University of Zurich Zurich Open Repository and Archive Winterthurerstr. 190 CH-8057 Zurich http://www.zora.uzh.ch Year: 2009 TNRT profiles with the Nucleus Research Platform 8 system Lai, W K; Dillier, N; Weber, B P; Lenarz, T; Battmer, R; Gantz, B; Brown, C; Cohen, N; Waltzman, S; Skinner, M; Holden, L; Cowan, R; Busby, P; Killian, M Lai, W K; Dillier, N; Weber, B P; Lenarz, T; Battmer, R; Gantz, B; Brown, C; Cohen, N; Waltzman, S; Skinner, M; Holden, L; Cowan, R; Busby, P; Killian, M (2009). TNRT profiles with the Nucleus Research Platform 8 system. International Journal of Audiology, (9:1-10):Epub ahead of print. Postprint available at: http://www.zora.uzh.ch Posted at the Zurich Open Repository and Archive, University of Zurich. http://www.zora.uzh.ch Originally published at: International Journal of Audiology 2009, (9:1-10):Epub ahead of print. TNRT profiles with the Nucleus Research Platform 8 system Abstract This study investigates the effect of the Nucleus CI24RE implant's neural response telemetry (NRT) system, which has less internal noise compared to its predecessor, the CI24M/R implant, on the NRT threshold (TNRT) profile across the array. CI24M/R measurements were simulated by ignoring CI24RE measurements with response amplitudes below 50 uV. Comparisons of the estimated TNRTs from the CI24RE measurements and the CI24M/R simulations suggest that, apart from a constant level difference, the TNRT profiles from the newer implant generally would not have differed very much from those of its predecessor. This view was also reflected by principal component analysis (PCA) results which revealed a 'shift' component similar to that reported by Smoorenburg et al (2002). On the whole, there is no indication that current practices of using the TNRT profiles for assisting with speech processor programming need to be revised for the CI24RE implant. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 1 of 32 TNRT profiles with the Nucleus Research Platform 8 system a a a b b c c d Lai WK , Dillier N , Weber BP , Lenarz T , Battmer R , Gantz B , Brown C , Cohen N , d e e f f Waltzman S , Skinner M , Holden L , Cowan R , Busby P , Killian M g a. Dept. of Otorhinolaryngology, University Hospital, Zürich, Switzerland b. Dept. of Otolaryngology, Medical University of Hannover, Germany c. Dept. of Otolaryngology, Head and Neck Surgery, University of Iowa, Iowa City, IO, USA d. Dept. of Otolaryngology, New York University Medical Center, NY, USA e. Dept. of Otolaryngology-Head & Neck Surgery, Washington University School of Medicine, St Louis, MO, USA f. Co-operative Research Centre for Cochlear Implant & Hearing Aid Innovation, Melbourne and Sydney, Australia g. Cochlear Technology Center, Mechelen, Belgium Keywords Cochlear implant, CI24RE, Neural Response Telemetry, Evoked Compound Action Potential, TNRT Profile, Principal Component Analysis Acroynms & Abbreviations ECAP Evoked Compound Action Potential RP8 Research Platform 8 NPE Nucleus Programming Enviroment DSP Digital Signal Processor NRT Neural Response Telemetry TNRT Neural Response Telemetry Threshold AGF Amplitude Growth Function PCA Principal Component Analysis Corresponding author: Wai Kong Lai ENT Department University Hospital Zürich Frauenklinikstr 24 CH-8091 Zürich Switzerland e-mail: waikong.lai@usz.ch Lai et al.: TNRT profiles with the Nucleus RP8 system Page 2 of 32 Abstract This study investigates the effect of the Nucleus CI24RE implant's neural response telemetry (NRT) system, which has less internal noise compared to its predecessor CI24M/R implant, on the NRT Threshold (TNRT) profile across the array. CI24M/R measurements were simulated by ignoring CI24RE measurements with response amplitudes below 50uV. Comparisons of the estimated TNRTs from the CI24RE measurements and the CI24M/R simulations suggest that, apart from a constant level difference, the TNRT profiles from the newer implant generally would not have differed very much from those of its predecessor. This view was also reflected by principal component analysis (PCA) results which revealed a "shift" component similar to that reported by Smoorenburg et al. (2002). On the whole, there is no indication that current practices of using the TNRT profiles for assisting with speech processor programming need to be revised for the CI24RE implant. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 3 of 32 Introduction The Nucleus CI24M cochlear implant introduced in 1996 features a self-contained system for intracochlear measurements of the electrically evoked compound action potential (ECAP), whose measurement data is sent via back-telemetry to the controlling Neural Response Telemetry (NRT) software. Although the measurement system is easy and reliable to use, it is not without its shortcomings. The measurement amplifier has a noise floor of around 30-40uV after averaging (Brown et al. 2000), limiting the accuracy in determining the ECAP threshold. Additionally, the amplifier gain can only be reduced from 60dB to 40dB when avoiding saturation that would otherwise arise from excessively large inputs, at the cost of a correspondingly poorer signal to noise ratio (SNR). The nominal temporal resolution of the measured neural