Key Words
cochlear implant - electrically evoked compound action potential - spread of excitation
- stimulus polarity
INTRODUCTION
Physiological spread of excitation (SOE) in the electrically stimulated cochlea can
be estimated using the electrically evoked compound action potential (ECAP) in individuals
with cochlear implants (CIs) ([Cohen et al, 2003]; [Abbas et al, 2004]; [Hughes and Stille, 2010]). SOE patterns are typically obtained using a forward-masking paradigm, where the
masker stimulus is roved across the array while the location of the probe and recording
electrode are fixed. With this method, the ECAP amplitude is usually largest when
the masker and probe are delivered to the same electrode. SOE patterns that are obtained
using forward masking reflect the amount of overlap between populations of stimulated
neurons, and can therefore be used to approximate the spatial resolution within the
electrically stimulated cochlea. Studies investigating how SOE patterns relate to
pitch perception have produced mixed results. [Hughes and Abbas (2006)] found no relation between the width of the SOE function and pitch ranking; however,
a follow-up study ([Hughes, 2008]) characterized the ECAP functions in a novel way by calculating the amplitude differences
between pairs of SOE functions, and then summing those differences across all masker
electrodes. Using this method to quantify the spatial separation between SOE functions,
the authors found a significant correlation between greater spatial separation of
SOE patterns and better pitch ranking. When a similar comparison was applied to adjacent
physical electrodes and intermediate channels, however, the relation was generally
positive but not statistically significant ([Goehring et al, 2014a],[b]). It was concluded that although the ECAP SOE has some utility for predicting pitch
ranking for physical electrodes, it is not sensitive enough to predict pitch ranking
for intermediate or virtual channels.
One variable that might affect the relation between physiology and perception is the
polarity of the stimulus. Commercially available CIs use cathodic-leading pulses,
likely because this type of stimulus has been shown to be most effective at stimulating
the auditory nerve in animal models ([Miller et al, 1999]; [Klop et al, 2004]). Conversely, recent physiological evidence in humans using the ECAP suggests that
anodic-leading pulses may be more effective at stimulating the auditory nerve ([Macherey et al, 2008]; [Undurraga et al, 2010]; [2012]; [Hughes et al, 2017]). Results from these studies suggest that more effective stimulation might result
in more effective masking when the forward-masking paradigm is used to derive the
SOE ([Undurraga et al, 2012]). In a recent study using symmetrical biphasic pulses with an interphase gap (Cochlear
devices), [Hughes et al (2017)] found larger ECAP amplitudes using the forward-masking method when the masker and
probe were both anodic leading versus cathodic leading.
With the forward-masking subtraction method ([Abbas et al, 1999]), four stimulus frames are applied: (A) probe alone (to elicit the ECAP), (B) masker
and probe with a short masker–probe interval (typically 400–500 μsec) to isolate the
probe artifact, (C) masker alone (to obtain a recording of any residual neural response
and artifact produced by the masker), and (D) zero-amplitude pulse (to isolate system
artifact). The ECAP is derived by the following formula: A − B + C − D. When the masker
and probe are delivered to the same electrode, both stimuli presumably recruit the
same population of neurons. The response to the masked probe in frame B should therefore
only contain stimulus artifact because the neurons are driven into a refractory state
by the preceding masker. This results in a maximal ECAP amplitude when subtracted
from the probe-alone trace in frame A. When the masker and probe are spatially separated,
the response to the probe is not fully masked in frame B. Fibers responding to the
probe in this case represent neurons that are not recruited by the masker. The partial
ECAP response to the probe in frame B is subtracted from the response in frame A,
resulting in a reduced ECAP amplitude that represents the spatial overlap between
masker and probe electrodes. If anodic-leading pulses result in greater activation
of the auditory nerve (as seen in recent human studies), responses to both the masker
and probe should be enhanced, producing broader regions of excitation. Therefore,
the partial ECAP response to the probe in frame B should be smaller for anodic stimulation
than for cathodic because of more effective masking. In the subtracted trace, ECAPs
should be larger for anodic than for cathodic-leading stimulation, resulting in an
overall broader excitation pattern for anodic-leading stimuli. Using asymmetric pulses
in monopolar mode, [Undurraga et al (2010)] were only able to record ECAPs using anodic maskers, suggesting that stimulation
with this phase resulted in the most effective masking. The goal of the present study
was to investigate the extent to which stimulus polarity affects basic characteristics
of ECAP SOE patterns using symmetrical biphasic pulses that are used clinically. If anodic-leading stimuli produce a more accurate
estimation of the SOE, we may see a stronger relation between physiological SOE patterns
and perceptual measures.
Previous research on polarity sensitivity has shown differing results in animals and
humans. Research on the electrically stimulated auditory nerve in animal models has
shown that action potentials generated by cathodic stimuli typically arise from the
peripheral axon ([Miller et al, 1998]; [1999]; [Rattay et al, 2001a]; [Klop et al, 2004]). In humans, electrophysiological studies with pseudomonophasic pulses have consistently
demonstrated greater sensitivity to the anodic phase, yielding shorter latencies and
larger amplitudes than for cathodic stimulation ([Macherey et al, 2008]; [Undurraga et al, 2010]; [2012]; [2013]). Modeling work by [Rattay et al (2001b)] showed that biphasic cathodic and anodic stimuli are equally effective in generating
action potentials when peripheral processes are present. However, once peripheral
axons have been lost, the cathodic phase requires more current to overcome the unmyelinated
soma for action potential propagation. Because the anodic phase preferentially stimulates
the central axon, less current is needed to generate an action potential. Peripheral
loss in animal experiments tends to be minimized because of acute deafening procedures
([Miller et al, 1998]; [Klop et al, 2004]), whereas peripheral loss tends to be much more extensive for most CI users due
to longer durations of hearing loss and deafness ([Hinojosa and Marion, 1983]; [Nadol, 1997]; [Ng et al, 2000]; [Khan et al, 2005]). These differences likely contributed to conflicting findings between humans and
animals.
