Advertisement

Sequence of visual cortex stimulation affects phosphene brightness in blind subjects

Open AccessPublished:April 01, 2022DOI:https://doi.org/10.1016/j.brs.2022.03.008

      Highlights

      • Psychophysics of artificial vision elicited through cortical stimulation in the blind can inform the design of cortical prostheses.
      • Sequential stimulation of visual cortex electrodes produces replicable biases in phosphene brightness perception.
      • The temporal effects were subject specific and had interactions with the electrodes' relative current amplitude.

      Abstract

      Background

      Visual cortical prostheses (VCP) could potentially benefit a majority of the blind population. Feasibility testing of these VCP opens new avenues to characterize stimulation of visual cortex in blind subjects.

      Objective/hypothesis

      To determine if sequential stimulation of visual cortex produces a perception bias in phosphene brightness.

      Methods

      We stimulated three blind subjects implanted with the Orion array with sequences of two and three electrodes and asked them to determine the brighter phosphene, using interval forced-choice paradigms. We selected a set of reference electrodes as the constant stimuli across sequences and compared across three different amplitude levels keeping all other stimulation parameters fixed across electrodes.

      Results

      For two subjects, we measured a significant increase in the probability of perceiving a lower-level amplitude just as bright or brighter than a higher-level amplitude when stimulated later in the sequence (p < 0.001, Wilcoxon rank sum test). The probability of reference electrodes selected as brighter was also higher during the second phase, across most amplitude comparisons. For the third subject, there were measurable but not significant changes, where the first stimuli were perceived as brighter. The effects were consistent within subjects in the three-electrode sequences, where the probability of the reference electrode selected as brighter was correlated to when it was presented in the sequence.

      Conclusions

      We showed evidence of temporal interactions in non-overlapping sequences of electrodes, where the direction of the effect was subject specific but consistent across a variety of electrode locations and current amplitude levels.

      1. Introduction

      The prospect of providing artificial vision for the blind has led to significant interest in developing visual cortical prostheses (VCP), with several ongoing and planned clinical trials [
      • Fernández E.
      • Alfaro A.
      • Soto-Sánchez C.
      • González-López P.
      • Ortega A.M.L.
      • Peña S.
      • et al.
      Visual percepts evoked with an Intracortical 96-channel microelectrode array inserted in human occipital cortex.
      ,
      • Fernández E.
      • Normann R.A.
      CORTIVIS approach for an intracortical visual prostheses.
      ,
      • Lowery A.J.
      Introducing the monash vision group's cortical prosthesis.
      ,
      • Lowery A.J.
      • Rosenfeld J.V.
      • Rosa M.G.
      • Brunton E.
      • Rajan R.
      • Mann C.
      • et al.
      Monash vision group's gennaris cortical implant for vision restoration.
      ,
      • Niketeghad S.
      • Pouratian N.
      Brain machine interfaces for vision restoration: the current state of cortical visual prosthetics.
      ,
      • Pouratian N.
      • Yoshor D.
      • Niketeghad S.
      • Dornm J.
      • Greenberg R.
      Early feasibility study of a neurostimulator to create artificial vision.
      ,
      • Troyk P.R.
      The intracortical visual prosthesis project.
      ]. A number of these groups are currently testing the safety of their systems as a long-term VCP and evaluating their functionality inside and outside clinical settings [
      • Fernández E.
      • Alfaro A.
      • Soto-Sánchez C.
      • González-López P.
      • Ortega A.M.L.
      • Peña S.
      • et al.
      Visual percepts evoked with an Intracortical 96-channel microelectrode array inserted in human occipital cortex.
      ,
      • Pouratian N.
      • Yoshor D.
      • Niketeghad S.
      • Dornm J.
      • Greenberg R.
      Early feasibility study of a neurostimulator to create artificial vision.
      ]. These systems provide a unique opportunity to understand the psychophysics of visual cortical stimulation and artificial vision in blind subjects. These insights can improve the design of next-generation devices and inform the stimulation patterns used to generate artificial vision.
      Some of these VCP currently implement asynchronous stimulation [
      • Pouratian N.
      • Yoshor D.
      • Niketeghad S.
      • Dornm J.
      • Greenberg R.
      Early feasibility study of a neurostimulator to create artificial vision.
      ], which delivers fast-rate non-overlapping sequential pulses across the array. This stimulation paradigm leverages experience from other sensory implants, for which asynchronous stimulation has helped improve the device functionality and reduce electric field interactions [
      • Fisher L.E.
      • Tyler D.J.
      • Triolo R.J.
      Optimization of selective stimulation parameters for multi-contact electrodes.
      ,
      • Horsager A.
      • Greenberg R.J.
      • Fine I.
      Spatiotemporal interactions in retinal prosthesis subjects.
      ,
      • Wilson B.S.
      • Finley C.C.
      • Lawson D.T.
      • Wolford R.D.
      • Eddington D.K.
      • Rabinowitz W.M.
      Better speech recognition with cochlear implants.
      ]. Asynchronous stimulation is also favored due to remaining uncertainty of what happens perceptually when simultaneous multi-electrode stimulation is delivered [
      • Bosking W.H.
      • Beauchamp M.S.
      • Yoshor D.
      Electrical stimulation of visual cortex: relevance for the development of visual cortical prosthetics.
      ,
      • Najarpour Foroushani A.
      • Pack C.C.
      • Sawan M.
      Cortical visual prostheses: from microstimulation to functional percept.
      ,
      • Moleirinho S.
      • Whalen A.J.
      • Fried S.I.
      • Pezaris J.S.
      The impact of synchronous versus asynchronous electrical stimulation in artificial vision.
      ]. Sequential stimulation has worked under the assumption that any temporal interactions in sequences of phosphenes and their effect on perception are negligible. We posit that these temporal interactions may indeed be significant and have conducted experiments focusing on whether the temporal sequences of stimulation generate consistent biases in more qualitative aspects of the perception, e.g., phosphene brightness.
      Previous evidence of serial dependencies (SDs) and systematic biases in visual perception [
      • Fischer J.
      • Whitney D.
      Serial dependence in visual perception.
      ] and in other sensory systems [
      • Motala A.
      • Zhang H.
      • Alais D.
      Auditory rate perception displays a positive serial dependence.
      ] suggest that perception in artificial vision could also be affected by stimulation order and previous stimuli history. Temporal effects on brightness are of particular interest since an ideal VCP should deliver the stimulation such that distinct brightness levels are perceived as different, while equal brightness levels are perceived as similar. Characterization of the psychophysics of sequential stimulation is also of interest since it is a likely candidate to generate form vision in blind individuals [
      • Beauchamp M.S.
      • Oswalt D.
      • Sun P.
      • Foster B.L.
      • Magnotti J.F.
      • Niketeghad S.
      • et al.
      Dynamic stimulation of visual cortex produces form vision in sighted and blind humans.
      ,
      • Wilke R.
      • Gabel V.-P.
      • Sachs H.
      • Bartz Schmidt K.-U.
      • Gekeler F.
      • Besch D.
      • et al.
      Spatial resolution and perception of patterns mediated by a subretinal 16-electrode array in patients blinded by hereditary retinal dystrophies.
      ], as well as a useful method to determine phosphene locations to build more accurate spatial maps [
      • Oswalt D.
      • Bosking W.
      • Sun P.
      • Sheth S.A.
      • Niketeghad S.
      • Salas M.A.
      • et al.
      Multi-electrode stimulation evokes consistent spatial patterns of phosphenes and improves phosphene mapping in blind subjects.
      ]. Experience from some of these sequential stimulation experiments has suggested that subjects perceive differences in brightness during multi-electrode sequences, even if the current amplitudes are initially set at similar levels and produce similar brightness levels when stimulated individually ([
      • Oswalt D.
      • Bosking W.
      • Sun P.
      • Sheth S.A.
      • Niketeghad S.
      • Salas M.A.
      • et al.
      Multi-electrode stimulation evokes consistent spatial patterns of phosphenes and improves phosphene mapping in blind subjects.
      ], personal communication). Furthermore, there have been perceptual differences reported between the bright regions perceived during video stimulation and those estimated from single-electrode stimulation [
      • Barry M.P.
      • Armenta Salas M.
      • Patel U.
      • Wuyyuru V.
      • Niketeghad S.
      • Bosking W.H.
      • et al.
      Video-mode percepts are smaller than sums of single-electrode phosphenes with the Orion® visual cortical prosthesis.
      ], when video stimulation is tested at the highest amplitude level. Although these sequential stimulations do not replicate the asynchronous stimulation that is used in the retinal or the cortical prosthesis, they provide valuable models to characterize the psychophysics of artificial vision and serve as proxy to understand the interactions that arise during real-world use of these prosthesis.
      In this report, we tested direct-stimulation experiments with three blind subjects enrolled as part of the Orion Early Feasibility (EFS) clinical trial (NCT03344848, clinicaltrials.gov) at the University of California Los Angeles (UCLA) site. We ran three sets of experiments to determine whether certain aspects of stimulation sequencing (e.g., order of the electrode in the sequence and relative current amplitude) produce bias in phosphene perception, specifically their perceived relative brightness. In the first two sets, we show that in two-electrode sequences, the current amplitude and stimulation order could interact to make a phosphene appear brighter than stimulation amplitude alone would predict, and this bias is consistent when stimulating either the same or two different electrodes. In the third set of experiments, we measured a cumulative effect of the stimulation order effect on brightness perception, with either later or earlier phases in the sequences perceived as brighter with similar stimulation levels across electrodes. We show that two of the three subjects experience a significant effect on the perceived phosphene brightness due to stimulation sequence, and this stimulation order seems to interact with the current amplitude level and is robust across different phosphene locations.

