J Am Acad Audiol 26:205-212 (2015)
Evaluation of iPod-Based Automated Tinnitus Pitch Matching DOI: 10.3766/jaaa.26.2.9 Robert Wunderlich* Alwina Stein* Alva Engell* Pia Lau* Lea Waasem* Alex Shaykevichf Claudia Rudacld: Christo Pantev*
Abstract Background: Tinnitus is the perception of sound unrelated to any external source. Diagnostic approaches to assess tinnitus characteristics such as tinnitus pitch are crucial for new attempts of tinnitus therapy. Purpose: The purpose of this study was to develop and evaluate reliable tinnitus pitch-matching pro cedures. Existing procedures usually require audiometric equipment and are time consuming. However, some patients with tinnitus find it hard to match their tinnitus in one single session. Therefore, we devel oped an iPod-based application for seif-administered tinnitus pitch matching and compared it with a standardized audiometric procedure. Study Sample: A total of 17 patients with chronic tonal tinnitus participated in two sessions including both pitch-matching procedures. Method: In the conventional audiometric test, the investigator adjusted the frequency and loudness of pure tones led by the responses of the patient. For the iPod-based procedure, we used a recursive twointerval forced-choice test that required no interaction with an investigator. Both procedures included loudness matching and testing for octave confusion. Results: The iPod-based procedure resulted in lower pitch matches as compared with the conventional audiometry. Psychometric qualities such as test-retest reliability of both methods were comparable. Par ticipants rated the iPod-based procedure as easier to perform and more comfortable to use. Conclusions: In conclusion, we find that the use of self-administered tinnitus pitch-matching procedures on a mobile device is feasible and easier in practice without any loss of reliability and validity. A major advantage is the possibility of repeated measurements without expensive equipment and experienced staff. Repeated measurements of tinnitus pitch can provide more information about the stability of the tinnitus perception and may improve the ability of participants to match their tinnitus. Key Words: Tinnitus, tinnitus pitch, pitch matching, reliability, iPod, smartphone Abbreviations: CAT = conventional audiometric test; ENT = ear, nose, and throat; RIFT = recursive two-interval forced-choice test; SD = standard deviation; SPL = sound pressure level; TMNMT = tailormade notched music training
‘ Institute for Biomagnetism and Biosignalanalysis, University Hospital of Munster, Munster, Germany; fUniversity of Western Australia, Crawley Perth, Western Australia, Australia; ^Department of Otolaryngology, University Hospital of Munster, Munster, Germany Univ.-Prof. Dr. Christo Pantev, Institute for Biomagnetism and Biosignal Analysis, University Hospital of Munster, Malmedyweg 15 D-48149 Munster, Germany; Tel: 0049 251 83 56885; Fax: 0049 251 83 56874; E-mail: [email protected] This research was supported by the Interdisziplinares Zentrum fur klinische Forschung Munster, Medical Faculty, University of Munster, Germany (Project No. CRA05). This work was presented at the TRI 2013 - 7th International TRI Conference on Tinnitus, May 15-18, 2013, Valencia, Spain.
205
Journal o f the American Academy o f Audiology/Volume 26, Number 2, 2015
INTRODUCTION innitus is a phenomenon that draws a large amount of attention in current neuroscientific research. It refers to the perception of a sound that is unrelated to any external source (Eggermont and Roberts, 2004). Although tinnitus may disappear in some cases, for 10-15% of the general population, tin nitus is an unremitting condition (Heller, 2003). An estimated 1-3% of those affected with tinnitus experi ence a significant loss in quality of life, involving sleep disturbance and even depression (Dobie, 2003). Although many questions cannot be answered yet, our understanding of the underlying neurophysiologi cal mechanisms of tinnitus is improving. Initially, tin nitus may be triggered by damage to the inner ear hair cells, but the neural generators of noise-induced tinnitus appear to be located in the central regions of the brain (Eggermont and Roberts, 2004; Norena, 2011). From the individual perspective, tinnitus is described as a whistling or beeping sound, but because there is no physical sound, it is not possible to measure its physical qualities. However, it has perceptual qualities such as pitch, loudness, and bandwidth or pulsing described by patients, which are comparable to real physical sounds. The measurement of these qualities relies on subjective information given by the individual with tinnitus. As knowledge about the neurophysiology of tinnitus grows, recent research has also brought some new per spectives on tinnitus treatment. Many of these treat ments try to reverse changes of maladaptive plasticity. The tailor-made notched music training (TMNMT), which has been applied in tonal tinnitus, aims at sup pressing hyperactive neurons by enhancing lateral inhibition from frequencies lower and higher than the tinnitus (Okamoto et al, 2010; Teismann et al, 2011; Pantev et al, 2012). Pathological neuronal syn chrony is the target of acoustic coordinated reset neuro modulation (Tass et al, 2012; Tass and Popovych, 2012). However, both types of training have not been evaluated in randomized clinical trials with large samples. Yet both sound therapies mentioned above share a common approach that addresses specific regions of the cochlea and the auditory cortex. Often these regions are related to the perceptual qualities of the tinnitus. Theoretically, the TMNMT can only work if the tinnitus frequency is covered by the notch filter in the music. The coordinated reset neuromodulation also focuses on frequencies in a specific distance to the tinnitus frequency. Conse quently, in order to be effective, such training requires reliable estimates of the tinnitus pitch. Different approaches have been made to assess the tinnitus pitch. However, up to now no procedure has been established as a clinical standard, although stan dardization is requested (Vernon and Fenwick, 1984).