Few studies have examined stimulus polarity effects on ECAP responses obtained in
humans using symmetrical biphasic pulses that are used clinically. [Undurraga et al (2010)] found larger ECAP amplitudes and shorter latencies for anodic-leading pulses in
a group of four Advanced Bionics (Valencia, CA) CI recipients. In contrast, [Macherey et al (2008)] reported similar ECAP amplitudes for anodic- and cathodic-leading pulses in a group
of six Advanced Bionics recipients, despite both studies using the same stimulus paradigm.
The lack of consensus between studies suggests the effects of polarity on the auditory
nerve are poorly understood and may be influenced by a number of factors, including
the degree of neural degeneration and differences in suprathreshold versus threshold
responses ([Carlyon et al, 2013]).
To date, only one study has examined the effect of stimulus polarity on ECAP SOE patterns.
[Undurraga et al (2012)] used symmetric cathodic- or anodic-leading biphasic pulses with a long interphase
gap (2.1 msec) presented in bipolar mode to simulate monophasic stimulation. The authors
described the resultant SOE patterns in terms of the centroid of the SOE function
and the SOE width (the number of electrodes between 25% and 75% in the cumulative
SOE function). The results showed that the centroid corresponded to the electrode
that delivered the anodic phase, although the width of the function was not different
between polarities. These results held true for pseudomonophasic stimuli and when
pulses were presented in a narrow bipolar mode (three-electrode separation versus
nine). The authors concluded that the anodic polarity is the most effective phase
of a biphasic pulse, likely due to more effective stimulation and/or masking.
The purpose of the present study was to examine the effects of polarity on basic characteristics
of ECAP SOE patterns using clinically relevant stimuli. Much of the previous work
in humans has been done with stimuli that simulate the monophasic pulses used in most
animal research (e.g., pseudomonophasic pulses or symmetric biphasic pulses with unusually
long interphase gaps). Because these pulse designs are not used clinically, the effects
of polarity remain largely unknown for standard symmetrical biphasic stimuli delivered
in monopolar mode. Based on recent findings that showed the anodic phase of a symmetrical
biphasic pulse is more effective if it is presented first ([Hughes et al, 2017]), we predicted that anodic-leading pulses will produce SOE patterns with larger
overall amplitudes, broader curves (due to stimulation of the central axon), and therefore
less spatial separation between patterns when compared to cathodic-leading pulses.
It was also hypothesized that polarity would have no significant effect on shifting
the peak of the SOE function for monopolar stimulation. For bipolar stimuli, forward-masking
functions can produce dual-peaked functions ([Chatterjee et al, 2006]; [Zhu et al, 2012]). [Undurraga et al (2012)] showed that the centroid of bipolar SOE functions occurred around the electrode
that delivered the anodic phase. Because monopolar stimulation does not tend to produce
forward-masking functions with two peaks, the location of the peak electrode was not
expected to be affected by polarity.
The results from this study will provide the foundation for subsequent investigations
into the role of polarity on the relation between physiological and perceptual measures.
If cathodic-leading symmetrical biphasic pulses produce less effective masking, the
subtraction method used to derive the ECAP will result in smaller amplitudes and narrower
SOE functions. As such, SOE functions for cathodic-leading stimuli might not accurately
represent the actual stimulation patterns that contribute to place-pitch estimates.
MATERIALS AND METHODS
Participants
Data were collected for 16 ears in 15 CI recipients ages 13–77 yr (mean = 55.1 yr).
Six participants were implanted bilaterally. For participant F10/F11, both ears were
tested. For the other five participants with bilateral CIs (F1, F5, F17, FS22, and
N11), only the ear listed in [Table 1] was used because the opposite ear either had an older generation device or very
small/absent ECAPs. All participants were users of newer generation Cochlear Ltd.
(Macquarie, NSW, Australia) devices. Six participants used the Nucleus 24RE(CA), five
participants used the CI422, three participants used the CI512, and one participant
used the CI522 (the CI422 and CI522 are straight arrays, whereas the others are perimodiolar).
Importantly, all devices used the same internal electronics package. Demographic data
for all participants can be found in [Table 1]. This study was approved by the institutional review board of Boys Town National
Research Hospital under protocol 03-07-XP, and written informed consent was obtained
from all participants.