      2. Methods

      The protocol for the Orion EFS was reviewed and approved by UCLA Institutional Review Board. Four subjects were enrolled and consented to the study at UCLA, and three (one female and two males, 29–64 years old) of them participated in the experiments reported here. Table 1 shows a summary of the demographics and implant information for the participating subjects. The subjects were bilaterally blind, with no-light or some bare-light perception, and they had lost their sight due to non-cortical etiology and therefore had normal neuroanatomy.
      Table 1Subjects’ demographics.
      Subject IDMonths since implantation
      As of last experimental data collection.
      Reason for vision loss
      As reported by clinical sites.
      Number of active electrodes
      As of last experimental data collection.
      Years range since loss of vision (at time of implant)
      S0242Head/eye trauma593–6
      S0341Endophthalmitis590–3
      S0440Congenital glaucoma59>6
      a As reported by clinical sites.
      b As of last experimental data collection.

      2.1 Orion implant surgery and direct-stimulation experiments

      The subjects were implanted with a 60-electrode subdural array which is part of the Orion system developed by Second Sight Medical Products. Orion is composed of a set of internal components (electrode array, cable, electronics package and receiving antenna) and external components (glasses with mounted video camera, communication RF antenna and video processing unit). Each electrode is a 2-mm platinum disc, with a 3-mm diagonal pitch between electrode contacts, center to center. The array was placed on the medial wall of the left occipital lobe, targeting the primary and secondary visual cortices, straddling the calcarine sulcus. Fig. 1A shows the estimated array placement from co-registered pre-surgical MRI and post-surgical CT for subject S02 [
      • Blenkmann A.O.
      • Phillips H.N.
      • Princich J.P.
      • Rowe J.B.
      • Bekinschtein T.A.
      • Muravchik C.H.
      • et al.
      iElectrodes: a comprehensive open-source toolbox for depth and subdural grid electrode localization.
      ,
      • Huff T.
      • Gale S.
      FreeSurfer.
      ]. Three to four weeks after surgery, the subjects returned to the site for initial device activation, i.e., first direct-electrode stimulation. After a 12-week period for device testing and video programs fitting, the subjects returned to the site for research sessions, on a weekly or bi-weekly basis.
      Fig. 1
      Fig. 1Electrode and phosphene locations, and stimulation sequences. A: Estimated electrode placement from pre- and post-implantation imaging for S02. Highlighted blue and orange electrodes show reference and test electrodes, respectively. Additional electrode labels are placed to illustrate array orientation. B: Exemplar distribution of empirical phosphene locations for S02, blue circles show the electrodes used as references, and orange squares show the pair and sequence electrodes tested. Red arrows indicate the exemplar pair from panel A. C: Timing sequences when the same electrode was stimulated at different amplitude levels (top-left traces). Schematic of sequential stimulation of a reference and a pair electrode (bottom-left traces). Diagram of 3IFC sequence of a reference electrode stimulated with other two test electrodes (right traces). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
      During on-site research sessions, we bypassed the camera input, and used the clinical programing computer to communicate via RF antenna to directly activate individual electrodes on the array. This allowed us to send specific sequences to a pair or group of electrodes, as those highlighted in Fig. 1A. For all stimulation sequences, we used a symmetric cathodic-first square pulse, with a phase width of 200 μs and inter-phase gap of 40 μs? The pulse trains for each electrode were 40 ms in duration, delivered at 120 Hz. For each subject, we selected a subset of reference electrodes that spread across the array and spanned a range of threshold amplitudes and locations across each subject's visual field. An example of this subset of electrodes is shown in Fig. 1B (blue circles), along with the test pair electrodes (orange squares).
      We determined phosphene thresholds using an algorithm, which is part of the research testing tool used for direct stimulation. The algorithm starts stimulation at a low current amplitude with incremental increases in amplitude in fixed steps if the subject reports not seeing phosphenes for at least two out of three trials. When the subject reports seeing a phosphene for two out of three trials, the amplitude is systematically decreased, using a modified binary search, until reaching the last current where the subject reported not seeing a phosphene. The last current amplitude associated with perceiving a phosphene is determined as the electrode's current threshold. The stimulation uses a symmetric cathodic-first square pulse, with 200 μs phase duration, 40 μs inter-phase gap, and total duration of 250 ms at 20 Hz. We used the latest available thresholds to determine the amplitudes used for our sequential-stimulation experiments. Supplementary Table 1 shows the range of thresholds and test amplitudes used for each subject.