response was increased by using a doublebuffer scheme which in turn introduced non-linearity complexities when combining the two lowresolution buffers (Lai 2004). The CI24R implant, introduced in 2000 and featuring a smaller receiver/stimulator package as well as a perimodiolar (Contour) electrode array, used the same telemetry circuitry for measuring the ECAP. The Nucleus Research Platform 8 (RP8) system consisting of the CI24RE cochlear implant, the L34 body-worn speech processor, and the Nucleus Programming Environment (NPE) software (Battmer et al. 2004), was developed specifically for research purposes, such as investigating the effect of an improved measurement amplifier. The CI24RE implant's measurement amplifier has a lower noise floor for the same amount of averaging (around 2uV (Patrick et al. 2006)), as well as improved linearity which reduces the susceptibility to saturation. The difference in the measurement noise between the CI24M/R and the CI24RE implants is illustrated in Figure 1. The lower noise floor results in faster measurements, since less averaging is needed to obtain a clear recording of the neural response. In addition, finer Gain adjustment settings allow better optimization of the measurements to maximize the SNR. The sampling rate for NRT recordings has also been increased, allowing higher temporal resolution for the measured neural response without having to resort to additional manipulation. The neural response threshold (TNRT) is usually estimated using either linear extrapolation of the amplitude growth function (AGF), or visual detection of the smallest measurable neural response. The Lai et al.: TNRT profiles with the Nucleus RP8 system Page 4 of 32 reduced noise floor of the CI24RE implant allows neural responses with smaller amplitudes to be recorded than otherwise possible with the CI24M/R implant. The result should be more accurate estimates of TNRT from either extending the lower portion of the AGF, or a lower visual threshold. Either way, the TNRTs derived from measurements using the CI24RE implant are expected to be different compared to those from the previous CI24M/R implant. Although the TNRT does not correspond directly to the perceptual threshold, modest correlations between TNRT and map T & C values for the CI24M/R implant have been reported by various studies (e.g. Brown et al. 2000, Hughes et al. 2000, Franck & Norton 2001, Smoorenburg et al. 2002, Cafarelli-Dees et al. 2005, King et al. 2006, Potts et al. 2007). These have encouraged the increasingly widespread use of TNRT profiles in assisting with speech processor programming, especially for very young children (e.g. Hughes et al. 2000, Mason et al. 2001, Gordon et al. 2002, Kaplan-Neeman et al. 2004) as well as the elderly (e.g. Pedley et al. 2007). Differences in TNRT between the CI24RE and CI24M/R implants would thus lead to differences in their corresponding TNRT profiles as well. Unless the changes in TNRT values were uniform across the electrode array, resulting in only a level shift in the TNRT profiles, this could change the correlations between the TNRT profiles and map T & C values. This, in turn, could then affect the way the CI24RE implant's TNRT profiles are applied to speech processor programming. There is therefore a clinical interest to know whether the relationship between TNRT and the map T & C values would be affected by the (more accurate) TNRT values obtained with the CI24RE implant. This study is therefore aimed at examining the relationship between the CI24RE implant's TNRT values with corresponding map T & C values, and to investigate whether the CI24RE implant's TNRT profiles differ from those of its predecessor CI24M/R. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 5 of 32 Method Amplitude growth functions (AGFs) from 5 electrodes (typically e20, e15, e10, e5 and e3) spread across the intracochlear electrode array were obtained postoperatively for each participating CI user. All NRT recordings were made using the Forward Masking artefact-cancellation paradigm, each averaged over 50 sweeps. The measurement amplifier's Gain was kept high (60dB) whenever possible (whilst avoiding amplifier saturation) in order to maximize the SNR. The recording site was always located 2 electrodes basal to the stimulation site (e.g. stimulation on electrode 20, recording on electrode 18). The Masker pulse was always 10 Current Level units larger than the Probe pulse. The stimulation levels were reduced in 5 Current Level steps to generate the AGF, beginning at the Loudest Acceptable Presentation Level (LAPL) as indicated subjectively by the CI user. A stimulation rate of 80Hz was used by default, but this could be reduced to 35Hz when necessary to yield a higher stimulation level without exceeding LAPL. This reduction was only necessary with 1 single subject. The TNRT values were then estimated from the AGF by linear regression extrapolation. The AGFs were collected at regular 4 week intervals after and including the initial tune-up session (denoted as week 0). The data collection was terminated at Week 12 after tune-up. The Map T & C level data for the same electrodes were also collected during the same sessions so as to be able to