Table 1
Demographic Information for Study Participants
|
Participant
|
Internal Device
|
Ear
|
Duration of Deafness (yr, mo)
|
Age at IS (yr, mo)
|
Duration of CI Use (yr, mo)
|
Etiology/Onset
|
|
F1
|
24RE(CA)
|
L
|
11, 3
|
60, 7
|
10, 2
|
Unknown/progressive
|
|
F2
|
24RE(CA)
|
R
|
10, 6
|
60, 3
|
9, 0
|
Unknown/progressive
|
|
F5
|
24RE(CA)
|
R
|
7, 7
|
48, 3
|
7, 11
|
Unknown/sudden from established HL
|
|
F10
|
24RE(CA)
|
R
|
1, 10
|
1, 10
|
10, 7
|
Waardenburg syndrome/congenital
|
|
F11
|
24RE(CA)
|
L
|
8, 3
|
8, 3
|
17, 0
|
Waardenburg syndrome/congenital
|
|
F17
|
24RE(CA)
|
R
|
11, 0
|
42, 11
|
10, 4
|
Congenital/progressive
|
|
F27
|
24RE(CA)
|
L
|
10, 5
|
56, 2
|
2, 0
|
Otosclerosis/progressive
|
|
N7
|
CI512
|
R
|
10, 11
|
69, 9
|
5, 8
|
Unknown/progressive
|
|
N11
|
CI512
|
L
|
6, 0
|
67, 5
|
5, 0
|
Unknown familial and noise exposure/progressive
|
|
N23
|
CI512
|
R
|
1, 10
|
70, 5
|
2, 7
|
Meniere’s
|
|
FS22
|
CI422
|
L
|
2, 3
|
13, 9
|
3, 2
|
Meningitis/progressive
|
|
FS28
|
CI422
|
R
|
3, 8
|
72, 6
|
1, 10
|
Unknown/progressive
|
|
FS31
|
CI422
|
R
|
30
|
76, 4
|
1, 3
|
Noise induced/hereditary
|
|
FS32
|
CI422
|
L
|
8, 4
|
58, 9
|
3, 2
|
Unknown/hereditary and progressive
|
|
FS33
|
CI422
|
R
|
0*
|
12, 0
|
1, 11
|
Auditory neuropathy spectrum disorder/congenital
|
|
NS20
|
CI522
|
R
|
28, 9
|
31, 9
|
0, 8
|
Illness/unknown
|
Notes: *Participant had mild–moderate hearing loss. L = left; R = right; IS = initial stimulation;
HL = hearing loss.
Equipment Setup
All ECAP measures were made using the commercial Custom Sound EP (v 4.3) software
(Cochlear Ltd., Macquarie, New South Wales, Australia). Stimuli were presented through
a laboratory Freedom sound processor and programming pod.
Stimuli and Procedure
ECAPs were obtained using symmetric, biphasic current pulses presented in monopolar
mode. Both the masker and probe stimuli generally used the following parameters: 25-μsec/phase
pulse duration, 50-dB gain, 400-μsec masker–probe interval, 80-Hz stimulation rate,
and 100 averages. Monopolar stimulation was relative to the extracochlear ball electrode,
MP1, and recording was relative to the extracochlear case electrode, MP2. The intracochlear
recording electrode was fixed two electrode positions apical to the location of the
probe. For some probe electrodes, the recording electrode was three electrodes apical
to the probe to optimize waveform morphology by further reducing stimulus artifact.
Pulse duration was set to 50 μsec/phase for some participants (F10, F11, F17, FS28,
FS32, N5, N7, and N23) due to issues with exceeding voltage compliance (described
further below). The polarity of the leading phase was either cathodic (e.g., 5-MP1)
or anodic (e.g., MP1-5) for both the masker and probe. The traditional forward-masking
subtraction technique was used to remove stimulus artifact from the recording ([Abbas et al, 1999]). SOE patterns were obtained by fixing the probe and recording electrode locations
and roving the masker electrode across the array. All parameters except recording
delay (which was individually adjusted to optimize the ECAP) were kept the same within
a participant across SOE patterns.
An ascending loudness-scaling technique was used to determine stimulation levels for
the SOE measures. Because our goal was to compare SOE measures between polarities
at the same current level (CL), loudness measures were only obtained using the anodic-leading
stimulus because it is typically louder than a cathodic-leading stimulus ([Macherey et al [2006]] for pseudomonophasic pulses; [Macherey et al [2008]] for pseudomonophasic pulses with a long interphase gap; [Carlyon et al [2013]] for triphasic pulses). Five sweeps of the four-frame masker–probe anodic-leading
stimulus were presented to a single electrode beginning at an inaudible level and
then slowly increased in 5-CL steps. Participants were instructed to use a 10-point
loudness-rating scale to indicate when the sound was an “8” (“loud”). This procedure
was repeated until an “8” level had been determined for all 22 electrodes. All masker
and probe stimuli for both leading polarities were presented at the corresponding
“8” level for each electrode obtained using the anodic-leading stimulus. For one participant
with a mild developmental delay (FS33), the reported “8” levels were much lower than
those from a previous session, yielding ECAP amplitudes that were difficult to distinguish
from the noise floor. Stimulation levels were therefore set between the “8” levels
given at the two visits, and the participant did not indicate that any of the stimuli
were too loud. For some participants who exceeded voltage compliance before an “8”
level could be reached (F17, FS31, N23, F2, and NS20), stimuli were presented at an
equal loudness level across the array that was just below voltage compliance limits.
If ECAPs were not large enough to interpret just below compliance limits at a pulse
duration of 25 μsec/phase, a pulse duration of 50 μsec/phase was used for the participants
previously mentioned. For each participant, SOE curves were collected for 14 probe
electrodes for both polarities. This resulted in a total of 28 SOE functions per participant.