T
206
The reason is the poor reliability within and between sessions of most procedures even in a controlled envi ronment. Standardized methods such as the forcedchoice double-staircase procedure have been shown to cause less within-session variability than methods of manual adjustment (Jesteadt, 1980; Penner and Bilger, 1992; Penner and Klafter, 1992; Henry et al, 2013). With a wider use of computer-automated procedures, it has also become easier to test different adaptive procedures. Among those, the two-alternative forcedchoice method provided the best reliability (Henry et al, 2001, 2004). The recursive two-interval forcedchoice test (RIFT) procedure uses a complex algorithm to bracket the pitch match to smaller intervals until only one frequency remains (Diesch et al, 2004). How ever, its psychometric qualities have not yet been com pared with those of other methods. It has also been proposed that a frequency spectrum measured with a likeness rating could be used to describe the tinnitus percept (Norena et al, 2002). The resulting curve often mirrors the hearing loss and is stable across time (Moffat et al, 2009), although it does not always provide a clear peak that can be interpreted as the prominent tinnitus pitch. Furthermore, patients are biased to rate higher frequencies generally as more similar to their tinnitus (Norena et al, 2002). To summarize, currently no proce dure is available that provides tinnitus pitch matching with high within- and between-session reliability. Besides the different procedures that are used for pitch matching, there is also no common standard to decide which ear should be tested. Although some authors rec ommend using the contralateral ear (Henry et al, 2004), others use the ipsilateral ear (Norena et al, 2002), claim that it does not make a difference, or recommend testing both ears separately (Vernon and Fenwick, 1984). Un fortunately, there is no study comparing pitch-matching reliability systematically for these options. The valida tion of tinnitus pitch-matching procedures is also gen erally difficult. Up to now, there is no objective way to measure tinnitus frequency; therefore, it is hard to define a validation criterion with which to compare pitch-matching results. Traditionally, tinnitus pitch matching is performed with audiometric equipment as part of an audiological evaluation, including measuring hearing thresholds, loudness matching, and minimal masking level. How ever, the procedure is time consuming and repeating measurements is often not practicable in clinical rou tine. Therefore, using different platforms for tinnitus pitch matching that are independent of clinical equip ment could be beneficial. Certainly, these platforms have to meet clinical standards for this kind of measure ment. Recently, smartphones have been increasingly used as a support for professionals and patients in the health care context (Mosa et al, 2012). All kinds of specific applications (apps) can be installed. These
Evaluation of iPod-Based Automated Tinnitus P itch Matching/Wunderlich et al
apps can help with disease diagnosis, management of chronic conditions, clinical communication and decision making, and standardized diagnostics (Boulos et al, 2011; Field et al, 2014). Apps for automated audiometry have already been tested. It was shown th at 96% of the thresholds measured with an iPod-based app were within 10 dB of the threshold measured with conven tional audiometry, depending on whether or not mea surem ents were performed in a sound booth (Foulad et al, 2013). The downloadable UHear app tends to over estimate hearing loss, but the authors claim th at it can be used to screen for hearing loss (Szudek et al, 2012). On a subjective level, most participants preferred iPodbased audiometry because they found it simpler, more comfortable, less intimidating, more convenient, and more controllable (Foulad et al, 2013). For pitch matching, the calibration of loudness is not as important as for measuring audiometric thresholds because loudness is not the main outcome variable. However, it might be possible th at the use of a mobile device has certain advantages beyond the user’s subjec tive preference. Because all methods of pitch matching cause variability, more repetitions can lead to more pre cise estimates of the “true” tinnitus frequency and also give information about the reliability of a single partic ipant (Tyler and Conrad-Armes, 1983). With a proce dure based on a mobile device, it would become much easier to obtain repeated measurements because indi viduals can perform testing in their own environment without much special equipment. Although there is a great potential of this new platform for improving tin nitus diagnostics and treatm ent, detailed evaluation is still needed. The goal of this study was to develop and evaluate an application for automated iPod-based tinnitus pitch matching and to compare its reliability and validity with a conventional audiometric test (CAT) that is based on a forced-choice multiple staircase procedure. The procedure chosen for the iPod-based pitch matching was the RIFT. We hypothesized higher test-retest reliability with RIFT compared with the CAT. Furthermore, we expected that RIFT would be less time consuming and would allow more repetitions during the same time.