compare them directly against their corresponding TNRT counterparts. Note that the AGFs at Week 0 (tune-up) were indeed based on data collected on that same day rather than from intraoperative data. ACE maps with a stimulation rate of 1200 pps per channel were used for the entire duration of the study. One of the aims of this study was to compare the TNRT profiles from the CI24RE implant against the TNRT profiles from the predecessor CI24M/R implant. Ideally, such comparisons should be made using data from the two different implants in the same ear, which is impossible in practice. However, it is possible to use the recording data from the CI24RE implant to simulate the data as though it had been recorded with the CI24M/R implant in the following manner: Lai et al.: TNRT profiles with the Nucleus RP8 system Page 6 of 32 The main difference between NRT recordings from the two implants lies in the newer CI24RE implant's ability to measure much smaller response amplitudes, compared to the older CI24M/R implant which has a higher noise floor of typically 30uV to 40uV (Brown et al. 2000). NRT recordings with the CI24M/R implant are typically made with 100 averages (NRT3.1 software default), while the CI24RE recordings in this study involved only half this amount of averaging. To allow for the likelihood of greater noise due to this difference in averaging, the noise floor threshold was rounded up to a more conservative 50uV for the simulations in this study. Assuming that measurements with response amplitudes above 50uV would be similar with either implant, the AGF from the newer CI24RE implant can be truncated below 50uV to simulate the NRT recordings as though they had been made in the same ear with the older CI24M/R implant. Such simulations would only be approximations as the remaining (untruncated) recordings would still contain less noise than their real CI24M/R counterparts. This will have to be kept in mind when interpreting the data later. Other differences between the two implant generations' circuitry such as linearity and resolution could theoretically also affect the measurements but these are assumed to have lesser impact than the noise floor. The truncation process is illustrated in Figure 2. A typical AGF measured using the CI24RE implant is shown in Figure 2a. Data points above and below 50uV are distinguished by solid and empty diamonds respectively. The truncated AGF with only data points above 50uV is shown in Figure 2b. Truncating the AGF in this manner generally results in the extrapolated TNRT value being shifted higher, as is also illustrated in Figure 2. The exact amount of shift in the TNRT depends on the shape and the rate of growth of the AGF. At higher response amplitudes, the AGF can be approximated by a linear function, whereas at lower response amplitudes, the AGF often exhibits an asymptotic tail towards the x-axis. Differences in the AGF from one site to another would cause the TNRTs from the truncated AGFs to be shifted by different amounts for each site along the array, resulting in the TNRT profiles computed from the complete and the truncated AGFs to be not parallel to one another. Note also that the asymptotic tail at lower response amplitudes means that the slope becomes less steep as responses to higher stimulation levels are omitted from the calculation of the regression line used to extrapolate TNRT. If there were a smaller number of suprathreshold data points in the AGF, the effect of truncating the AGF would be more pronounced (see Figure 2c and 2d). Lai et al.: TNRT profiles with the Nucleus RP8 system Page 7 of 32 Subjects Complete TNRT and Map data (ACE1200) from a total of 17 subjects from 6 centers (Hannover, Iowa City, Melbourne, New York, St Louis and Zürich) were compiled for this study. The subjects were all postlingually deafened adults aged between 27 & 72 years, with less than 20 years of deafness. Their aetiologies are summarized in Table 1. Results i) Truncating the AGF: Estimating the "simulated" CI24M/R TNRT value As described in the Methods section above, AGFs for the CI24M/R implant were simulated by truncating (ignoring) all response amplitude values below 50uV. The truncated AGF yields a different TNRT value, denoted as TNRT50. This TNRT50 is typically higher than the corresponding TNRT, particularly when the AGF is not linear down past the point of truncation to the abscissa, such as when an asymptotic tail is present in the AGF at lower response amplitudes. This was generally the case, as can be seen from Figure 3, which plots TNRT versus TNRT50 pooled over all 4 sessions. The values before truncation (i.e. TNRT) and after truncation (i.e. TNRT50) showed a very good correlation (R=0.98) with one another, suggesting that the truncation has in general not produced radically different TNRT50 values. Note that truncation in this manner is only applicable to AGFs with at least 2 data points with response amplitudes above 50uV, otherwise, the AGF would no longer be defined after truncation and the TNRT50 can not be computed. As the AGFs were all measured post-operatively with awake subjects, it was not always possible to guarantee that the maximum response amplitude of an AGF would be over 