Three electrode regions were tested. Regions were designated as basal (electrodes
5–9), middle (9–13), or apical (14–18). For one participant (F2), electrodes 3–7 were
used for the basal set due to an open circuit on electrode 8. Electrodes 13–17 were
used for the apical set for this participant to maintain 14 distinct probe electrodes
per participant. For each region, the middle electrode in the set (e7 for basal, e11
for middle, and e16 for apical) was used as a reference electrode for comparison to
the other four electrodes to calculate the spatial separation between probe-electrode
SOE patterns within each polarity. For example, in the basal group the reference electrode
was 7, so comparison pairs were 5–7, 6–7, 8–7, and 9–7 ([Figure 1E]). These pairs were chosen because they will also be used for a subsequent study
comparing the effects of stimulus polarity on the relation between physiological spatial
separation and pitch ranking.
Figure 1 Schematic of outcome metrics. Each curve represents a hypothetical SOE pattern. (A)
Mean amplitude was calculated by averaging ECAP responses across all masker electrodes
(black circles), excluding the recording electrode (empty space). (B) Peak electrode
was the masker electrode number that generated the largest ECAP amplitude across each
SOE function (black star). (C) Area under the curve (sum of vertical lines) was calculated
as the cumulative sum of normalized ECAP amplitudes in the function after interpolating
for the recording electrode. (D) Spatial separation between SOE functions was calculated
as the sum of the absolute value of the difference in normalized amplitude between
paired functions for each masker electrode location (sum of vertical lines). (E) Depiction
of the electrode array showing probe pairs used to calculate spatial separation. The
three regions are denoted by three sets of arrows. Reference electrodes (e7, e11,
and e16) are shown using black rectangles. Arrows point to the other electrodes (shown
using gray rectangles) that are compared to these reference electrodes.
Data Analysis
ECAP amplitudes were calculated as the difference between the first negative trough
(N1) and the following positive peak or plateau (P2). These peaks are automatically
marked in the Custom Sound EP software; however, all peak markers were verified or
adjusted as necessary by the investigators. Data were then exported and processed
using custom Matlab scripts (Mathworks, Natick, MA). The following calculations were
made for each SOE function (see [Figure 1] for a schematic of each metric):
-
A) Mean ECAP amplitude: For each probe-electrode SOE function, amplitudes were averaged
across all masker electrodes (excluding the recording electrode) to obtain the mean
ECAP amplitude. [Figure 1A] shows a schematic of the 21 ECAP amplitudes that are averaged together to generate
the mean. Mean amplitudes were compared between polarities to determine if one polarity
yielded larger average amplitudes than the other.
-
B) Electrode location of the peak of the SOE function: The peak of the SOE function
was the masker electrode number that generated the largest ECAP amplitude across each
SOE function. [Figure 1B] shows a schematic of the location of the peak (indicated by the star). The location
of the peak was compared between SOE patterns produced by anodic- and cathodic-leading
stimuli to determine whether polarity shifted the location of the peak of the function.
-
C) Area under the SOE curve: In most cases, the ECAP cannot be recorded from the same
electrode that provides the stimulus because of excessive artifact (typically amplifier
saturation). Because the location of the recording electrode differed in some cases,
and measures from all electrodes are needed to determine the area under the SOE curve
using our earlier method ([Hughes, 2008]), the amplitude of the ECAP from the recording electrode was estimated using linear
interpolation (also for the open-circuit electrode for participant F2). This was done
by averaging the two ECAP amplitudes for masker electrodes directly adjacent on either
side of the recording electrode. (Although others [e.g., [Cohen et al, 2003]] have used more elaborate curve-fitting methods to estimate values for the recording
electrode [or other missing data points], we feel a simple interpolation is sufficient
for the measures of interest here.) Next, all ECAP amplitudes were normalized to the
peak amplitude within each SOE function to control for overall amplitude differences
across participants, electrodes, and polarities so that the area-under-the-curve measure
more directly reflects the breadth of each function. The area under the curve was
calculated as the cumulative amplitude of all normalized ECAP amplitudes in the function.
(Note this is not a true “area” calculation; however, we will use the term “area under
the curve” throughout this report for simplicity.) [Figure 1C] depicts a schematic of the data points used to determine area under the curve. This
metric was used in favor of measuring the width of the curve at a designated down
point because it is less influenced by asymmetric curves ([Hughes and Abbas, 2006]). The areas under the curve for SOE functions obtained with anodic-leading versus
cathodic-leading stimuli were compared to determine if polarity systematically affected
the overall breadth of the function.
-
D) Spatial separation between SOE functions: The goal of this metric was to determine
whether the spatial separation between pairs of cathodic-leading SOE functions differed
from that between pairs of anodic-leading SOE functions. To quantify the spatial separation
between pairs of SOE functions within each polarity, ECAP amplitudes were first normalized
to the single peak amplitude across both functions within each comparison pair. This
method preserves relative amplitude differences between comparison pairs (see [Hughes, 2008]), but avoids the confound of overall amplitude differences between polarities. Because
ECAP amplitudes are generally larger for anodic-leading than for cathodic-leading
stimuli ([Macherey et al, 2008]; [Undurraga et al, 2010]; [Hughes et al, 2017]), the spatial separation between anodic pairs versus cathodic pairs will be affected
by the raw amplitude differences between polarities if the data are not normalized.
The absolute value of the difference in normalized amplitude between paired functions
for each masker electrode location was then calculated ([Hughes, 2008]; [Hughes et al, 2013]; [Goehring et al, 2014a]). These differences were summed together to yield the spatial separation index,
Σ. [Figure 1D] shows a schematic of this calculation. In total, there were 12 comparison pairs
per participant, with 4 pairs in each region (basal, middle, and apical). Each electrode
was compared to the reference electrode in the center of that region, as depicted
in [Figure 1E]. The Σ values for all anodic-leading SOE pairs were compared to those for the respective
cathodic-leading SOE pairs to determine if polarity affected the amount of spatial
separation between functions.