MATERIALS AND METHODS Participants A total of 17 participants with tinnitus took part in two sessions of pitch-matching tasks. Participants were recruited via advertisements in local newspapers, our homepage, and distribution of flyers to local ear, nose, and throat (ENT) practitioners. Potential participants were asked to fill out an anamnestic questionnaire th a t we developed for this purpose. The questionnaire covered questions about the characteristics, duration,
onset, possible cause, and treatm en t of the tin n itu s as well as preexisting illnesses, hyperacusis, current medication, and alcohol and drug consumption. The participants were then scheduled for the study. Partici pants had to match the following criteria: The tinnitus had to be present for at least 3 mo. We selected only par ticipants who reported to have a tonal tinnitus (whistling or beeping sound) th at was stable in pitch over time. Hearing loss had to be less than 70 dB HL for all fre quencies between 125 Hz and 8 kHz. No participants with ENT, neurological, or psychiatric disorders were included. The sample included 5 females and 12 males (mean age = 49.44 yr, age range, 26-63 yr, standard deviation [SD] = 11.45 yr). Three participants had musical training in the past, without being professional musicians. The study was conducted according to the Declaration of Helsinki, and the study protocol was approved by the ethics committee of the Medical Faculty of the University of M unster. Each participant signed an approved informed-consent form before study enrollm ent.
Procedures The first session started with an interview to explore the occurrence, development, and history of tinnitus treatm ent as well as the criteria already specified in the questionnaire described above. Each participant received an initial ENT evaluation th at included an ear examination performed by an otorhinolaryngologist from the ENT Department. The purpose of this exami nation was to remove earwax from the participants’ ear th a t otherwise might have influenced hearing thresh olds. We also wanted to check for any illnesses of the ear canal or the tympanic membrane. Audiometric test ing was conducted in a double-walled sound-attenuated chamber (ISO 8253-1). Hearing thresholds were mea sured manually by an audiologist for the frequency range of 125 Hz to 16 kHz with an appropriate audiometer (MADSEN Astera, Denmark) and calibrated headphones (Sennheiser HDA 200). Two runs (each method once) of tinnitus pitch matching were obtained with a half-hour break for the participant to relax outside of the sound booth. A second session took place 1 wk after the first ses sion. At the end of the second session, participants were additionally tested with a tinnitus pitch likeness rating, which was not included in the analysis because of techni cal problems. The order in which the two pitch-matching m ethods were used in each session was random ly assigned. All pitch-matching procedures were performed with the ear ipsilateral to the tinnitus perception. With bilateral tinnitus, participants were asked in which ear they perceived their tinnitus as more dominant. If they were not able to name one ear as more dominant, the ear with less average hearing loss was chosen for testing. All testing sounds were pure tones. Each session lasted approximately 2.5 hr.
50 7
Journal of the American Academy of Audiology/Volume 26, Number 2, 2015
Participants were asked to rate the two procedures on four different visual analog scales (0-100) concerning comprehensibility, ease of use, certainty about the result, and comfort level of the procedure.
CAT The CAT was conducted with the same audiometric equipment used for measuring hearing thresholds, and the participants were informed about the different steps of the procedure. We used seven frequencies as starting tones in the following order: 4, 12.5, 1, 10, 2, 8, and 6 kHz. The participants were presented with the starting tone and were asked if the tinnitus pitch was higher or lower. After the participants provided their answers, the frequency of the tone was adjusted. Adjusting steps were initially half an octave until the participant changed the direction of the answer or rated the presented tone as equal to the tinnitus. The proce dure continued in one-twelfth octave steps until the participant indicated th at the presented tone was again similar to the tinnitus. This procedure was repeated for the seven different starting tones. The highest possible frequency presented to participants was 16 kHz. Ini tially, all tones were presented 10 dB above the individ ual hearing threshold. The maximal presentation level was 90 dB SPL. For the adjusted frequency, we matched loudness to the individual tinnitus loudness in 1 dB steps by asking the participant whether the tinnitus was softer or louder. The resulting seven frequencies were then con tinually entered into a two-alternative forced-choice test where the participant was to choose which frequency was more similar to the tinnitus, starting with the lowest two frequencies until only one frequency remained. To avoid octave confusion, the resulting frequency was afterward tested in a two-alternative forced-choice fashion against its octaves. Beforehand, the octaves were also matched to the loudness of the tinnitus (Vernon and Fenwick, 1984).