50uV. Consequently, only 56% of the AGFs remained after truncation. It could be argued that using a lower truncation threshold (e.g. 30uV or 40uV) would have resulted in slightly more truncated Lai et al.: TNRT profiles with the Nucleus RP8 system Page 8 of 32 AGFs. However, the present choice of 50uV is meant to account for the smaller amount of averaging used with the CI24RE recordings. ii a) Correlations with Map T & C levels The most common form of comparison of TNRT values versus psychophysical Map T & C levels found in current literature is simple correlations (e.g. Brown et al 2000, Hughes et al 2000, Franck & Norton 2001, Cafarelli Dees et al 2005, King et al 2006). However, level differences in the TNRT, T and C values across subjects need to be taken into account before the data can be pooled together. Otherwise, the resultant pooled data would have greater scatter than appropriate. To take into account level differences across subjects, the following was done: For each subject, the mean difference between the TNRT and its corresponding T value (referred to as the T-offset) across the array was first computed. This mean T-offset was then added to the T values, effectively correcting the mean level differences between the TNRT and T level profiles across the array. This correction does not change the correlation between the TNRT and T values for a particular subject. Similarly, the mean difference between the TNRT and its corresponding C value (referred to as the Coffset) across the array was also computed. This mean C-offset was then subtracted from the C values, effectively correcting the mean level differences between the TNRT and C level profiles across the array. As with the T-offset correction, this does not change the correlation between the TNRT and C values for a particular subject. The above corrections to the T and C values are similar to those employed by Franck & Norton (2001), and were computed for both TNRT as well as TNRT50 in order to examine the effect of truncating the AGFs on the correlations with map T & C values. The offset-corrected map T & C levels yielded correlations of R=0.78 and 0.80 respectively when compared with the corresponding TNRT values (pooled data over all 4 sessions) (Figure 4). The same offset-corrected map T & C levels produced very similar results when compared with the estimated TNRT50 values, with correlations of also R=0.78 and 0.80 respectively (Figure 5). The similarity between the correlations for both TNRT and TNRT50 indicate that the presence of the Lai et al.: TNRT profiles with the Nucleus RP8 system Page 9 of 32 smaller response amplitudes below 50uV measurable with the CI24RE implant has not affected the correlations between TNRT and the corresponding Map T & C levels. ii b) Changes over time Figure 6 illustrates how the averaged TNRT, TNRT50, T and C profiles change over time. Over a period of 12 weeks from tune-up, the average TNRT and TNRT50 values were quite stable, while the average T and C levels typically showed a small but gradual increase (up to 20 Current Level units) as the subjects adapted to the CI's stimulation. A one-way analysis of variance showed that the mean differences between TNRT and the Map T & C levels did not vary significantly between sessions (T-TNRT: p=0.004, C-TNRT: p=0.006). On the other hand, the mean differences between TNRT50 and the Map T & C levels were not so consistent over the four sessions (T-TNRT50: p=0.318, C-TNRT50: p=0.469). iii) Principal component analysis The data was also subjected to principal component analysis (PCA) to examine the similarities between TNRT, T level and C level profiles across the array, in the same manner as reported by Smoorenburg et al. (2002). The analysis was then repeated for the TNRT50 profiles, compared with the same T level and C level data, in order to assess the difference between TNRT and TNRT50. Typically, only components with eigenvalues > 1 are considered to be significant. It is important to note at this point that Smoorenburg et al. (2002) used profiles consisting of 20 electrodes, while the data here only involve 5 electrodes spread out across the array per profile, raising the significance threshold from 5% to 20% of the total variance. The first 2 components from PCA of the individual TNRT, TNRT50, T and C data are summarized in Table 2, including also the results from Smoorenburg et al. (2002) for convenience. For the present data, only component 1, which accounted for 70.3%, 67.0%, 84.2% and 77.3% of the total variance for TNRT, TNRT50, T level and C level respectively, was found to be significant. As expected, given the Lai et al.: TNRT profiles with the Nucleus RP8 system Page 10 of 32 smaller number of data points per profile, these values are less than those reported by Smoorenburg et al. (2002). Interpretation of the principal components involves plotting them as a function of the electrode number, as illustrated by Figure 7. In all cases, component 1 is fairly constant across the array, suggesting it could be interpreted as an "offset" or "shift" component. In Smoorenburg et al. (2002), two main components were found for each of the TNRT, T and C data which were interpreted as a "shift" and