Statistical analyses were performed in SigmaPlot (v. 12.5; Systat Software, San Jose,
CA). For each outcome measure (amplitude, peak location, area, and spatial separation),
a nonparametric version of the paired Student’s t test (Wilcoxon signed-rank test) was used to compare measures made with anodic- versus
cathodic-leading stimuli. A nonparametric test was used because the data were not
normally distributed. For nonparametric tests, medians are typically reported in lieu
of means because means can be skewed when the data do not follow a normal distribution.
We have, however, reported both medians and means where possible. Last, electrode
was not considered as a factor in the analysis because variations in insertion depth
and electrode-modiolar distance (particularly for perimodiolar arrays) are large and
unsystematic across individuals ([Saunders et al, 2002]), which precludes clear interpretation of any potential statistical effects of electrode
as a factor.
RESULTS
[Figure 2] shows ECAP waveforms for anodic-leading (left) and cathodic-leading (right) stimuli.
Data are from probe e10 in participant N23. Masker electrode numbers are indicated
at the right side of the figure in an apical (top) to basal (bottom) direction. Bolded
waveforms represent the ECAP obtained with the masker and probe on the same electrode
(e10). The corresponding SOE patterns are shown as an inset at the bottom right of
the figure. In this example, ECAP amplitudes were larger overall for anodic-leading
(white circles in the inset) than for cathodic-leading (black circles) stimuli. ECAPs
were obtained for more masker electrodes for the anodic-leading polarity than for
cathodic, leading to a broader SOE pattern. The peak of the SOE function obtained
with anodic-leading stimuli occurred at e10 (which was also the probe electrode),
whereas the peak for the cathodic-leading condition occurred at e11.
Figure 2 Example waveforms and SOE patterns (figure inset) for participant N23. ECAP responses
obtained for anodic- (left column) and cathodic-leading (right column) pulses with
the probe fixed on electrode 10. Masker electrodes are listed at the far right from
apical (top) to basal (bottom). The bolded waveforms represent the masker and probe
on the same electrode. The SOE patterns derived from these waveforms are shown in
the inset at the bottom right corner. Peak-to-peak amplitudes are plotted as a function
of masker electrode from left (basal) to right (apical). White symbols, anodic-leading
pulses; black symbols, cathodic-leading pulses.
[Figure 3] shows the mean ECAP amplitudes for anodic-leading compared to cathodic-leading stimuli.
For each of the 16 ears, SOE functions for 14 probe electrodes ([Figure 1E]) were collected for each polarity. The solid diagonal line represents equal amplitude
for both polarities. Data points that fall above the line represent probe electrodes
that exhibited larger amplitudes for anodic-leading stimuli. Although most raw ECAP
amplitudes were <100 μV, some participants had relatively large mean amplitudes (∼300–500
μV). For all but the largest ECAP responses, anodic-leading stimuli generally produced
larger amplitudes. A Wilcoxon signed-rank test indicated that the median ECAP amplitude
for anodic-leading (77.9 μV) pulses was significantly larger than the median amplitude
for cathodic-leading (38.0 μV) pulses (z = −11.704, p < 0.001). The mean ECAP amplitudes for anodic- and cathodic-leading stimuli were
124.1 and 86.7 μV, respectively. This result confirmed the hypothesis that anodic-leading
pulses produce larger ECAP responses.
Figure 3 ECAP amplitudes for anodic-leading compared to cathodic-leading stimuli. Each dot
represents the mean ECAP amplitude across the SOE pattern for one probe electrode
for one participant (14 probes per ear; 224 total data points). The diagonal line
denotes equal amplitudes for both polarities. Data points above the diagonal line
represent larger amplitudes for anodic-leading stimuli.
The top panel of [Figure 4] shows the peak electrode location compared between polarities. Each ear contributed
14 SOE functions to the data analysis. The solid diagonal line represents the same
peak location for both polarities. Note that some of the data points overlie each
other. There was no significant effect of polarity on peak electrode location (Wilcoxon
signed-rank test, z = 1.464, p = 0.144), as expected. The bottom panel of [Figure 4] is a bubble plot that indicates the degree to which the peak was shifted relative
to the location of the fixed probe. In this figure, negative shifts reflect peaks
that occurred basal to the probe and positive shifts reflect peaks that occurred apical
to the probe. The size of the bubbles reflects the number of SOE patterns at each
coordinate. The peak occurred at the same electrode for both polarities for slightly
more than half (114/224) of the pairs of SOE functions. For 66 of these pairs, the
peak occurred at the probe electrode (zero-electrode shift on both axes). For 27 pairs,
the peak was shifted one electrode position apically (+1) for both polarities; for
17 pairs, the peak was shifted one electrode position basally (−1) for both polarities;
and for 4 pairs, the peak was shifted two electrode positions basally (−2). The eight
smallest bubbles each represent one cathodic/anodic pair of SOE functions. The minimum
shift in peak electrode location from the probe electrode location was zero electrodes
for both polarities. The maximum shift was −3 to +1 electrodes for the anodic-leading
stimulus and −6 to +4 electrodes for the cathodic-leading stimulus. The average shift
in electrode location was 0.34 electrodes for the anodic polarity and 0.83 electrodes
for the cathodic. In summary, stimulus polarity did not shift the peak of the SOE
function in any predictable way.