      2.2 Two-interval forced-choice experiments

      For the 2IFC experiments, two sequential visual cortical stimuli were delivered (referred herein as phase1 and phase 2 stimuli). The subjects were trained to use a game controller to select which phase they perceived as brighter. The trials were self-paced, with a new trial starting shortly after the subjects’ response. We did not provide any auditory cues before stimulation and instructed the subjects to verbally indicate if they only perceived one phosphene on a given trial. When this occurred, the run was cancelled, and a new set of electrodes were selected.
      Fig. 1C shows schematics of the pulse sequences used for the 2IFC experiments. Here we show examples of same-electrode (top row) and different-electrodes (bottom row) stimulations. We tested three relative current amplitude levels. Low indicated current amplitudes ∼1.1 times the electrode threshold; medium or med corresponded to current amplitudes ∼1.3 times threshold; and high was for current amplitudes at least 1.5 times threshold. The amplitude multipliers were selected from experience with video-mode stimulation, where we generate stimulation programs with increasing brightness levels, and these threshold multipliers were those that subjects historically report perceiving as increasingly brighter. The values were not exact due to firmware and maximum current limitations. The current delivered were adjusted by the research tool to the closest current amplitude level, which is determined by each subject's implant characteristics and the range of the requested current. The actual current delivered is reported by the research tool, along with the other stimulation parameters and the subject's responses. In post-hoc analysis, the amplitude current reported was used to re-label, if necessary, the trials as low, medium, or high amplitude level.
      For same-electrode trials, we randomized the order of the comparisons and balanced the trials, so each current amplitude was tested at both phases. Similarly, for the different-electrode 2IFC trials, we randomized the comparisons orders and balanced the trials, so each current amplitude was tested at both phases, and each electrode (reference and test) was stimulated during both phases. For all 2IFC experiments, we used a 500 ms inter-stimulus interval (ISI) between phases. Although the trials were structured to be balanced across amplitude comparisons and electrode pairs, changes in the number of trial runs during the session occurred due to human factors. To account for bias due to number of trials, in post-hoc analysis we resampled the responses using bootstrapping with replacement (N = 10000) to draw twelve responses per electrode pair for each amplitude comparison and estimated their empirical probability as the median of the bootstrapped distribution.
      For the analysis of the 2IFC same-electrode comparisons with same amplitude levels, we calculated the proportion of responses per level where the subject selected phase 2 as brighter. For same-electrode at different amplitude levels, we calculated the proportion of responses per comparison where the lower amplitude level was selected as brighter. For example, in the med-high comparisons we grouped the trials where the med level was stimulated during phase 1 and the high level during phase 2, and vice versa. From these trials, we computed the empirical probability of the med amplitude selected as brighter when stimulated in phase 1 and phase 2.
      From previous experience during single-electrode stimulation and underlying differences in the brightness-amplitude relationship for each electrode, we expected that some electrodes could be consistently dimmer than others or saturate their perceived brightness even if stimulated at high amplitude levels. To test whether this inherent relative brightness difference affected any of the observed temporal interactions, we conducted further sub-analyses in the different-electrode comparisons. For each amplitude comparison, we sorted trials by reference electrodes in three different groups: one where the reference electrodes were consistently selected as brighter (cumulative empirical probability of 0.6 or higher), a second one where the reference electrodes were less bright (cumulative probability of 0.4 or lower), and third group including the electrodes around chance level (cumulative probability between 0.4 and 06). For all groups, we calculated the empirical probability of the reference electrodes selected as brighter, when it was stimulated at each phase.

      2.3 Three-interval forced-choice experiments

      To test any potential temporal interactions in longer sequences, we ran a set of experiments where a reference electrode was stimulated along with two other test electrodes in sequence (Fig. 1C). The trials were balanced such that the reference electrode was stimulated in all three phases. All the electrodes were stimulated at ∼1.4–1.5 times threshold, and we used 500 and 1000 ms ISI between each phase. The trials with different ISI were interleaved in the same run. The subjects were trained to provide a verbal response to indicate which phase they perceived as brighter. A new trial started shortly after the subject's response was entered to the research tool. Given the longer duration of the sequences and the additional concentration it requires from the subjects, in a few initial sessions some subjects reported perceiving less than three phosphenes. To account for this, the subjects were instructed to indicate when fewer than three phosphenes were perceived. If more than two trials in a row were flagged, the run was stopped. Otherwise, the run continued but the flagged trials were excluded from analysis. To analyze these experiments, we grouped the responses across both ISIs and calculated the empirical probability of the reference electrodes selected as brighter, when stimulated at each phase.

      2.4 Statistical analysis

      Statistical assessments were done using MATLAB (The Mathworks Inc., Natick, MA. 2021). To evaluate whether the stimulation phase had a significant effect on brightness perception, we used a Wilcoxon signed rank test for the null hypothesis that the responses for phase 2 significantly differed from chance level (p = 0.50) in the 2IFC same-electrode with same-amplitude comparisons. We employed the Wilcoxon rank sum test to assess whether the empirical probability of selecting the lower amplitude (2IFC same electrode with different amplitudes) or the reference electrode (2IFC different electrodes) as brighter was significantly different between phases. For the 3IFC experiments, we used a Kruskal-Wallis non-parametric one-way analysis of variance, where the treatments were the phase number of when the reference electrode was stimulated. We also used the Kruskal-Wallis test to check for any significant effect the type of amplitude level comparison (e.g., low-med or high-med) had on the empirical probability delta between the first and the second phase, in the 2IFC experiments with different amplitude levels. We corrected for multiple comparisons using Bonferroni critical value. We measured the pairwise Spearman correlation between the empirical probability and the stimulation phases in the 3IFC experiments. For all the tests, significance was determined at an alpha of at least 0.05.
      To address the effect that a preference of a stimulation phase (i.e., bias) might have on the empirical probabilities, we computed the discrimination indices Az and da. The latter provides a measure of how well a signal can be detected against background noise, if the signal and noise inputs have unequal variances (see Eq. (1) in Supplementary material, taken from Ref. [
      • Verde M.F.
      • Macmillan N.A.
      • Rotello C.M.
      Measures of sensitivity based on a single hit rate and false alarm rate: the accuracy, precision, and robustness of', A z, andA.
      ]). This metric is useful to account for the effect an unintended bias towards a particular phase might have on the responses, when interpreted as noise in the signal. While Az is the area under best-fitting ROC curve, which is also robust to unequal variances when estimated from da (see Eq. (3)). To estimate da, we used the same-amplitude same-electrode trials as a noise signal to compare against the same-electrode trials with different amplitudes. We matched the amplitude levels for all the reference electrodes tested. For example, we used the empirical probability same-electrode trials with low amplitude level as our noise input, and the probability of the low amplitude selected as brighter during a particular phase (phase 2 for S02 and S03, and phase 1 for S04) as the signal input. For each reference electrode at each amplitude, we estimated da and Az, and used a t-test to determine if the distribution of Az across electrodes was significantly different than 0.5, which is the case where the noise and signal have the same distribution so the observed differences in phase responses can be attributed to noise. For different-electrode trials, we used the responses of selecting the reference electrode when stimulated in one phase, and the conditional probability of the reference electrode selected when stimulated during the other phase as signal. When comparing phases, we used phase 2 as the signal input for S02 and S03, and phase 1 as the signal for S04.

