal to reduce the variations among different subjects. The response to current threshold showed that the addition of IPI diminished the activation ratio in both LE (Fig. 6b) and F1 groups (Fig. 6d). In the F1 group, responses were notably weak at IPIs of 1000 µs and 1500 µs when stimulation was applied at the threshold current (Fig. 6d). To further investigate the spatial effects of varying IPIs, further analyses were conducted using a stimulation current of 200 µA. The two-way mixed ANOVA test demonstrated the significant effect of IPI on the spatial ratio (**p < 0.01). The results showed an overall decreasing trend in activation ratio for both LE and F1 groups as the IPI increased (Fig. 6c,e). Specifically, in LE group (16 trials, 8 rats; Fig. 6c), the activation ratio declined significantly from 1 to 0.80 ± 0.05 (for 500 µs IPI, **p < 0.01), 0.71 ± 0.06 (for 1000 µs IPI, **p < 0.01), and 0.72 ± 0.09 (for 1500 µs IPI, **p < 0.01). As for F1 group, Fig. 6e showed that at 200 µA stimulation, the activation ratio evoked by retinal stimulation without IPI was significantly higher than that elicited by stimulation with 1000 µs and 1500 µs.

Fig. 6
figure 6

Cortical response to stimuli with various interphase intervals (0, 500, 1000, and 1500 µs) in the LE (16 trials, 8 rats) and F1 (8 trials, 4 rats) groups at 5 Hz and 1000 µs phase duration stimulation. (a) Activation threshold as a function of varying interphase intervals in LE and F1 groups. (b) and (c) show the normalized activation ratio of the responses elicited by stimulation with different interphase intervals (0, 500, 1000, and 1500 µs) at threshold current and 200 µA in LE group. The activation ratio is normalized to that elicited by the stimulation with 0 µs IPI. (d) and (e) are same to (b) and (c), but for the F1 group. The error bars represent the standard error of the mean. Two-way mixed ANOVA with Tukey post-hoc comparison is performed; *p < 0.05, **p < 0.01, ***p < 0.001.

Full size image

Discussion

This study demonstrates that utilizing electrical stimulation on epi-retina with a short phase duration (500 µs) can lower the charge threshold required to activate the visual cortical neurons (Fig. 3) and achieve a broader spatial activation area (Fig. 4) in both LE and F1 rats. The contribution of IPI significantly decreased the activation ratio, while it did not show any prominent effect on the threshold (Fig. 6), possibly owing to the choice of phase durations in the present study. All results were obtained from in vivo experiments with rats exhibiting retinal degeneration, with almost entirely lost photoreceptors, and rats with normal eyesight. Following, we discuss the potential applicability of the present results to a future retinal implant and limitations, as well as the possible directions of future research. Simpson’s Paradox describes a statistical phenomenon in which an observed association between two variables in a population can emerge, disappear, or even reverse when the population is divided into subgroups39. This highlight the importance of careful data analysis and consideration of subgroup differences39. To mitigate this effect, we applied both the multi-factor mixed ANOVA and two-factor mixed ANOVA in our statistical analysis.

Visual implants hold promise for partially restoring vision in patients with retinal degeneration2,4,9,1040–45. However, these devices still face significant limitations, including the low resolution of artificial vision45,46,47,48. Therefore, the improvement of stimulation efficacy can have far-reaching benefits, not only for visual prosthetics but also for other areas within the field of neuroprosthetics. It is known that employing an optimal stimulation strategy with a lower charge threshold can also reduce power consumption for the implants, making it a practical consideration. Existing studies24,35,3849–51 showed the capacity of optimizing stimulation parameters to improve the performance of retinal implants. The major limitations of these studies include only involving healthy subjects and in vitro settings. Hence, this study’s primary focus is on conducting in vivo experiments on both LE and retinal degenerated F1 animals to examine the effects of electrical stimulation phase duration, frequency, and interphase interval on the V1 cortical neural response. Our previous work examined the optimal distance between the stimulating electrode and the retinal surface to ensure safe and efficient stimulation based on the optimal electrode-retina impedance34. Building upon this, the present study further investigated the optimal stimulation parameters that can effectively evoke visual cortical responses from the perspective of threshold, response level, and spatial characteristics. The findings indicated that optimizing the stimulation parameters, including phase duration, stimulus frequency, and interphase interval, can reduce the threshold and achieve lesser extended cortical activation in both LE and F1 rats. These results suggested a promising retinal stimulation strategy for achieving efficient cortical activation.