Figure 4 Top: Electrode location of the peak of the SOE function for anodic-leading compared
to cathodic-leading stimuli. Each dot represents the peak electrode location of an
SOE pattern for one probe electrode for one participant (14 per ear; 224 total data
points). The diagonal line denotes equal peak electrodes for both polarities. Bottom:
Bubble plot indicating the degree to which the peak of the SOE function shifted relative
to the probe electrode. Negative and positive numbers represent the number of electrodes
by which the peak shifted in the basal or apical direction, respectively. Bubble size
corresponds to the number of occurrences at each coordinate, where the largest bubble
represents 66 SOE functions and the smallest eight bubbles each represent one SOE
function.
[Figure 5] shows six individual examples of pairs of SOE patterns obtained with each leading
polarity. In each panel, both SOE patterns have been normalized to the peak of each
respective function so that the area under each curve could subsequently be calculated
and compared. These examples show the range of patterns that were observed. Participant
numbers and probe electrodes are indicated in each panel. For participants F1, FS22,
and FS33, the SOE function obtained with the anodic-leading polarity yielded broader
patterns on both the apical and basal sides of the function. For F2 (P10), the cathodic-leading
polarity yielded a broader function on the apical side of the peak. In contrast, the
anodic-leading polarity yielded a broader function on the basal side of the peak for
FS32. Finally, virtually no polarity effects were demonstrated for probe e5 in participant
F2.
Figure 5 Individual examples of pairs of normalized SOE patterns obtained with anodic-leading
(open circles) and cathodic-leading (filled circles) stimuli. Each pattern has been
normalized to the peak of its own function to allow for comparisons of area under
the curve while controlling for overall amplitude differences. Participant number
and probe electrode are indicated on each panel.
[Figure 6] shows the mean curve areas for normalized SOE functions produced by anodic- and
cathodic-leading pulses. The data represent SOE functions from 14 probe electrodes
per ear. The solid diagonal line represents equivalence between polarities. Data points
that fall above the diagonal line represent broader functions (larger curve areas)
for anodic-leading stimuli. The median curve area was significantly greater for anodic-leading
(8.9) than for cathodic-leading (8.2) stimuli (Wilcoxon signed-rank test, z = −8.358, p < 0.001). Mean curve areas were 9.1 for anodic-leading and 8.5 for cathodic-leading
stimuli. This result is consistent with the hypothesis that anodic-leading pulses
produce broader excitation patterns than cathodic-leading stimuli.
Figure 6 Area under the curve (see text) for normalized SOE functions obtained with each polarity.
Each dot represents an SOE pattern for one probe electrode for one ear (14 per ear;
224 total data points). The diagonal line denotes equal curve areas for both polarities.
Data points above the diagonal represent larger curve areas (broader SOE patterns)
for anodic-leading stimuli.
[Figure 7] shows the calculated spatial separation index, Σ, for pairs of SOE functions obtained
with cathodic-leading (abscissa) and anodic-leading (ordinate) stimuli. The 192 data
points represent Σ values from 12 SOE probe pairs per ear. As expected, more data
points fell below the diagonal line, indicating less spatial separation between comparison
pairs for anodic-leading than for cathodic-leading stimuli. A Wilcoxon signed-rank
test showed that the median spatial separation for cathodic-leading pulses (2.51)
was significantly larger (z = 3.16, p = 0.002) than for anodic-leading stimuli (2.43). The mean cathodic-leading spatial
separation (2.8) was also slightly larger than the mean anodic-leading spatial separation
(2.6). These results confirm the hypothesis that more effective masking by anodic
stimuli leads to larger (i.e., broader) curve area and thus less spatial overlap between
SOE patterns for anodic-leading stimuli.
Figure 7 Spatial separation between SOE functions (normalized by comparison pairs). Each dot
represents the spatial separation between two SOE patterns for one ear (12 pairs per
ear; 192 total data points). The diagonal line denotes equal peak electrodes for both
polarities. Data points below the diagonal represent greater spatial separation for
cathodic-leading stimuli.
DISCUSSION
In this study, ECAP SOE patterns for cathodic-leading symmetrical biphasic pulses
were compared to those for anodic-leading pulses to examine the effects of polarity
on basic characteristics of SOE patterns using clinically relevant stimuli. In general,
anodic-leading pulses produced larger ECAP amplitudes, broader SOE patterns, and consequently
less spatial separation between functions. These results appear to support existing
evidence that the anodic phase preferentially and more effectively stimulates the
central axon in the deafened ear. When the anodic phase is presented first as part
of a biphasic pulse, it produces greater masking than when presented immediately following
a cathodic phase. Finally, polarity appears to have no consistent effect on the location
of the peak electrode in the SOE curve.
Amplitudes
Larger ECAP amplitudes were generally observed for the anodic-leading polarity compared
with the cathodic-leading polarity ([Figure 3]), consistent with the findings of [Undurraga et al (2010)] for symmetrical biphasic pulses. This supports the central axon as the site of excitation,
as anodic stimuli preferentially stimulated the axon in the degenerated auditory nerve
models described by [Rattay et al (2001a],[b)]. Furthermore, auditory neurons are more tightly bundled at the central axon, which
may lead to greater SOE for direct axonal stimulation ([Rattay et al, 2001b]).