      3. Results

      The 2IFC and 3IFC experiments were completed across 4 months during the on-site UCLA research sessions, testing up to four different reference electrodes per session. For the same-electrode experiments 2IFC, we tested a total of 29 reference electrodes across the three subjects (μ = 9.3, s.d. = 1.15) with six amplitude levels comparisons (average trials per electrode per amplitude level for S02 = 17.9 ± 6.0, S03 = 33.9 ± 15.6, S04 = 16.8 ± 5.4). For different-electrode 2IFC experiments, we tested an average of 76 unique pairs per subject (s.d. 28.3) at nine amplitude level comparisons (average of trials per electrode pair per amplitude level for S02 = 19.0 ± 7.4, S03 = 19.9 ± 12.1, S04 = 20.9 ± 12.9). To compensate for the variability in the number of trials per comparison across the electrode pairs in the 2IFC experiments, we estimated the proportion of responses by resampling with replacement 12 trials per comparison per electrode pair (N = 10000) for each subject.

      3.1 Effects on brightness perception in two-electrode sequence

      Fig. 2 shows the summary of the subjects' responses for the same-electrode 2IFC experiments. Each comparison level is represented with a different color, and each subject's data points with different markers (S02 = circle, S03 = square and S04 = diamond). Fig. 2A shows the median empirical probability of the second phase being selected as brighter, with standard error bars. We observed mixed responses when a single electrode was tested at the same amplitude level. For example, subjects S02 and S04 had responses mostly at chance levels. However, S03 was more likely to choose the second phase as brighter across all amplitude levels, and it was significant for the med and high amplitudes (Wilcoxon rank-sum test, p < 0.05). It is unclear if this bias is due to a true perceptual difference due to the first stimulation altering the perception of the second phase or if it served as primer, so the second stimuli was more salient, or if the subject simply preferred the second phase. If only the latter were true, we would observe similar distribution in the responses regardless of the stimulation amplitude. However, we measured a linear relationship between the medians and the amplitude level. Although this relationship was not significant (Kruskal-Wallis test, Χ2(2,22) = 2.83, p > 0.29), it suggests that the responses might be both due to a phase-preference bias and a change in the perception of the second stimuli.
      Fig. 2
      Fig. 2Responses for same-electrode 2IFC trials with the same and different amplitude levels. A: shows the median of empirical probability of the second phase selected as brighter for each subject, across same-amplitude comparisons. The black dashed line shows chance level at 0.5. B: average empirical probability of the lower amplitude level selected as brighter, given that it was stimulated at each phase. The asterisks on top display the number of amplitude comparisons that were significantly different than chance (A) or phase 1 vs. phase 2 (B), and the error bars represent the 25th and 75th response quantiles. Data for subjects S02, S03 and S04 are represented by circle, square and diamond markers, respectively.
      Fig. 2B shows the median of the empirical conditional probability of the lower amplitude selected as brighter given it was presented first or second in the sequence. These are for same-electrode sequences stimulated at different amplitudes. Across all subjects, we observed a trend for the lower-level amplitude to be perceived just as bright or brighter than the higher-level amplitude, when the lower level was stimulated during the second phase (significant for S02 and S03, Wilcoxon rank-sum test, p ≤ 0.05). For example, in S02 the med amplitude was more likely to be perceived brighter than the high amplitude (pmed-phase2 = 0.62), if the med level was stimulated on phase 2 (red circles in Fig. 2B); whereas it would be perceived consistently as less bright when it was stimulated during phase 1 (pmed-phase1 = 0.12). For subject S03, all of the lower levels responses in phase 2 were significantly higher than in phase 1, and the largest changes in the median of the empirical probabilities happened for the high-med amplitude level (μS02_Δphases = 0.50 and μS03_Δphases = 0.65) (Supplemental Fig. S1). Although the trend was also measured for S04 responses, none of the differences were significant. To further verify that the trends observed were significant, we computed a sensitivity index for each amplitude level comparison known as Az. Fig. 3A show a summary of the Az metric for the same-electrode experiments, where a value of 0.5 means the signal and noise are detected at equal rates. In the comparisons between med-high levels and low-med levels of S02, the perception of the lower amplitude in the pair was higher if this was stimulated during phase 2, and it was significant for the low-med levels (right-tail t-test, p < 0.05). Subject S03 only had significant higher Az index for the med-high comparison (right-tail t-test, p < 0.05), which was the comparison with the largest probability increase (Fig. 2B). In S04, we also observed a trend of the signal being detected at higher than chance rates (Az > 0.5) when the low amplitude was stimulated in the first phase for med-high and low-med amplitudes. However, none of the amplitude comparisons were significant.
      Fig. 3
      Fig. 3Sensitivity Index. Estimated area under the ROC curve Az metric. A: Mean index values for same-electrode comparisons at different amplitude levels, using same-amplitude responses as noise signal. Dashed black line shows chance level performance, and error bars depict one standard deviation. B–D: mean index values for different-electrode comparisons at all amplitude levels, for low brightness (LB), high brightness (HB) and Chance level (CL) groups, respectively. Asterisks a top each data point signals a significance (Az > 0.5). Square, rectangular and diamond markers show data for S02, S03 and S04, respectively.
      For the different-electrode comparisons, we split the responses on reference electrodes in groups of those selected as brighter at higher than chance across all amplitude comparisons (higher brightness = HB elecs.), and electrodes selected at lower than chance level (lower brightness = LB elecs.), and the electrodes selected around chance level (Chance level = CL elecs). These categories were split per amplitude level, so not all groups had trials in all amplitude comparisons. Fig. 4 summarizes the responses for the different-electrode 2IFC experiments, with the top row showing the LB electrodes, the middle row the HB electrodes and the bottom row the CL electrodes. We observed an effect of the stimulation order on brightness perception across multiple amplitude comparisons for S02 and S03, in all three electrode groups. Fig. 4A, C and 4E summarize the data when the reference and pair electrodes were stimulated with the same amplitude levels. They illustrate the median of the empirical conditional probability of the reference electrode chosen as brighter when stimulated at each phase for LB, HB, and CL electrode groups respectively, with error bars depicting 25th and 75th quantiles. For two subjects (S02 and S03), the reference electrode was more likely to be chosen as brighter if it was stimulated during the second phase. In LB electrodes the change in high comparisons was significant for S02, but for S03 it was significant for the low level comparisons (Wilcoxon rank sum test, p < 0.05). In the HB electrodes, we observed similar trends, with significant differences in the low and high comparisons for S02 and S03, respectively. For the CL electrodes, the increase in probabilities was significant for the med and high levels in S02, and in the med level for S03. Although for S04 we observed an increase in the probability that the reference electrode would be chosen as brighter, it was in the opposite direction as the other two subjects. The trend was present across most of the electrode groups and comparison levels, but it was only significant in the low level for the CL electrodes. When we look at the sensitivity index Az for these comparisons in Fig. 3C–F, we observe that the high and low level differences in S02 remain significant in the LB and HB electrodes, as well as the med and high in the CL electrodes (Az > 0.5, right-tail t-test, p < 0.05). Similarly, the differences in probabilities in the low and med comparisons are significant for S03 in LB and CL groups, respectively (Az > 0.5, right-tail t-test, p < 0.05). For S04 the higher probability of choosing the reference electrode if stimulated in phase 1, was also significant for the low amplitude level in the CL electrodes (Az > 0.5, right-tail t-test, p < 0.05).
      Fig. 4
      Fig. 4Summary of responses for different-electrode 2IFC trials. Top-row panels summarize trials of reference electrodes in the low brightness (LB) group, the middle-row panels the trials for high brightness (HB) electrodes, and the bottom-row panels show the chance level (Ch) group. A, C, E: Median of the empirical conditional probability of reference electrode chosen as brighter given the stimulation phase, when ref. and pair electrodes were stimulated at the same amplitude. B, D, F: Median of the empirical conditional probability of reference electrode chosen as brighter given the stimulation phase, when both electrodes were stimulated at different amplitudes. Error bars show 25th and 75th quantiles. Circles, squares, and diamonds show data points for S02, S03 and S04, respectively. Markers on top of responses highlight the number of significant differences between phase 1 and phase 2 responses: asterisks showing for up to three levels.
      Fig. 4B, D and 4F display the median of the empirical probabilities when comparing different amplitudes for LB, HB and CL groups, respectively. Error bars depict the 25th and 75th quantiles. Similar to the same-amplitude comparisons, we observed a consistent increase in the probability of the reference electrodes chosen as brighter in two subjects (S02 and S03) when the reference was stimulated during the second phase. A reference electrode with a lower amplitude level could be perceived just as bright or brighter than a higher amplitude level if it was stimulated during the second phase. For example, S02 responses in the low-med comparison (brown circles in Fig. 3F) were consistently lower than chance for phase 1 (prefElec-phase1 = 0.32), but they were perceived as brighter than the pair electrode when stimulated in the second phase (prefElec-phase2 = 0.67). Similar effects were observed for S02 and S03 in the HB and CL electrode groups, across different amplitude comparisons. In S04, for LB electrodes we measured a consistent trend where the reference electrodes were slightly more likely to be chosen as brighter if stimulated during the first phase, when comparing different amplitudes. However, these were only significant in the CL group for the refHigh-pairLow trials. In addition, for one of the subjects when the electrodes were compared at the highest amplitude levels, they displayed the largest changes in empirical probability between stimulation phases. For S02, the median increase when the reference electrode was stimulated at med or high was significantly higher than for most other comparisons in both the LB (medianS02_Δphases = 0.35) and HB (medianS02_Δphases = 0.25) groups (Supplemental Fig. S2). Finally, Fig. 3B–D show the sensitivity index Az that include the different-amplitude comparisons in the different-electrode trials. We measured that the difference between the phase 1 and phase 2 responses was significant for several amplitude comparisons, across all subjects. This was determined as the mean of the Az index being significantly larger than 0.5 (right-tail t-test, p < 0.05). In LB electrodes, for S02 the increase in probability in phase 2 was significant for the refMed-pairHigh comparison. For HB electrodes in S03, the increase was significant for refMed-pairLow and refHigh-pairMed trials. For CL electrodes, we measured more amplitude comparisons being significant for all subjects (S02 = reflow-pairMed, all refMed comparisons, and refHigh-pairMed; S03 = all refMed comparisons, and refHigh-pairLow; S04 = refMed-pairLow and refHigh-pairLow).