RIFT Testing was performed with an iPod Touch fourthgeneration Model A1367 8GB, running a custom appli cation with all necessary features specifically designed for tinnitus pitch matching. The headphone used was a Sennheiser HD 201. As a safety constraint, the ampli tude of the sounds was fixed. All sound stimuli had a maximal sound pressure level (SPL) between 75 and 90 dB SPL, measured with a 2 cm3 ear simulator (Brad & Kjaer, model 4157). Participants were first instructed how to use the application with a short manual of two pages. Additionally, pop-up windows guided the par ticipant through each step. The testing itself did not require interaction with an examiner. In the first step, participants had to match the loudness of every quarter octave from 1—16 kHz to the perceived
SO B
tinnitus loudness with a slider. Loudness values were stored as default loudness to be used later in the testing. The participant still had the opportunity to change the loudness of the tones during the actual testing. The loudness values for the frequencies between the quarter octaves were interpolated. The frequency range was lim ited to the highest frequency th at the participant was able to hear. The actual procedure was a two-alternative forced-choice test in which participants were instructed to select the frequency th at was more similar to their tinnitus (see Figure 1). For the comparisons, the fre quency range from 1-16 kHz was bisected into two equally large subintervals. Depending on the decisions, bisection and two-interval forced-choice testing was reapplied to the low subinterval, the high subinterval, or a new middle interval that was bound by the mid points of the low and the high subinterval. For example, if the participant was presented with 1 versus 4 kHz (2 octaves) as the first interval and 4 versus 16 kHz (2 octaves) as the second interval and chose 1 kHz as the first decision and 4 kHz as the second decision, the next intervals would have been 1 versus 2 kHz (1 octave) and 2 versus 4 kHz (1 octave). This procedure was repeated until one final pitch was selected. The smallest step was a twelfth octave. The testing was com pleted with an octave confusion test. If the participant’s choices were contradictory twice in a row, the test was aborted and the procedure was restarted. The RIFT was repeated three times or until the participant completed one pitch matching. For the analysis, we used the first valid pitch match and, additionally, the mean across all valid runs.
Statistical Analysis In preparation of the analyses, all frequencies were transformed into a logarithmic scale to calculate means and SDs and retransformed for better interpretation. Differences of the mean pitch matches of both methods were analyzed with a repeated-measures analysis of
iPod
10:08 M essung
Back
m
\
i
Welcher der beiden Tone ist Ihrem Tinnitus in der Tonhohe ahnlicher?
w J it,
*
fit
w . *
V Ton A
OK
Ton B
Figure 1. Graphical user interface of the RIFT. English transla tions: “Measurement. Which of the two tones is closer to your tin nitus in pitch? Tone A, Tone B.”
Evaluation of iPod-Based Automated Tinnitus Pitch Matching/Wunderlich et al
variance with two within-participant factors session (Session 1 versus Session 2) and method (CAT versus RIFT first pitch match versus RIFT mean of all runs). Test-retest reliability was calculated with the Pearson product-moment correlation coefficient between Session 1 and Session 2 for each method. Reliability of the methods was further analyzed with Cronbach’s a. Furthermore, we calculated f-tests on the means of the time duration and subjective evaluation.
RESULTS earing thresholds of the sample are displayed in Figure 2. There was no significant difference in the hearing thresholds of the right and left ears across all participants. Nine participants reported to have uni lateral left-sided or left-dom inant tinnitus, and six reported unilateral right-sided or right-dom inant tin nitus. Two participants reported th eir tin n itu s as bilateral. The mean pitch matches measured with CAT and RIFT are shown in Table 1. The analysis of variance revealed a significant main effect for the method. As the assumption of sphericity was violated, revealed via the Mauchly test (p < 0.000), the degrees of freedom were Greenhouse-Geisser corrected for the withinfactor method resulting in F a 156;18 496) = 6.132, p < 0.05. The session did not have a significant effect on the pitch matches [F(1>16) = 0.995, p > 0.5]. Planned simple contrasts revealed th at CAT and the first pitch match of RIFT provided significantly different pitch matches [F(1j16) = 7.375, p < 0.05], and CAT was differ ent from the average over three runs of RIFT [F(116) = 5.656, p < 0.05]. Averaged absolute differences were cal culated to illustrate the variability of the different methods (Table 1).