a "tilt" component respectively. The individual correlations between TNRT (as well as TNRT50) and the corresponding T and C levels were then examined for component 1, and the results are summarized in Table 3. The correlations between TNRT versus the T & C levels were a modest 0.40 and 0.57 respectively. Similar correlations (0.58 and 0.55 respectively) were obtained for TNRT50 versus the T & C levels. Both sets of correlations compared reasonably well with those reported by Smoorenburg et al. (2002). The correlation between TNRT and the C levels for the present data was statistically significant at the 95% confidence level. PCA was also performed on the collective {TNRT, T & C} data, as well as on the collective {TNRT50, T & C} data in the same manner as Smoorenburg et al. (2002). Here, only a single component was found which accounted for 70.4% and 70.7% of the total variance, compared to 94.1% for the first component reported by Smoorenburg et al. (2002), as summarized in Table 4. The similarity between the {TNRT, T & C} and the {TNRT50, T & C} data sets can also be seen in Figure 8, which also shows that the significant component can clearly be interpreted as a "shift" component. Discussion The truncation process for simulating the NRT recordings from the CI24M/R implant yields only an approximation, as there would still be less noise in the remaining (untruncated) CI24RE recordings than in actual CI24M/R recordings. Consequently, the corresponding TNRT50 estimates, while Lai et al.: TNRT profiles with the Nucleus RP8 system Page 11 of 32 producing less scatter due to the reduced noise, would at the same time be compromised by being derived using a smaller number of remaining data points. The specific combined effect of the reduced scatter and the reduced number of data points are not investigated further here. It is important to bear in mind that TNRT50 is also approximate in nature. The NRT profiles from the CI24RE implant have been analyzed using various methods which yield different degrees of details. With simple correlations, no distinction is made between different sites along the electrode array, nor are inter-subject differences accounted for. Nevertheless, the high correlation found between TNRT and TNRT50 suggest that the truncation of the AGF below 50uV response amplitude generally resulted in little more than a level shift between TNRT and TNRT50. The observation that TNRT50 was generally larger than TNRT indicates that the truncated portion of the AGF either had an asymptotic nature as the response amplitudes approached zero, or had at least a slope lesser than that of the untruncated portion. Had the AGF been linear with the same slope all the way down to the abscissa, there would have been very little difference between the extrapolated values for both TNRT and TNRT50. Correlations between TNRT and Map T & C levels were similar to those reported in previous studies (e.g. Brown et al 2000, Hughes et al 2000, Frank & Norton 2001). The improved sensitivity of the NRT measurement circuitry has therefore apparently not altered the correlations dramatically. This again suggests that the CI24RE implant's TNRT profiles are merely lower in level, and that the changes in TNRT along the electrode array may possibly be regarded as being uniform. This view is further reinforced by the correlations between TNRT50 and the Map T & C levels, which are also similar to those found with the predecessor CI24M implant. Bearing in mind that the TNRT50 values are only approximate in nature, it would seem that truncating the AGF was a reasonable way to simulate the neural response measurements as if they had they been made with the predecessor CI24M/R implant. The TNRT being smaller than TNRT50 is a direct consequence of the improved measurement sensitivity. However, TNRT remains higher than the behavioral threshold (Map T level). A one-to-one Lai et al.: TNRT profiles with the Nucleus RP8 system Page 12 of 32 correspondence between TNRT and the behavioral threshold is unlikely to be achieved as measurements of the ECAP necessitate a certain degree of synchrony in firing of the activated neural population, whereas such synchronous activity is probably not required for the perceptual system to detect the activity at the behavioral threshold. The relationship between TNRT and the Map C levels is more difficult to determine as the latter is very much subjective in nature. Correlations between TNRT and the Map C levels are indirectly an indicator of the relationship between corresponding Map T and C levels. The results over time showed that there were not very large changes in TNRT or TNRT50 over time, suggesting that postoperative measurements of the ECAP are quite stable over time. The observed increases in both T & C levels over the same period are typical of CI users as they adapt to the CI stimulation and are gradually able to accept slightly higher levels of stimulation with time. PCA provides more information than simple correlations because it also takes into account variations along the electrode array. The overall results (see Table 4 and Figure 8) show that about 70.4% or 70.7% of the variance between either TNRT or TNRT50 and the T & C levels