Although the stimuli for both leading polarities were presented at the same CL, the
respective loudness levels might not have been equal between polarities. Indeed, anodic
pulses have been shown to be louder than cathodic pulses at equal CL for triphasic
([Carlyon et al, 2013]) and pseudomonophasic ([Macherey et al, 2006]; [2008]; [Undurraga et al, 2013]) pulse shapes. Louder percepts could result from recruitment of a larger population
of neurons (i.e., from axonal stimulation), resulting in larger ECAP responses and
broader SOE patterns than for softer percepts. If the CL of the anodic-leading stimulus
is reduced to match the loudness of the cathodic-leading stimulus, then the anodic
SOE patterns would likely have smaller amplitudes and narrower widths at these lower
levels ([Hughes and Stille, 2010]). Therefore, SOE patterns might look similar for both polarities when CLs are set
according to similar loudness ratings (as is done during clinical programming). If
the cathodic-leading polarity truly produces less effective masking ([Undurraga et al, 2012]), then the breadth of the cathodic SOE obtained with present measurement techniques
might be underestimated. Consequently, at equal loudness levels, it is possible that
the spatial excitation patterns might be different. It is unclear what effect polarity
would have on speech perception using contemporary processing strategies.
Poor nerve survival, as measured by spiral ganglion cell counts, has been shown in
individuals with severe–profound hearing loss ([Kawano et al, 1998]; [Nadol et al, 2001]), such as the CI recipients who participated in this study. Spiral ganglion cells
counts have also been shown to be reduced for older individuals and for longer durations
of deafness ([Nadol et al, 1989]). Half of the participants in this study (8 of 15) were >60 yr of age at the time
of participation, and almost half (7 of 15) had durations of deafness >10 yr before
implantation ([Table 1]). Collectively, the demographic factors for most of the participants in this study
would suggest some loss of peripheral processes; however, we have no clear way to
measure auditory nerve survival in living humans.
Peak Location
The present study found no consistent effect of polarity on the location of the peak
electrode, consistent with the hypothesis ([Figure 4]). The peak occurred at the same electrode for both polarities for slightly more
than half (114/224) of the pairs of SOE functions. For the remaining pairs, there
was no systematic shift in the peak of the SOE function across polarities. Physiological
([Cohen et al, 2003]; [Abbas et al, 2004]; [Hughes and Stille, 2010]; [Undurraga et al, 2012]) and perceptual ([Kwon and van den Honert, 2006]; [Hughes and Stille, 2008]; [Nelson et al, 2008]; [Macherey et al, 2010]) forward-masking functions obtained with symmetrical, cathodic-leading, biphasic
pulses delivered in monopolar mode typically result in functions with a single peak
at or near the location where the masker and probe are delivered to the same electrode.
However, these results may have been different if polarity was manipulated. Different
degrees of degeneration in the populations of neurons responding to each polarity
may affect the location of the peak of the SOE function. For these reasons, we did
not have a clear prediction as to whether polarity would have a notable effect on
the location of the peak of the SOE functions. To date, no other studies have systematically
evaluated the effects of stimulus polarity on forward-masking patterns for monopolar
stimulation using symmetrical biphasic pulses. However, with bipolar stimulation (particularly
with wide spacings), dual-peaked functions have been obtained ([Chatterjee et al, 2006]; [Nelson et al, 2008]; [Zhu et al, 2012]). For ECAP recordings using a wide bipolar stimulation mode (BP+9), [Undurraga et al (2012)] found that the centroid of the SOE function occurred closest to the electrode delivering
the anodic phase, suggesting that the anodic phase provided the most masking. Given
that the present study used monopolar stimulation, it was not surprising that the
peak electrode location did not change with polarity.
Area under the Curve
Anodic-leading pulses generally produced broader SOE patterns than cathodic-leading
pulses ([Figures 5] and [6]). This result is consistent with the hypothesis that the anodic-leading polarity
likely provides more effective masking than the cathodic polarity, and therefore larger
ECAP amplitudes across the entire SOE pattern. However, it should be noted that it
is difficult to determine whether greater excitation is in response to the masker,
probe, or both. To fully elucidate the polarity effects of the masker and probe, polarity
would need to be assessed separately for the masker and probe (e.g., anodic-leading
masker with cathodic-leading and anodic-leading probes). To isolate masker polarity
effects, [Macherey et al (2008)] and [Undurraga et al (2010)] used symmetrical biphasic pulses with long interphase gaps to effectively create
a monophasic masker using the second phase of the pulse. Both studies found that ECAP
responses could only be obtained in response to an anodic masker; the cathodic masker
did not produce ECAPs. In some cases, the cathodic masker even appeared to produce
facilitation rather than masking. With more effective masking provided by the anodic
phase, larger ECAPs will result. When the forward-masking paradigm is applied, the
result is a broader SOE pattern, as observed in the present study.
In contrast to the present findings, [Undurraga et al (2012)] found that the width of the SOE function was not different between polarities. There
are several notable stimulus differences between their study and the present study,
however. First, the probe stimulus in their study was always a cathodic-leading symmetrical
biphasic pulse, while the polarity and configuration (monopolar/bipolar) of the masker
was varied. In our study, the masker and probe were identical stimuli, and both reversed
in polarity together. Additionally, they used several different configurations of
masker pulses (symmetrical biphasic in monopolar and bipolar mode; and pseudomonophasic
and symmetrical biphasic with a long interphase gap in only bipolar mode), but did
not report polarity effects on the SOE width for the monopolar condition (most similar
to our study). Last, their results were based on data from only 5 participants, whereas
the present study had 15 participants (16 ears). To our knowledge, this is the first
study to assess polarity effects on ECAP SOE patterns using clinically relevant stimuli.