H
All test-retest correlations were significant, as shown in Table 1. The difference in the size of the Pearson r’s was not significant between the different methods. To calculate the internal consistency for the CAT, we ana lyzed the first seven matched frequencies. Cronbach’s a was 0.912 for the CAT for Session 1 and 0.859 for Ses sion 2. Because a single run of the RIFT did not provide any intermediate pitch matches, Cronbach’s a was only calculated for the three runs. The RIFT reached a Cron bach’s a of 0.864 during the first session and 0.796 dur ing the second session. It has to be noted th at for the RIFT, five participants provided only one or two valid pitch matches during Session 1. In Session 2, one par ticipant provided only one pitch match. These partic ipants had to be excluded from the calculation of Cronbach’s a, resulting in an overall sample size of 12 for the first session and 16 for the second session. The mean range of pitch matches provided by a par ticipant across both methods (CAT and RIFT) and both sessions was almost 1 octave (mean = 1935 Hz, SD = 1500 Hz). However, there was no difference between female and male participants [t(15) = 0.851, p > 0.05]. In the first session, the participants needed 20.06 min (SD = 3.09 min) to complete the pitch matching with the CAT. In the second session, the matching took 17.20 min (SD = 2.70 min). For the RIFT, participants needed 20.69 min (SD = 4.64 min) in the first session and 20.13 min (SD = 5.78 min) in the second session. Differences between the two methods were only significant for the second session (t = 2.067, p < 0.05). On a subjective level, participants rated both m eth ods as good to comprehend (Table 2). The certainty th at the procedure yielded the correct pitch match did not differ between the two methods. Participants also rated both methods as highly comprehensible but favored the RIFT concerning simplicity (t = 3.05, p < 0.01) and com fort (t = 3.18, p < 0.01).
DISCUSSION wo procedures for pitch matching were investi gated in this study. Both methods had comparable reliability. According to the criteria of Gliner et al (2009), a test-retest reliability above 0.8 can be consid ered as high, which applies only for the RIFT with three repetitions. The CAT demanded a final decision in each session by means of two-alternative forced-choice test ing. However, like any other previous decision, this decision was also confounded by a measurement error that could not be minimized in this procedure. Averag ing over all runs of the RIFT provided a better chance of canceling out this error. The internal consistency for the CAT was good, whereas the consistency for the RIFT was only acceptable. It should be taken into account th at the low number of items th at were used for the calculation of Cronbach’s a, especially in the case
T
Figure 2. M e a n h e a r i n g t h r e s h o l d s (in dB S PL ) of all 17 p articip an ts.
209
J o u rn a l o f th e A m erica n A cad em y o f A udiology/Volum e 26, N um ber 2, 2015
Table 1. Summary of Pitch-Matching Results Mean Pitch Match Pitch-Matching Method
Session 1 (SD)
Session 2 (SD)
Average Absolute Difference
Pearson r
CAT RIFT, First Pitch Match RIFT, Mean of All Runs
7,692 (1,702) 5,816 (1,504) 6,126 (1,484)
8,287 (1,632) 6,262 (1,550) 6,140 (1,479)
468.31 530.89 339.82
0.649* 0.561f 0.825*
Note: Means and SDs are expressed in Hz. *p < 0.05, one-tailed. f p < 0.01, one-tailed.
of the RIFT, deflates the value of Cronbach’s a (Cortina, 1993). Surprisingly, the RIFT resulted in lower pitch matches. A reason for this could be th at the first two decisions of this tw o-alternative forced-choice test always presented very distant frequencies. Errors in these decisions can usually result in a restriction of the frequency spectrum th at does not allow returning to lower or higher frequencies. In particular, the upper boundary of the first high interval might have caused a certain bias because it was always near the edge of the individual’s hearing. On the other hand, there might have been a bias toward high frequencies in the CAT. In the second part of this procedure, when frequencies were entered into a two-alternative forced-choice test, higher frequencies were presented later than lower fre quencies. Therefore, lower frequencies had to pass more decisions than higher frequencies. Additional variance might also have been introduced because of the different headphones used for both pro cedures. For the RIFT, we wanted to increase ecological validity by using a headphone that was not solely for audio metry. We chose to use the Sennheiser HD201 because of its flat frequency spectrum, which does not attenuate specific frequencies. Furtherm ore, we w anted to use a circum aural headphone to shield from surrounding noise. The Sennheiser HD201 is also not too expensive (approximately 21€ in Germany). Nevertheless, the dif ferent equipm ent used for both procedures was a lim itation of this study. Because two different methods and two different setups of equipment were used, it is possible th a t the equipm ent also had an effect on the results. However, small differences in the loudness output should not have influences on the test-retest reliability. Another source of variance responsible for reduced reliability, besides the method itself, was the partici pant’s general ability to perform pitch matches. Pitch matching is not a simple task for most participants.