could be explained by a single component, interpreted as a "shift" component. Given that the TNRT50 results need to be interpreted with care because of its reduced data set after AGF truncation, the similarities between the {TNRT, T & C} and the {TNRT50, T & C} data suggest that the missing TNRT50 data has not significantly affected the principal components. Unlike Smoorenburg et al. (2002), only a single significant component which accounted for a smaller percentage of the total variance was found here. This is not surprising considering that the present data set is smaller, with only 5 instead of 20 electrodes to describe the profiles. The TNRT and TNRT50 data differed slightly from one another in that TNRT yielded significant correlations between its "shift" component and the corresponding C level, whereas the TNRT50 data, similar to the Smoorenburg et al. (2002) data, were not significantly correlated (see Table 3) with either T or C levels. One possible explanation for the improved correlations for the TNRT data could be the greater sensitivity of the CI24RE implant's measurement circuitry, since these significant correlations disappear when the lower response amplitudes were excluded (with TNRT50 as well as Lai et al.: TNRT profiles with the Nucleus RP8 system Page 13 of 32 the Smoorenburg et al. (2002) data) from the analysis. Note however that the TNRT50 data should not be equated with the Smoorenburg et al. (2002) data due to the smaller number of electrodes per profile as well as the reduction in TNRT50 data after AGF truncation. This is possibly a reason also for the lack of significant correlations for TNRT50. Another difference between the Smoorenburg et al. (2002) data and the present study is that the former exclusively involved the CI24M implant which had a straight array of full-ring banded electrodes. In the present study, the CI24RE implant is equipped with the perimodiolar Contour/Contour Advance electrode array, which is expected to provide more variation along the length of the electrode array compared to the straight electrode array. For instance, Smoorenburg (2007) more recently reported a "curvature" component in the Map data where the majority (127 versus 88) of the implants had a perimodiolar electrode array. On the other hand, Willeboer & Smoorenburg (2006) involved Contour/Contour Advance electrode arrays but reported only significant "shift" and "tilt" components, possibly due to the smaller population size of only 18 subjects. Unfortunately, the present data with only 5 data points for representing each profile is insufficient for revealing features more subtle than the "shift" component, particularly features that are related to longitudinal variations such as "tilt" or "curvature". In fact, from Smoorenburg et al. (2002) it could already be seen that the "tilt" component accounted for a substantially smaller percentage of the total variations than the "shift" component. On the whole, PCA results were similar to those found with the CI24M implant, again reinforcing the view that the TNRT profiles with the CI24RE implant have not been significantly altered to provoke any changes in the clinical application of these TNRT profile. Conclusions The aim of this study was to determine if the TNRT profiles have changed with the new CI24RE implant compared to its predecessor CI24M/R implant. Truncation of the AGF was employed to produce TNRT50 profiles in order to simulate the TNRT profiles of the earlier implant. Both TNRT and Lai et al.: TNRT profiles with the Nucleus RP8 system Page 14 of 32 TNRT50 data produced correlations similar to those reported for the CI24M implant, suggesting that, apart from a level change, there are no great differences in the TNRT profiles between the two implant generations. Due to the reduction in the number of data points representing each profile, principal component analysis was able to only reveal a single principal component, with similar significance to that found for the predecessor implant. Significant correlations found with the Map C levels are likely to be related to the CI24RE implant's improved measurement amplifier, but this has little impact on the general observations above. Thus, it is not expected that the use of TNRT profiles in assisting with clinical speech processor programming will be altered greatly for the CI24RE implant. Acknowledgments This study was supported by a research grant from Cochlear AG, Basel as well as a research grant No. 320000-110043 from the Swiss National Science Foundation The authors would also like to thank Bas van Dijk, Kerrie Plant & Sarah Ashburn Reed for part of the data collection, as well as Professor Guido Smoorenburg for comments on the principal component analysis. Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper Lai et al.: TNRT profiles with the Nucleus RP8 system Page 15 of 32 References Battmer RD, Dillier N, Lai WK, Weber BP, Brown C, Gantz BJ, Roland JT, Cohen NJ, Shapiro W, Pesch J, Kilian MJ & Lenarz T. 2004. Evaluation of the neural response telemetry (NRT) capabilities of the Nucleus Research Platform 8: initial results from the NRT trial. Int J Audiol, 43(Suppl 1), S10-15. Brown CJ, Hughes ML, Luk B, Abbas PJ, Wolaver A & Gervais J. 2000. The relationship between EAP and EABR thresholds and levels used to program the Nucleus 24 speech processor: Data from adults. Ear Hear, 21(2), 151-163. Cafarelli Dees D, Dillier N, Lai WK, von Wallenberg E, van Dijk B, Akdas F, Aksit M, Batman C, Beynon A, Burdo S, Chanal J-M, Cradock L, Cullington H, Deggouj N, Fraysse B, Grabel S, Kiefer J, Kiss JG, Lenarz T, Mair A, Maune S, Müller-Deile J, Piron J-P, Razza S, Tasche C, Thai-Van H, Toth F, Truy E, Uziel A & Smoorenburg GF. 2005. Normative findings of electrically evoked compound action potential measurements using the neural response telemetry of the Nucleus CI24M cochlear implant system. Audiol Neurootol, 10(2), 105-116. Franck KH & Norton SJ. 2001. Estimation of psychophysical levels using the electrically evoked compound action potential with the Neurel Response telemetry scapabilities of Cochlear Corporation's CI24M device. Ear Hear, 22(4), 289-299. Gordon KA, Ebinger KA, Gilden JE & Shapiro WH. 2002. Neural response telemetry in 12- to 24month.old children. Ann Otol Rhinol Laryngol, 189(Suppl), 42-48. Hughes ML, Brown CJ, Abbas PJ, Wolaver AA & Gervais JP. 2000. Comparison of EAP thresholds with map levels in the Nucleus 24 cochlear implant: Data from children. Ear Hear, 21(2), 164-174. Kaplan-Neeman R, Henkin Y, Yakir Z, Bloch F, Berlin M, Kronenberg J, Migirov L, Hildesheimer M & Muchnik C. 2004. NRT-based versus behavioral-based MAP: a comparison of parameters and speech perception in young children. J Basic Clin Physiol Pharmacol, 15(1-2), 57-69. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 16 of 32 King JE, Polak M, Hodges AV, Payne S & Telischi FF. 2006. Use of neural response telemetry measures to objectively set the levels in the Nucleus 24 cochlear implant. J Am Acad Audiol, 17(6), pp413-431. Lai WK. 2004. An NRT Cookbook: Guidelines for making NRT measurements, 2nd Edition for Version 3.1. Cochlear, Basel. Mason SM, Cope Y, Garnham J, O'Donoghue GM & Gibbin KP. 2001. Intra-operative recordings of electrically evoked auditory nerve action potentials in young children by use of neural response telemetry with the Nucleus CI24M cochlear implant. Br J Audiol, 35(4), 225-235. Patrick JF, Busby PA & Gibson PJ. 2006. The development of the Nucleus Freedom Cochlear implant system. Trends Amplif, 10(4), 175-200. Pedley K, Psarros C, Gardner-Berry K, Parker A, Purdy SC, Dawson P & Plant K. 2007. Evaluation of NRT and behavioral measures for MAPping elederly cochlear implant users. Int J Audiol, 46(5), 254262. Potts LG, Skinner MW, Gotter BD, Strube MJ & Brenner CA. 2007. Relationship between Neural Response Telemetry Thresholds, T- and C-Levels and loudness judgements in 12 adult Nucleus 24 cochlear implant recipients. Ear Hear, 28(4), 495-511. Smoorenburg GF, Willeboer C & van Dijk JE. 2002. Speech perception in Nucleus CI24M cochlear implant users with processor settings based on eletrically evoked compound action potential thresholds. Audiol Neurootol, 7(6), 335-247. Smoorenburg GF. 2007. T- and C-level profiles across the electrode array: Fitting the speech processor by profile parameter adjustment. Chapter 2, Cochlear Implant Ear Marks. University Medical Centre Utrecht, The Netherlands. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 17 of 32 Willeboer C & Smoorenburg GF. 2006. Comparing cochlear implant users' speech performance with processor fittings based on conventionally determined T and C levels or on compound action potential thresholds and live-voice speech in a prospective balanced crossover study. Ear Hear, 27(6), 789798. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 18 of 32 List of Tables Table 1: Subjects summary Table 2: Principal components analysis results – Component 2 for the present data (black cells) are not significant. The present data's significant components account for a smaller proportion of the total variances compared to those from Smoorenburg et al. (2002). Table 3: Correlations between TNRT and the corresponding map T & C levels for the CI24RE implant (TNRT), the simulated CI24M/R implant (TNRT50) and the actual CI24M/R implant (Smoorenburg et al. (2002)). Table 4: Combining TNRT and Map levels, only a single significant principal component waas found for all 3 instances. The overall data with either TNRT or TNRT50 from the present data showed very similar results. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 19 of 32 Table 1 Age (years) Duration of Deafness (years) 27 2 Autoimmune 42 7 Unknown 43 5 Unknown 43 2 Measles 43 20 Familial 43 12 Unknown 53 0 Meniere's Disease 61 12 Unknown 63 4 Unknown 64 12 Unknown 67 0 Familial 67 2 Unknown 71 6 Trauma 72 8 Ototoxic Drugs 72 1 Familial 74 5 Unknown 74 1 Unknown Aetiology Lai et al.: TNRT profiles with the Nucleus RP8 system Page 20 of 32 Table 2 Percentage variance accounted for Present Data Smoorenburg et al. (2002) TNRT TNRT50 T C TNRT T C Component 1 70.3 67.0 84.2 77.3 90 90.0 88.3 Component 2 15.0 18.5 10.6 17.0 6.4 6.7 7.5 Sum (significant) 70.3 67.0 84.2 77.3 96.4 96.7 95.8 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 21 of 32 Table 3 TNRT Component Interpretation Shift Tilt * 0.05 TNRT50 TNRT (Smoorenburg et al. 2002) T Level C Level T