Regarding the threshold required to evoke 50% of maximum response, our results indicated that the current threshold decreased with an increase in phase duration, as demonstrated in Fig. 3a–c. Conversely, the charge threshold increased with phase duration as seen in Fig. 3d–f. These findings aligned with prior in vivo studies of intraorbital ON optic nerve electrical stimulation in rabbits37,38 as well as in vitro experiments performed in rodents’ retinas52. Utilizing shorter phase duration stimuli could minimize the total injecting charge required to activate retinal cells, thus reducing potential damage to neural tissue and providing a possible strategy to lower the power consumption of retinal implants. However, given the short phase duration, a relatively higher current amplitude is necessary to reach the activation threshold. This highlighted the opportunity to design advanced electronic devices capable of high-current delivery53.

Our aggregate results showed the LE group exhibited lower thresholds compared to the F1 group, consistent with conclusions from previous in vitro studies on isolated rodents’ retinas35,54,55. Another study reported a similar but not the same finding that rd10 retinas with low spontaneous activities required an increased charge to evoke responses, while the thresholds of rd10 retinas and wild-type mice with high spontaneous activities were similar56. In addition, we noticed that the threshold difference between the two groups decreased as the stimulation frequency increased from 1 to 10 Hz, and the thresholds were nearly equal at 10 Hz (Fig. 3). Similarly, the response level exhibited a discrepancy between the healthy and retinal degenerated groups at low-frequency stimulation and demolished as stimulation increased to 20 Hz (Fig. 5c–j). The difference between the two groups at the low frequency might be attributed to selectively activating bipolar cells within the inner retina at a low stimulation frequency while avoiding direct activation of RGCs that survive during retinal degeneration57,58. Moreover, an earlier in vitro study reported that surviving RGCs during retinal degeneration progression functioned similarly to those in the healthy group, indicating a preferential activation of RGCs at high frequencies when the charge-balanced biphasic rectangular current stimulation was applied51which could account for the diminished difference between two groups at high frequency in the present study. However, the comparable threshold for healthy and retinal degeneration mice was also reported by in vitro works on voltage stimulation51 and current stimulation at 10 Hz56 and 5–10 Hz59. The differences might be partially attributable to retinal remodeling, though the potential changes in the intrinsic physiological characteristics of the retinal ganglion cells remained unclear. These results implied that retinal cell sensitivity to low-frequency stimulation is influenced by retinal degeneration. Considering this, we suggested the higher stimulation frequency (i.e., 10 Hz) as the efficient parameter to activate the visual cortical neurons in retinal degeneration rats at lower charge injection because larger charge injection increased power consumption and necessitated electrode materials with a high charge capacity45.

Previous work in mice19 and the current study in the LE and F1 group (Fig. 5) showed that the response level of EEPs in V1 monotonically declined as the stimulation frequency increased, which was also reported by the in vivo study in rabbit60. In contrast, Nanduri et al. suggested that increasing frequency always increased brightness in patients with epiretinal implants47. It was noteworthy that other clinical trials on epiretinal prostheses reported less brightness over time at higher stimulus frequencies61aligned with our finding. In addition, the patient with subretinal prosthesis also claimed their preference to the lower refresh rates of 5–7 Hz41,62. The contributors to these discrepant clinical reports might be the duration of the different experiments. In specific, less brightness over time was obtained from the experiments of 10-s stimulation61 while Nanduri et al. applied stimulation for 500 ms47which might not be long enough to observe the reduction. Ryu et al. reported a suppressive effect of stimulation frequency increment on network-mediated cortical response to repeated stimulation on the corneal surface in healthy mice19. As the response used for analysis averaged over 10 repeated consecutive stimulations, the decrease in response level here possibly resulted from suppression of response due to repetitive stimulation with frequency increased. It raises the necessity to check the response level pulse-by-pulse further. In addition, a similar reduction of spike activity in RGC (referred to as desensitization) in response to repetitive stimulation with frequency increase was also observed in RGCs of isolated retinas63,64. It suggested the RGCs desensitization as one possible contributor to the decrease in cortical response displayed here, whereas it still required further study. Similarly, the auditory studies also showed the decline of the late auditory evoked potential in normal hearing human subjects following stimulus repetition65as well as electrically evoked compound action potentials in cochlear implant users at a high stimulus rate66. The neural adaptation was thought to contribute to the reduction of response evoked by repeated stimulation65,66which raised another possible factor to the response decrement observed in the present study. However, our results did not uncover the underlying mechanism behind this suppression of cortical response. Therefore, it would be intriguing to investigate the underlying cause of the desensitization observed in cortical responses to repetitive retinal electrical stimulation.