      3.2 Stimulation order effect on brightness in three-electrode sequences

      For the 3IFC experiments, we tested an average of 7 reference electrodes (s.d. = 1.7) per subject. This included a mean of 55.7 unique sequences (s.d. = 29.7) per subject, with an average of 10.3 trials per sequence (s.d. = 0.6). We did not measure any significant differences in the responses when using 500 or 1000 ms ISI, and the responses were pooled across ISI durations.
      Fig. 5 shows the average proportion of responses where the reference electrode was chosen as brighter in each phase with standard error bars. For subjects S02 and S03, we measured a positive and significant correlation between the probability that the reference electrode would be perceived as brighter and the phase order at which it was stimulated (ρS02 = 0.33 and ρS03 = 0.30, p < 0.001). For these subjects, the reference electrode was consistently perceived as less bright when stimulated first (prefElec-phase1 < 0.2), and it was reported as brighter as it was stimulated at later phases (prefElec-phase2 ≈ 0.3, and prefElec-phase3 ≥ 0.37). We observed an opposite direction effect in S04, where the reference electrodes were more likely to be reported as brighter if it was stimulated during the first phase (prefElec-phase1 = 0.45) and likely to be perceived as brighter at chance levels during the other two phases (prefElec-phase2 = 0.25, prefElec-phase3 = 0.30).
      Fig. 5
      Fig. 5Summary of responses for 3IFC experiments. Data points show the median empirical probability of the reference electrodes selected as brighter, given that the electrode was stimulated at each phase. The error bars represent standard error. The asterisks highlight phases with significantly different averages between each of the stimulation phases for each subject. Circle, square and diamond markers show data for S02, S03 and S04, respectively.