Level C Level T Level C Level 0.40 0.57* 0.58 0.55 0.64 0.39 0.82 0.36 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 22 of 32 Table 4 Percentage variance accounted for Smoorenburg et al. (2002) Present Data Component 1 TNRT, T & C TNRT50, T & C TNRT, T & C 70.4 70.7 94.1 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 23 of 32 List of Figures: Figure 1: NRT recordings from typical CI users with the CI24RE (left column) and the CI24M/R (right column) implants for subthreshold as well as suprathreshold input signals. The examples here were selected to have comparable suprathreshold response amplitudes. The recordings were all made with the same 60dB amplifier gain setting, and involved 50 averages. The second row shows how a response of around 25uV peak-to-peak amplitude is clearly discernable with the CI24RE implant but not at all with the CI24M/R implant. Figure 2: The original AGF (a) with all data points as measured with the CI24RE implant yields an extrapolated TNRT of 138 CL. The threshold TNRT50 is shifted to 145 CL when the AGF is truncated below 50uV (b). The truncated AGF simulates the AGF as if it had been measured with the CI24M/R implant. If the highest stimulation level in the untruncated AGF were lower (resulting in less suprathreshold data points), truncating the AGF produces a larger difference between TNRT and TNRT50 (c and d). Figure 3: Correlations between original TNRT (untruncated AGFs) data and TNRT50 (AGFs truncated below 50uV). Note that TNRT50 was generally larger than the corresponding TNRT. Figure 4: Correlations between original TNRT (untruncated AGFs) data with offset corrected (left) T levels and (right) C levels. Figure 5: Correlations between TNRT50 (truncated AGFs) data with offset corrected (left) T levels and (right) C levels. Figure 6: Average TNRT and TNRT50 profiles appear to be stable over the 12-week period of the study, while average T- and C-Levels show a small but gradual increase over the same period. Lai et al.: TNRT profiles with the Nucleus RP8 system Page 24 of 32 Figure 7: Plots of the principal components across the array show how component1 can be interpreted as an "offset/shift" component for the individual TNRT, TNRT50, T Level or C Level data. Figure 8: Plots of the single principal component of the collective data sets show very similar component values, suggesting that the TNRT and TNRT50 data are very similar. Lai et al.: TNRT profiles with the Nucleus RP8 system Figure 1 Page 25 of 32 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 26 of 32 Figure 2 Only data points ≥ 50uV All data points 300 300 Data ≥ 50uV 250 Data < 50uV Response Amplitude uV Response Amplitude uV 250 200 150 100 TNRT=138 50 200 150 100 TNRT50=145 50 0 0 130 140 150 160 170 180 190 200 210 130 140 150 Current Level CL 160 170 180 190 200 210 Current Level CL a) b) Only data points ≥ 50uV (with less suprathreshold data) All data points (with less suprathreshold data) 300 300 Data ≥ 50uV 250 Data < 50uV Response Amplitude uV Response Amplitude uV 250 200 150 100 TNRT=132 50 200 150 100 TNRT50=142 50 0 0 130 140 150 160 170 180 190 200 210 130 c) 140 150 160 170 180 Current Level CL Current Level CL d) 190 200 210 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 27 of 32 Figure 3 220 200 R = 0.98 TNRT50 180 160 140 120 100 100 120 140 160 TNRT 180 200 220 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 28 of 32 220 220 200 200 180 C (less mean offset) T (plus mean offset) Figure 4 R = 0.78 160 140 120 100 100 180 160 R = 0.80 140 120 120 140 160 TNRT 180 200 220 100 100 120 140 160 TNRT 180 200 220 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 29 of 32 220 220 200 200 C (less mean offset) T (plus mean offset) Figure 5 180 R = 0.78 160 140 120 100 100 180 R = 0.80 160 140 120 120 140 160 TNRT50 180 200 220 100 100 120 140 160 TNRT50 180 200 220 Lai et al.: TNRT profiles with the Nucleus RP8 system Page 30 of 32 Figure 6 Changes over time: Average C levels 180 170 160 170 160 150 140 130 120 110 100 90 80 TNRT w 0 TNRT w 4 Level (CL units) Level (CL units) Changes over time: Average TNRT 180 TNRT w 8 TNRT w 12 150 140 130 120 Cw4 90 80 C w 12 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode Changes over time: Average T levels Changes over time: Average TNRT50 180 160 150 TNRT50 w 0 110 100 TNRT50 w 4 90 80 TNRT50 w 12 TNRT50 w 8 Level (CL units) Level (CL units) Cw8 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode 180 170 140 130 120 Cw0 110 100 170 160 Tw0 150 140 Tw8 130 Tw4 T w 12 120 110 100 90 80 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode Lai et al.: TNRT profiles with the Nucleus RP8 system Page 31 of 32 Figure 7 1 Component value Component value 1 0 TNRT data TNRT50 data Comp 1 2 3 4 5 6 7 8 1 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode 1 1 Component value Component value Comp 1 -1 -1 1 0 0 T Level data 0 C Level data Comp 1 -1 Comp 1 -1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode Lai et al.: TNRT profiles with the Nucleus RP8 system Page 32 of 32 Figure 8 1 Component value Component value 1 0 Combined TNRT, T & C data 0 Comp 1 -1 Combined TNRT50, T & C data Comp 1 -1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Electrode