The largest differences can be observed when musicians and nonmusicians are compared (Moreti et al, 2012). Playing an instrum ent trains the auditory system, including the auditory cortex, in a very special way that leads to an increased cortical representation and also morphological changes (Pantev et al, 1998; Schneider et al, 2002). Musicians have a lower threshold for dis criminating frequencies than nonmusicians; nonethe less, nonmusicians can be trained to perform better (Micheyl et al, 2006). Unfortunately, our sample did not give us the opportunity to analyze how musical training affected our data. Untrained participants also often confuse frequency differences with loudness dif ferences (Vernon and Meikle, 1988). Special training before pitch-matching tasks might prevent participants from mixing up both dimensions or at least ensure that participants who are not able to do so can be identified and excluded (Henry et al, 2001). Another factor that reduces the ability to match frequencies is hearing loss th at is often present in patients with tinnitus. In the region of hearing loss, the frequency discrimination is considerably reduced (Moore et al, 2001). Significant hearing loss can even result in dead regions, which leads to distorted processing of pure tones (Huss and Moore, 2005). This is even more problematic because a connection between the region of hearing loss and the frequency of tinnitus is assumed (Henry et al, 1999; Norena et al, 2002; Eggerm ont and Roberts, 2004; Konig et al, 2006; Roberts et al, 2006; Moffat et al, 2009; Schaette and Kempter, 2009). Considering these issues, pitch matching is a challenging task for the tested participants. This is reflected in the certainty ratings, which were only moderate. To minimize varia bility introduced by the tinnitus itself, it is better to exclude p a rticip a n ts w ith unstable or even atonal tinnitus and too-pronounced hearing loss. It is also a possibility to include only participants who are known to provide reliable pitch matches because they have
Table 2. Subjective Evaluation of Pitch-Matching Methods (VAS Means and SDs) Pitch-Matching Method CAT RIFT
SIO
Certainty
Comprehensibility
Simplicity
Comfort
70.65 (19.37) 74.41 (19.85)
91.81 (6.85) 92.00 (8.06)
51.94 (29.64) 74.82 (17.36)
74.41 (20.83) 89.29 (9.67)
Evaluation of iPod-Based Automated Tinnitus Pitch MatchingAVunderlich et al
been tested before. Unlike in some other studies (e.g., Henry et al, 2000, 2001), we had no prior knowledge about the ability of the participants to do pitch matches. Although this probably led to lower reliability coeffi cients, it should allow a more realistic view on pitch matching reliability in a clinical sample. However, the reliability of both methods is still comparable to reliabil ities of pitch-m atching procedures found in other studies (Henry et al, 2000; Henry et al, 2001, 2004). The RIFT provides the possibility to carry out repeated measurements to achieve a higher accuracy and at low cost, whereas performing CAT repetitions is almost unrealistic. Additionally, the RIFT could allow ana lyzing within-participant variability more deeply and efficiently. In the end, whether the accuracy of the pitch matching is considered as sufficient depends also on the purpose of the diagnostic. The subjective evaluation of both methods provided clear results. The iPod-based pitch-matching procedure was as well understood and performed as the procedure guided by the examiner. Information in the manual and the pop-up help windows within the RIFT application provided sufficient information for the tinnitus partici pants to perform the RIFT correctly. In addition, partici pants even rated the RIFT as easier to perform. The higher difficulty of the CAT might also have caused the saving of time in the second session that was observed for the CAT but not for the RIFT. The RIFT was probably already very economic considering time effort. Time was not limited for making a decision in both procedures. However, participants were completely independent during the RIFT, which might have left more time to listen to tones again, if a decision could not be made at once. In con trast, the presence of an examiner during the CAT might have created more pressure on the participants. The more comfortable feehng during the RIFT procedure also con firms this interpretation and is in line with research on iPod-based audiometry (Foulad et al, 2013). However, we have to add th at the items we used for the subjective evaluation have not been validated. To further evaluate and compare the usability of pitch-matching procedures, we would recommend using validated measures such as the system usability scale (Brooke, 1996).
CONCLUSIONS he purpose of this study was to investigate whether it is feasible to perform tinnitus pitch matching using portable devices and to achieve positive effects for the reliability of these procedures. The main result was that, although both methods had methodological weaknesses, pitch matching with the iPod-based RIFT procedure was as reliable as the standardized audio metric procedure in a controlled environment. However, we could not confirm that the iPod-based procedure was more reliable th an the audiom etric procedure. The
T
strongest argument for mobile devices is th at repeated measurements can be obtained easily, and thus, the error of a single measurement can be reduced. The forced-choice double-staircase procedure could also be realized in an iPod-based application and might even provide more reliable results. It is also possible to run more sophisticated probabilistic approaches. At the moment, the benefit of spending extra time on tin nitus pitch matching as a clinician may not be obvious because this information is currently not needed for therapy. However, if training such as TMNMT proves to be effective against tinnitus, this might change. F ur thermore, to receive reliable measurable information about the individual tinnitus could already mean a reduction of distress for some patients. With mobile testing, the entire diagnostic and treatm ent prepara tions could be done with less time and effort needed for both patients and attending clinical personnel. In future research, it should be analyzed if repeated mea surements can also be done completely independently by the patient at home using the same setup as in our study. Data could be sent via e-mail to the clinic, where the data could be analyzed before the training is set up for the patient. Also, it would be possible to con nect pitch-matching apps directly with apps for tinnitus treatm ent. Thus, the accessibility of standardized diag nostics and treatm ents could be largely increased.