Furthermore, we explored the effect of adding IPI on the threshold. Our results showed no significant impact on the threshold for both LE and F1 rats (Fig. 6a), which can be attributed to the selected phase duration in the present study. Based on Weitz et al.’s findings from experiments on isolated salamander retinas, a computational model, and human testing24their study revealed an inverse relationship between threshold and interphase gap (IPG) (also known as IPI) duration. Another study showed that IPIs shorter than the phase duration did not yield significant differences52. This suggests that the discrepancy between our study and Weitz et al.’s might be attributed to the varying ratios of IPI and phase duration. The IPIs used in Weitz et al.’s work ranged from 26% to 400% of phase duration, while our IPIs only extended up to 1500 µs (150% of phase duration). Another possible factor might be the phase duration, as John et al. found that the effectiveness of IPIs highly depended on the phase duration52. Specifically, they determined four ranges of phase duration based on chronaxie and noticed a significant effect of IPI at three ranges of phase duration, but not significant in the remaining phase duration. Our results confirmed the findings from John et al. that the IPI of certain phase durations had no considerable effect on the cortical response. Further study presents an opportunity to explore the efficacy of longer IPIs (> 150%) and IPIs with different phase durations in visual cortical responses. Furthermore, to our knowledge, the effect of IPIs on cortical spatial activation remains unexplored. The present work provides the first insight into the impact of phase duration and IPI on cortical spatial activation in both healthy and retinal degenerated rats. We found that the adding IPI reduced the cortical activation ratio, and a longer IPI led to a smaller activated area (Fig. 6b–e). It is suggested that the inclusion of an IPI, especially the relatively longer IPI (i.e., 1000 and 1500 µs), could result in a lesser spread of cortical activation.

Previous clinical studies reported that increasing stimulation amplitudes results in brighter and larger phosphenes47underscoring the importance of achieving efficient spatial activation at lower stimulation levels. Building on this, identifying optimal stimulation parameters is essential for targeting specific retinal regions with minimal current, thereby expanding the dynamic range for current adjustment and reducing the risk of retinal tissue damage. Elongated phosphenes reported by patients are thought to be associated with the activation of RGCs’ axon bundles. Two common strategies have been proposed to avoid such axonal activation: one involves indirect stimulation of RGCs to bypass axon bundle, while the other aims to increase the activation threshold difference between axon bundles and RGC somas, enabling selective activation of RGCs located near the electrodes. In addition to modifying the electrode configurations51,67,68,69 several in vitro studies using isolated retina have explored the spatial effects of varying phase durations and frequencies35,36,70. Notably, one study on epiretinal stimulation demonstrated that increasing phase duration led to more focal activation35. Their clinical trials further validated these findings, showing that longer phases produced more localized percepts35. Consistent with these results, our findings indicate that longer phase durations (1000 and 1500 µs) reduce the spatial spread of cortical activation in both LE and F1 rat models. However, another in vitro study using fiber electrodes showed axonal activation regardless of phase duration. This discrepancy may be attributed to differences in electrode geometry; the fibrous structure used in that study may have a lower activation threshold for axon bundles compared to the disk electrodes used in our experiments and previous studies. In terms of in vivo research, studies have been conducted in cats and rabbits to examine the effects of stimulation parameters on visual cortical responses. Su et al.’s group used optical intrinsic signal imaging to evaluate the spatial activation in the secondary visual cortex of cats, finding that increasing current intensity expanded the response areas under sinusoidal stimulation71. Similarly, Sun et al. recorded electrophysiological signals from the primary visual cortex of healthy rabbits and investigated the relationship between spatial activation stimulation current phase ratio38. However, in vivo studies on spatial activation in RD models remain limited. The present study addresses this gap by providing in vivo evidence from RD models. Our results suggest that retinal electrical stimulation using longer phase duration or incorporating an IPI can effectively reduce the spatial extent of cortical activation in response to the epiretinal stimulation.

This study aimed to investigate the effects of retinal electrical stimulation parameters on visual cortical responses by analyzing activation thresholds and activation ratios. Our findings suggest that stimulation with a short phase duration (i.e., 500 µs) offers an effective strategy for eliciting the desired cortical response at lower charge thresholds, potentially minimizing the risk of retinal tissue damage. These results were consistent across both LE and F1 rat models, providing valuable insights for optimizing visual cortical activation in the context of retinal degeneration. However, our analysis of spatial characteristics related to IPI was limited to a stimulation frequency of 5 Hz. The influence of stimulation frequency on activation ratios remains unexplored and warrants further investigation. Additionally, while our study indicate that higher stimulation frequencies lead to a marked reduction in cortical response amplitude, the underlying mechanisms were not addressed in this study and represent an important direction for future research.