      4. Discussion

      4.1 Evidence of stimulation sequence effect on brightness perception

      Our results provide novel insights to the psychophysics of phosphene perception in blind subjects. In other visual prostheses systems, when studying brightness perception most of the interest has laid on persistence, which refers to how quickly a particular phosphene might fade after sequential stimulation [
      • Freeman D.K.
      • Fried S.I.
      Multiple components of ganglion cell desensitization in response to prosthetic stimulation.
      ,
      • Horsager A.
      • Greenwald S.H.
      • Weiland J.D.
      • Humayun M.S.
      • Greenberg R.J.
      • McMahon M.J.
      • et al.
      Predicting visual sensitivity in retinal prosthesis patients.
      ,
      • Pérez Fornos A.
      • Sommerhalder J.
      • da Cruz L.
      • Sahel J.A.
      • Mohand-Said S.
      • Hafezi F.
      • et al.
      Temporal properties of visual perception on electrical stimulation of the retina.
      ]. The objective of this report was to explore a different aspect of the artificial vision perception, by testing the changes in relative brightness due to sequential stimulation, with evidence of a significant effect of sequence on brightness perception. Although the direction of the effect differed across subjects, there were quantifiable trends due to stimulation sequence in all of them. On one hand, in the 2IFC experiments, S02 and S03 were more likely to perceive a reference electrode or a lower amplitude level as brighter when delivered during the second phase (Fig. 2, Fig. 3, Fig. 4). On the other hand, S04 showed less consistent results, and the trends that were observed were in the opposite direction, where the electrodes and amplitudes stimulated first in the sequence were more likely to be selected as brighter. Looking closer at the demographics, it is notable that S04 is the subject in the category of the longest time since loss of vision, so it could be possible that these differences are due to changes in the cortical circuitry of visual cortex. However, if this was the case, we would expect other differences in phosphenes characteristics (e.g., poor or lack of retinotopy in the phosphene map) but this has not been observed in other experiments done with this subject. It could also be that the extended time without any visual input affected the cognitive processing of visual information in higher visual areas, but our current subject population does not allow us to explore this. Further studies with broader population would be key to understand if time since vision loss or other aspects of the subjects’ blindness could affect the functionality of a VCP.
      Results from the 3IFC experiments show evidence that perception bias measured in the 2IFC experiments extended to longer sequences and appeared to aggregate. This was measured as the significant positive correlation between responses and stimulation phase measured in S02 and S03 (Fig. 5). These results are limited by the small number of subjects, especially given the differences in their responses, so it would be desirable to replicate this in a larger patient cohort. However, given the restricted pool of blind subjects implanted with this type of device, we believe the results still provide valuable insights of the psychophysics of artificial vision in individuals who lost their sight.
      It is also possible that artificial vision perception displays similar subject-specific patterns that have been reported in visual percepts of sighted subjects [
      • Kosovicheva A.
      • Whitney D.
      Stable individual signatures in object localization.
      ,
      • Wang Z.
      • Murai Y.
      • Whitney D.
      Idiosyncratic perception: a link between acuity, perceived position and apparent size.
      ] and in other sensory modalities [
      • Rincon-Gonzalez L.
      • Buneo C.A.
      • Helms Tillery S.I.
      The proprioceptive map of the arm is systematic and stable, but idiosyncratic.
      ], where stable but idiosyncratic biases are present in spatial visual perceptions and in proprioceptive maps. If this is the case, the directions of the effect would not be uniform across subjects, but the effect would be replicable within each subject across different stimulation sequences, which is somewhat consistent with the current results. This dependency could bias the perception of the subjects towards phosphenes that appear later or earlier in the sequence, depending on the subject specific pattern. This bias could help explain some of the perceptual differences reported during longer asynchronous stimulations in video mode [
      • Barry M.P.
      • Armenta Salas M.
      • Patel U.
      • Wuyyuru V.
      • Niketeghad S.
      • Bosking W.H.
      • et al.
      Video-mode percepts are smaller than sums of single-electrode phosphenes with the Orion® visual cortical prosthesis.
      ], where all the electrodes were stimulated at the same suprathreshold amplitude level but only a fraction of the predicted area covered by individual phosphenes was reported by the subjects.

      4.2 Interactions between amplitude levels and stimulation order

      To explore the possibility that different attributes of the stimuli, e.g., the current amplitude, interact with the temporal sequence and further affect the brightness perception, we explicitly compared different amplitude levels in the 2IFC sequences. Firstly, when the same electrode was stimulated at different amplitude levels, two of the subjects consistently reported the lower amplitude as brighter more frequently during the second phase than when stimulated during the first phase (Fig. 2B). When accounting for any possible bias due to phase preference, we observe that the difference in the responses is still significant for at least one of the amplitude levels, low-med for S02 and med-high for S03 (Fig. 3A), with trends of higher signal detection (Az > 0.5) for most the other amplitudes. This finding is consistent with previous reports of brightness-matching experiments in retinal implants, where smaller charges were necessary to match sequential stimuli to a standard when the low amplitude was stimulated after a high amplitude [
      • Horsager A.
      • Greenberg R.J.
      • Fine I.
      Spatiotemporal interactions in retinal prosthesis subjects.
      ]. Secondly, for the two subjects that reported this effect, we also observed that the magnitude of the effect was larger with higher amplitudes. This was measured as larger increase in the conditional probability of the lower amplitude selected as brighter when stimulated at either med or high amplitudes in the same-electrode different-amplitude comparisons (see Fig. S1). This suggests that phosphenes that were already bright were affected more by the stimulation order. However, in the different-electrode trials (Fig. 4B, D and 4F), this trend was not as clear, and the additional sub-analysis only yielded significant differences for S02 (Fig. S2).
      These results also suggest that the order and amplitude effects are stable across different phosphene locations, as the reference and pair electrodes spanned the subjects’ visual space and across the physical array. We did not find any significant correlation between the electrodes difference in polarity and eccentricity, or the distance in visual space between electrodes, and the probability of the reference electrodes selected as brighter for any of the subjects. When looking at the correlation of the distance between electrodes from their array location and the probability selecting phase 2 as brighter (regardless of which electrode was selected) across all the experiments, we did not find a significant correlation for any of the subjects (see Fig. S3). This would suggest any temporal current summation due to stimulation in neighboring electrodes does not explain any bias for the later phase in brightness perception. Additionally, to account for possible differences due to the electrode locations and summation effect of the activated cortical area across different amplitude levels, we re-estimated the probabilities of the of all our experiments in electrode sets that were very close together, either same electrode or in adjacent rows or columns, or at least on either edge of the array (see Supplementary material, Fig. S4). We found similar trends across the different electrode groups in both close and far sub-groups of electrodes, and no significant difference in the overall responses (Wilcoxon rank-sum test, p > 0.5), and only a significant of electrode group (close or far) in the probability changes of the low-med amplitude comparison for S03 and S04 (Wilcoxon rank-sum, p = 0.02). Finally, the Az index we estimated for the different-electrode trials already considers the probability that the reference electrode will be chosen as brighter in the other phase, so if the effect was solely due to cortical area activation, we would not see a large difference between the phases responses and the index would stay around 0.5. These additional tests give us confidence that the effect cortical activation might have on the responses is not the main contributor of the stimulation order effect we measured in at least two of our subjects. Making it likely that the effects measured are due to perceptual changes, rather than cellular-level mechanisms.