Acknowledgments. The authors thank Andreas Wollbrink for his technical support.
REFERENCES Boulos MNK, Wheeler S, Tavares C, Jones R. (2011) How smart phones are changing the face of mobile and participatory health care: an overview, with example from eCAALYX. Biomed Eng Online 10:24. Brooke J. (1996) SUS-A quick and dirty usability scale. In: Jordan PW, Thomas B, McClelland IL, Weerdmeester BA, eds. Usability Evaluation in Industry. London, UK: Taylor and Fran cis, 189-194. Cortina JM. (1993) What is coefficient alpha? An examination of theory and applications. J Appl Psychol 78(1):98-104. Diesch E, Struve M, Rupp A, Ritter S, Hiilse M, Flor H. (2004) Enhancement of steady-state auditory evoked magnetic fields in tinnitus. Eur J Neurosci 19(4):1093-1104. Dobie RA. (2003) Depression and tinnitus. Otolaryngol Clin North Am 36(2):383-388. Eggermont JJ, Roberts LE. (2004) The neuroscience of tinnitus. Trends Neurosci 27(ll):676-682. Field LC, McEvoy MD, Smalley JC, et al. (2014) Use of an elec tronic decision support tool improves management of simulated in-hospital cardiac arrest. Resuscitation 85(1): 138-142. Foulad A, Bui P, Djalilian H. (2013) Automated audiometry using apple iOS-based application technology. Otolaryngol Head Neck Surg 149(5):700-706.
an n
Journal of the American Academy of Audiology/Volume 26, Number 2, 2015
Gliner JA, Morgan GA, Leech N. (2009) Research Methods in Applied Settings: A n Integrated Approach to Design and Analysis. 2nd ed. New York, London: Routledge. Heller AJ. (2003) Classification and epidemiology of tinnitus. Oto laryngol Clin North A m 36(2):239-248. Henry JA, Fausti SA, Flick CL, Helt WJ, Ellingson RM. (2000) Computer-automated clinical technique for tinnitus quantifica tion. A m J Audiol 9(1 j:36—49. Henry JA, Flick CL, Gilbert A, Ellingson RM, Fausti SA. (2001) Comparison of two computer-automated procedures for tinnitus pitch matching. J Rehabil Res Dev 38(5):557-566. Henry JA, Flick CL, Gilbert A, Ellingson RM, Fausti SA. (2004) Comparison of manual and computer-automated procedures for tinnitus pitch-matching. J Rehabil Res Dev 41(2): 121-138. Henry JA, McMillan GP, Thielman EJ, et al. (2013) Evaluating psychoacoustic measures for establishing presence of tinnitus. J Rehabil Res Dev 50(4):573-584. Henry JA, Meikle MB, Gilbert A. (1999) Audiometric correlates of tinnitus pitch: insights from the tinnitus data registry. In: Hazell J, ed. Proceedings o f the Sixth International Tinnitus Semi nar. Cambridge, 51-57. Huss M, Moore BCJ. (2005) Dead regions and noisiness of pure tones. Int J Audiol 44(10):599-611. Jesteadt W. (1980) An adaptive procedure for subjective judg ments. Percept Psychophys 28(l):85-88.
Norena A, Micheyl C, Chery-Croze S, Collet L. (2002) Psycho acoustic characterization of the tinnitus spectrum: implications for the underlying mechanisms of tinnitus. Audiol Neurootol 7:(6):358-369. Okamoto H, Stracke H, Stoll W, Pantev C. (2010) Listening to tailormade notched music reduces tinnitus loudness and tinnitus-related auditory cortex activity. Proc Natl Acad Sci USA 107(3):1207-1210. Pantev C, Okamoto H, Teismann H. (2012) Music-induced cortical plasticity and lateral inhibition in the human auditory cortex as foundations for tonal tinnitus treatment. Front Syst Neurosci 6:50. Pantev C, Oostenveld R, Engelien A, Ross B, Roberts LE, Hoke M. (1998) Increased auditory cortical representation in musicians. Nature 392(6678):811-814. Penner MJ, Bilger RC. (1992) Consistent within-session measures of tinnitus. J Speech Hear Res 35(3):694-700. Penner MJ, Klafter EJ. (1992) Measures of tinnitus: step size, matches to imagined tones, and masking patterns. Ear Hear 13:(6):410-416. Roberts LE, Moffat G, Bosnyak DJ. (2006) Residual inhibition functions in relation to tinnitus spectra and auditory threshold shift. Acta Otolaryngol 126:27-33. Schaette R, Kempter R. (2009) Predicting tinnitus pitch from patients’ audiograms with a computational model for the develop ment of neuronal hyperactivity. J Neurophysiol 101(6):3042-3052.