Spatial Separation
When comparing SOE patterns for two spatially separated probe electrodes, the larger,
broader patterns found with anodic-leading pulses lead to greater overlap of the respective
SOE patterns, and thereby yield less spatial separation between probe electrodes.
Broader patterns appear to be the product of more effective masking of the neural
response to the probe for anodic stimuli, likely due to stimulation of the central
axon of auditory nerve fibers ([Macherey et al, 2008]; [Undurraga et al, 2010]). It is possible that the narrower SOE patterns (and thus more spatial separation)
observed with cathodic-leading stimuli are due to less effective masking with the
forward-masking subtraction method. If a cathodic-leading stimulus is less effective
than an anodic-leading stimulus, then it would also be a less effective masker. Less
effective masking would result in more neurons responding to the probe in the masked
probe condition (i.e., a larger masked response). When subtracted from the probe-alone
trace, the result would be a smaller ECAP than would otherwise be derived if the masker
were more effective. If the raw ECAP amplitudes are smaller across the entire SOE
function for the cathodic-leading condition than for the anodic-leading condition
(as was demonstrated in [Figure 3]), then the resulting SOE pattern for the cathodic-leading condition will likely
be narrower. This effect can be seen in the inset panel in [Figure 2], which shows the larger amplitude SOE (open circles) for anodic-leading stimuli,
along with the smaller-amplitude SOE (filled circles) for cathodic-leading stimuli.
The smaller overall amplitudes for the cathodic condition result in the function falling
to 0 μV at masker electrodes 5 and 21 (producing a narrower SOE pattern), whereas
this occurs at masker electrodes 3 and 22 for the anodic function (producing a broader
SOE pattern). As a result, smaller ECAPs lead to generally narrower SOE patterns,
and therefore greater spatial separation between curves is expected. As noted above,
however, it is unclear whether the differences in spatial separation arise primarily
from polarity effects of the masker, probe, or both.
We must also consider the effects of neural health in the context of the present results.
CI recipients typically have severe-to-profound hearing loss, which tends to be accompanied
by substantial degeneration or loss of peripheral processes in the auditory nerve.
However, individuals with greater degrees of residual hearing are receiving CIs. The
models of [Rattay et al (2001b)] suggest that biphasic pulses of both polarities are equally effective in generating
action potentials when peripheral processes are present. Therefore, if the SOE patterns
for each leading polarity are similar, this could potentially indicate regions of
greater neural survival.
Limitations and Future Directions
A general limitation of the forward-masking method used to measure the SOE pattern
is that it does not differentiate between excitation due to the masker versus the
probe. Instead, it exploits some degree of spatial overlap between two electrodes.
Therefore, in our study, it is not possible to say that the effects of polarity on
ECAP amplitude, curve area, and spatial separation are due exclusively to the masker.
Rather, they are likely due to some combination of excitation from the masker and
the probe. In order to elucidate effects of the masker, other experimental conditions
where the masker and probe differ in polarity (e.g., masker cathodic-leading, probe
anodic-leading) would need to be carried out. For symmetric biphasic pulses with long
interphase gaps, [Undurraga et al (2010)] found that ECAPs could only be measured when the masker was anodic leading, regardless
of whether the probe was anodic or cathodic leading.
Two studies have proposed other methods to more accurately estimate SOE patterns.
[Biesheuvel et al (2016)] used a deconvolution method to separate excitation areas generated by the masker
and the probe. These areas, or excitation density profiles, were highly correlated
with SOE curves measured using the forward-masking method when the modeled excitation
patterns were exponential or Gaussian (as opposed to rectangular). This relationship
was diminished for normalized SOE curves, suggesting normalization may not be necessary
for deconvolution. [Cosentino et al (2016)] described a “panoramic” multistage optimization approach intended to model SOE,
as well as identify neural dead regions and cross-turn stimulation that could be unaccounted
for in the forward-masking method. The authors compared their model to the forward-masking
method for both simulated and human ECAP data. The model was found to be reliable
with minimal error, as long as the signal-to-noise ratio was >5 dB. While both reports
claim to correctly model SOE, future studies are needed to compare these methods against
each other, as well as to assess SOE patterns obtained with anodic-leading stimuli.
Perhaps a combination of anodic stimuli and improvements in the forward-masking method
for measuring SOE would improve the ability of the ECAP SOE to better predict electrode
discrimination on the basis of pitch.
CONCLUSIONS
Results from this study show that anodic-leading, symmetrical, biphasic pulses produce
larger ECAP amplitudes and broader excitation patterns compared with cathodic-leading
pulses when equal CLs are used. These combined factors reduce the spatial separation
between SOE functions obtained for different electrodes, and are likely a result of
more effective masking. Further research is needed to determine whether the spatial
separation between SOE patterns obtained with anodic-leading pulses better predicts
pitch discrimination between electrodes. Our work also has implications for clinical
practice. Given that anodic-leading pulses more effectively stimulate the auditory
nerve, future versions of software may allow clinicians to choose this type of stimulus
in coding strategies. Because less current is needed to achieve similar loudness levels
to cathodic stimulation ([Carlyon et al, 2013]), recipients may benefit from longer battery life and reduced nonauditory perceptions
if anodic-leading stimuli are used.
Abbreviations
CI:
cochlear implant
CL:
current level
ECAP:
electrically evoked compound action potential
SOE:
spread of excitation