      4.3 Sequence effect in artificial vision

      Evidence with sighted subjects have shown the existence of SD in the visual system across a variety of attributes [
      • Fischer J.
      • Whitney D.
      Serial dependence in visual perception.
      ,
      • Fornaciai M.
      • Park J.
      Serial dependence generalizes across different stimulus formats, but not different sensory modalities.
      ,
      • Manassi M.
      • Liberman A.
      • Kosovicheva A.
      • Zhang K.
      • Whitney D.
      Serial dependence in position occurs at the time of perception.
      ], where a response is biased towards previous stimuli to facilitate perceptual stability. This SD could explain the change in responses when the same electrode was stimulated at different amplitudes (Fig. 2B), where a lower amplitude would be perceived as closer in brightness to a previous higher amplitude. However, a purely attraction SD does not fully explain the responses when testing different electrodes (Fig. 4), since the increase in probability was measured when the preceding stimuli were at the same or at lower amplitude levels. It appears that similar to recent reports [
      • Czoschke S.
      • Fischer C.
      • Beitner J.
      • Kaiser J.
      • Bledowski C.
      Two types of serial dependence in visual working memory.
      ], both attraction and repulsion SDs are also integrated in the perception of stimulation-induced phosphenes, i.e. sequential stimuli perceived as either more similar or more different from each other. For example, S02 and S03 were most likely to report lower amplitude levels as brighter or just as bright as higher amplitude stimuli (effectively integrating the stimuli), and they would report more often a high amplitude level as brighter if stimulated in phase 2 (resulting in stimuli segregation). Similar effects have also been briefly described in sequential retinal stimulation with a smaller subset of electrodes, represented as faciliatory and suppressive effects [
      • Horsager A.
      • Boynton G.M.
      • Greenberg R.J.
      • Fine I.
      Temporal interactions during paired-electrode stimulation in two retinal prosthesis subjects.
      ]. For the design of stimulation paradigms in a VCP, it is important to understand whether faciliatory and suppressive processes, like those suggested above, are primarily triggered by the interactions of electrode order and current amplitude as those measured here or if other features of the stimulation, e.g., different ISIs, pulse frequency or number of pulses, could also contribute to changes in relative brightness perception. Additional studies will also need to evaluate if these interactions remain when more than two phosphenes are present. Moreover, although a lot of the serial interaction studies have been focused on visual stimuli, there is evidence of trial history and previous stimuli affecting subjects responses in multisensory and artificial sensation studies [
      • Kayser S.J.
      • Kayser C.
      Trial by trial dependencies in multisensory perception and their correlates in dynamic brain activity.
      ,
      • de Lafuente V.
      • Romo R.
      Neuronal correlates of subjective sensory experience.
      ]. While some of these effects could be due the reward scheme in the experiments [
      • de Lafuente V.
      • Romo R.
      Neuronal correlates of subjective sensory experience.
      ], other show how previous stimuli affect the perception of the next trial, and describe distinct neural correlates for this serial interaction [
      • Kayser S.J.
      • Kayser C.
      Trial by trial dependencies in multisensory perception and their correlates in dynamic brain activity.
      ]. Our experiments were not designed to explore the mechanisms correlated to changes in perception and it should not be discounted that local circuitry adaptation could be in part responsible for some of these reported changes. However, given the controls we analyzed post-hoc regarding cortical stimulation, electrode location and electrode preference, we consider that most of the changes are due to perceptual changes.

      4.4 Considerations for visual cortical prostheses

      Sequencing order is a relevant feature to incorporate in the selection of stimulation parameters for VCP, especially if the intent is to deliver consistent brightness information with the camera-captured images. Firstly, given the variable direction effects measured across our subjects, future VCPs might benefit by integrating quantitative aspects of the phosphenes, like the spatial map or phosphene thresholds, to the video programs, as well as subject-specific changes in perceptions due to sequential stimulation. Secondly, if multiple-electrode sequences are desirable to generate form vision with cortical stimulation [
      • Beauchamp M.S.
      • Oswalt D.
      • Sun P.
      • Foster B.L.
      • Magnotti J.F.
      • Niketeghad S.
      • et al.
      Dynamic stimulation of visual cortex produces form vision in sighted and blind humans.
      ], then a clear understanding of how the perception changes during sequential stimulation is key to implement these stimulation paradigms. These experiments do not represent an exhaustive exploration of the stimulation parameter pool. Although the subject's responses in the 3IFC experiments did not significantly change between the 500 and 1000 ms ISIs, the electrodes were tested at similar current levels, possibly diminishing any amplitude and temporal interaction effects. Moving forward, the effect of ISIs on brightness perception will need a more comprehensive evaluation, as some of these temporal interactions might disappear with very short or very long ISIs. Furthermore, the optimal stimulation rate of pulses, i.e., stimulation frequency, for VCP remains an open question and it would be important to explore if the frequency at which each electrode is stimulated affects these temporal effects in perception. Moreover, while the current focus is on brightness, future studies will have to critically evaluate sequence effects on other perceptual qualities, like shape, size, and persistence. Finally, it will be desirable to test different electrode sequencing order during video-mode stimulation, considering the relative amplitude differences, and evaluate its effects on subjects' qualitative perception and their performance on functional tasks.

      CRediT authorship contribution statement

      Michelle Armenta Salas: Conceptualization, Methodology, Investigation, Data curation, Formal analysis, Visualization, Writing – original draft, Writing – review & editing. Joseph Bell: Investigation, Writing – review & editing. Soroush Niketeghad: Investigation, Writing – review & editing. Denise Oswalt: Writing – review & editing. William Bosking: Writing – review & editing. Uday Patel: Investigation, Writing – review & editing. Jessy D. Dorn: Writing – review & editing, Project administration, Funding acquisition. Daniel Yoshor: Writing – review & editing, Project administration. Robert Greenberg: Writing – review & editing, Project administration, Funding acquisition. Ausaf Bari: Project administration. Nader Pouratian: Conceptualization, Methodology, Validation, Resources, Writing – review & editing, Supervision, Project administration, Funding acquisition.

      Declaration of competing interest

      The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dr Pouratian is a consultant for Abbott Laboratories and Second Sight Medical Products. Uday Patel and Jessy Dorn are employees of Second Sight Medical Products, the manufacturer of the Orion Visual Prosthesis.

      Acknowledgements

      Research reported in this publication was fully supported by National Institute of Neurological Disorders and Stroke of the National Institutes of Health under award number UH3NS103442 . The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

      Appendix A. Supplementary data

      The following is the Supplementary data to this article:

      References

        • Fernández E.
        • Alfaro A.
        • Soto-Sánchez C.
        • González-López P.
        • Ortega A.M.L.
        • Peña S.
        • et al.
        Visual percepts evoked with an Intracortical 96-channel microelectrode array inserted in human occipital cortex.
        J Clin Invest. 2021; https://doi.org/10.1172/JCI151331
        • Fernández E.
        • Normann R.A.
        CORTIVIS approach for an intracortical visual prostheses.
        in: Gabel V.P. Artif. Vis. Clin. Guide. Springer International Publishing, Cham2017: 191-201https://doi.org/10.1007/978-3-319-41876-6_15
        • Lowery A.J.
        Introducing the monash vision group's cortical prosthesis.
        IEEE, 2013: 1536-1539https://doi.org/10.1109/ICIP.2013.6738316
        • Lowery A.J.
        • Rosenfeld J.V.
        • Rosa M.G.
        • Brunton E.
        • Rajan R.
        • Mann C.
        • et al.
        Monash vision group's gennaris cortical implant for vision restoration.
        Artif. Vis., Springer, 2017: 215-225
        • Niketeghad S.
        • Pouratian N.
        Brain machine interfaces for vision restoration: the current state of cortical visual prosthetics.
        Neurotherapeutics. 2019; 16: 134-143
        • Pouratian N.
        • Yoshor D.
        • Niketeghad S.
        • Dornm J.
        • Greenberg R.
        Early feasibility study of a neurostimulator to create artificial vision.
        Neurosurgery. 2019; 66https://doi.org/10.1093/neuros/nyz310_146
        • Troyk P.R.
        The intracortical visual prosthesis project.
        Artif. Vis., Springer, 2017: 203-214
        • Fisher L.E.
        • Tyler D.J.
        • Triolo R.J.
        Optimization of selective stimulation parameters for multi-contact electrodes.
        J NeuroEng Rehabil. 2013; 10: 25https://doi.org/10.1186/1743-0003-10-25
        • Horsager A.
        • Greenberg R.J.
        • Fine I.
        Spatiotemporal interactions in retinal prosthesis subjects.
        Invest Ophthalmol Vis Sci. 2010; 51: 1223-1233https://doi.org/10.1167/iovs.09-3746
        • Wilson B.S.
        • Finley C.C.
        • Lawson D.T.
        • Wolford R.D.
        • Eddington D.K.
        • Rabinowitz W.M.
        Better speech recognition with cochlear implants.
        Nature. 1991; 352: 236-238https://doi.org/10.1038/352236a0
        • Bosking W.H.
        • Beauchamp M.S.
        • Yoshor D.
        Electrical stimulation of visual cortex: relevance for the development of visual cortical prosthetics.
        Annu Rev Vis Sci. 2017; 3: 141-166https://doi.org/10.1146/annurev-vision-111815-114525
        • Najarpour Foroushani A.
        • Pack C.C.
        • Sawan M.
        Cortical visual prostheses: from microstimulation to functional percept.
        J Neural Eng. 2018; 15021005https://doi.org/10.1088/1741-2552/aaa904
        • Moleirinho S.
        • Whalen A.J.
        • Fried S.I.
        • Pezaris J.S.
        The impact of synchronous versus asynchronous electrical stimulation in artificial vision.
        J Neural Eng. 2021; 18051001https://doi.org/10.1088/1741-2552/abecf1
        • Fischer J.
        • Whitney D.
        Serial dependence in visual perception.
        Nat Neurosci. 2014; 17: 738-743https://doi.org/10.1038/nn.3689
        • Motala A.
        • Zhang H.
        • Alais D.
        Auditory rate perception displays a positive serial dependence.
        Percept. 2020; 11 (2041669520982311)https://doi.org/10.1177/2041669520982311
        • Beauchamp M.S.
        • Oswalt D.
        • Sun P.
        • Foster B.L.
        • Magnotti J.F.
        • Niketeghad S.
        • et al.
        Dynamic stimulation of visual cortex produces form vision in sighted and blind humans.
        Cell. 2020; 181: 774-783
        • Wilke R.
        • Gabel V.-P.
        • Sachs H.
        • Bartz Schmidt K.-U.
        • Gekeler F.
        • Besch D.
        • et al.
        Spatial resolution and perception of patterns mediated by a subretinal 16-electrode array in patients blinded by hereditary retinal dystrophies.
        Invest Ophthalmol Vis Sci. 2011; 52: 5995-6003https://doi.org/10.1167/iovs.10-6946
        • Oswalt D.
        • Bosking W.
        • Sun P.
        • Sheth S.A.
        • Niketeghad S.
        • Salas M.A.
        • et al.
        Multi-electrode stimulation evokes consistent spatial patterns of phosphenes and improves phosphene mapping in blind subjects.
        Brain Stimul. 2021; 14: 1356-1372https://doi.org/10.1016/j.brs.2021.08.024
        • Barry M.P.
        • Armenta Salas M.
        • Patel U.
        • Wuyyuru V.
        • Niketeghad S.
        • Bosking W.H.
        • et al.
        Video-mode percepts are smaller than sums of single-electrode phosphenes with the Orion® visual cortical prosthesis.
        Invest Ophthalmol Vis Sci. 2020; 61 (927–927)
        • Blenkmann A.O.
        • Phillips H.N.
        • Princich J.P.
        • Rowe J.B.
        • Bekinschtein T.A.
        • Muravchik C.H.
        • et al.
        iElectrodes: a comprehensive open-source toolbox for depth and subdural grid electrode localization.
        Front Neuroinf. 2017; 11: 14https://doi.org/10.3389/fninf.2017.00014
        • Huff T.
        • Gale S.
        FreeSurfer.
        in: Kreutzer J. DeLuca J. Caplan B. Encycl. Clin. Neuropsychol. Springer International Publishing, Cham2017: 1-2https://doi.org/10.1007/978-3-319-56782-2_9064-2
        • Verde M.F.
        • Macmillan N.A.
        • Rotello C.M.
        Measures of sensitivity based on a single hit rate and false alarm rate: the accuracy, precision, and robustness of', A z, andA.
        Percept Psychophys. 2006; 68: 643-654
        • Freeman D.K.
        • Fried S.I.
        Multiple components of ganglion cell desensitization in response to prosthetic stimulation.
        J Neural Eng. 2011; 8016008https://doi.org/10.1088/1741-2560/8/1/016008
        • Horsager A.
        • Greenwald S.H.
        • Weiland J.D.
        • Humayun M.S.
        • Greenberg R.J.
        • McMahon M.J.
        • et al.
        Predicting visual sensitivity in retinal prosthesis patients.
        Invest Ophthalmol Vis Sci. 2009; 50: 1483-1491https://doi.org/10.1167/iovs.08-2595
        • Pérez Fornos A.
        • Sommerhalder J.
        • da Cruz L.
        • Sahel J.A.
        • Mohand-Said S.
        • Hafezi F.
        • et al.
        Temporal properties of visual perception on electrical stimulation of the retina.
        Invest Ophthalmol Vis Sci. 2012; 53: 2720-2731https://doi.org/10.1167/iovs.11-9344
        • Kosovicheva A.
        • Whitney D.
        Stable individual signatures in object localization.
        Curr Biol. 2017; 27: R700-R701https://doi.org/10.1016/j.cub.2017.06.001
        • Wang Z.
        • Murai Y.
        • Whitney D.
        Idiosyncratic perception: a link between acuity, perceived position and apparent size.
        Proc R Soc B Biol Sci. 2020; 287 (20200825)https://doi.org/10.1098/rspb.2020.0825
        • Rincon-Gonzalez L.
        • Buneo C.A.
        • Helms Tillery S.I.
        The proprioceptive map of the arm is systematic and stable, but idiosyncratic.
        PLoS One. 2011; 6e25214https://doi.org/10.1371/journal.pone.0025214
        • Fornaciai M.
        • Park J.
        Serial dependence generalizes across different stimulus formats, but not different sensory modalities.
        Vis Res. 2019; 160: 108-115https://doi.org/10.1016/j.visres.2019.04.011
        • Manassi M.
        • Liberman A.
        • Kosovicheva A.
        • Zhang K.
        • Whitney D.
        Serial dependence in position occurs at the time of perception.
        Psychon Bull Rev. 2018; 25: 2245-2253https://doi.org/10.3758/s13423-018-1454-5
        • Czoschke S.
        • Fischer C.
        • Beitner J.
        • Kaiser J.
        • Bledowski C.
        Two types of serial dependence in visual working memory.
        Br J Psychol. 2019; 110: 256-267https://doi.org/10.1111/bjop.12349
        • Horsager A.
        • Boynton G.M.
        • Greenberg R.J.
        • Fine I.
        Temporal interactions during paired-electrode stimulation in two retinal prosthesis subjects.
        Invest Ophthalmol Vis Sci. 2011; 52: 549-557https://doi.org/10.1167/iovs.10-5282
        • Kayser S.J.
        • Kayser C.
        Trial by trial dependencies in multisensory perception and their correlates in dynamic brain activity.
        Sci Rep. 2018; 8: 3742https://doi.org/10.1038/s41598-018-22137-8
        • de Lafuente V.
        • Romo R.
        Neuronal correlates of subjective sensory experience.
        Nat Neurosci. 2005; 8: 1698-1703https://doi.org/10.1038/nn1587