Konig O, Schaette R, Kempter R, Gross M. (2006) Course of hear ing loss and occurrence of tinnitus. Hear Res 221(l-2):59-64.
Schneider P, Scherg M, Dosch HG, Specht HJ, Gutschalk A, Rupp A. (2002) Morphology of Heschl’s gyrus reflects enhanced activation in the auditory cortex of musicians. N at Neurosci 5:(7):688-694.
Micheyl C, Delhommeau K, Perrot X, Oxenham AJ. (2006) Influ ence of musical and psychoacoustical training on pitch discrimina tion. Hear Res 219(l-2):36-47.
Szudek J, Ostevik A, Dziegielewski P, et al. (2012) Can Uhear me now? Validation of an iPod-based hearing loss screening test. J Otolaryngol Head Neck Surg 41(Suppl 1):S78-S84.
Moffat G, Adjout K, Gallego S, Thai-Van H, Collet L, Norena AJ. (2009) Effects of hearing aid fitting on the perceptual character istics of tinnitus. Hear Res 254(1-21:82—91.
Tass PA, Adamchic I, Freund H-J, von Stackelberg T, Hauptmann C. (2012) Counteracting tinnitus by acoustic coordinated reset neuromo dulation. Restor Neurol Neurosci 30(2):137-159.
Moore B, Huss M, Vickers D. (2001) Psychoacoustics of dead regions. In: Breebaart D, Houtsma A, Kohlrausch M, Prijs V, Schoonhoven R, eds. Physiological and Psychophysical Bases o f Auditory Function. Maastricht, Netherlands: Shaker, 419-A25.
Tass PA, Popovych OV. (2012) Unlearning tinnitus-related cere bral synchrony with acoustic coordinated reset stimulation: theo retical concept and modelling. Biol Cybern 106(l):27-36.
Moreti F, Pereira LD, Gielow I. (2012) Pitch-matching scanning: com parison of m usicians and non-m usicians’ perform ance. J Soc B ras Fonoaudiol 24(4):368-373. [Article in English, Portuguese]
Teismann H, Okamoto H, Pantev C. (2011) Short and intense tailor-made notched music training against tinnitus: the tinnitus frequency m atters. PLoS ONE 6(9):e24685. Tyler RS, Conrad-Armes D. (1983) Tinnitus pitch: a comparison of three measurement methods. Br J Audiol 17(2): 101-107.
Mosa ASM, Yoo I, Sheets L. (2012) A systematic review of health care applications for smartphones. BMC Med Inform Decis Mak 12:67.
Vernon J, Fenwick J. (1984) Identification of tinnitus: A plea for standardization. J Laryngol Otol 98(Suppl S9):45-53.
Norena AJ. (2011) An integrative model of tinnitus based on a central gain controlling neural sensitivity. Neurosci Biobehav Rev 35(5):1089-1109.
Vernon J, Meikle MB. (1988) Measurement of tinnitus: an update. In: Kitahara M, ed. Tinnitus. Pathophysiology and Management. Tokyo: Igaku-Shoin, 36-52.
S1S
Copyright of Journal of the American Academy of Audiology is the property of American Academy of Audiology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
Evaluation of iPod-Based Automated Tinnitus Pitch Matching DOI: 10.3766/jaaa.26.2.9 Robert Wunderlich* Alwina Stein* Alva Engell* Pia Lau* Lea Waasem* Alex Shaykevichf Claudia Rudacld: Christo Pantev*
Abstract Background: Tinnitus is the perception of sound unrelated to any external source. Diagnostic approaches to assess tinnitus characteristics such as tinnitus pitch are crucial for new attempts of tinnitus therapy. Purpose: The purpose of this study was to develop and evaluate reliable tinnitus pitch-matching pro cedures. Existing procedures usually require audiometric equipment and are time consuming. However, some patients with tinnitus find it hard to match their tinnitus in one single session. Therefore, we devel oped an iPod-based application for seif-administered tinnitus pitch matching and compared it with a standardized audiometric procedure. Study Sample: A total of 17 patients with chronic tonal tinnitus participated in two sessions including both pitch-matching procedures. Method: In the conventional audiometric test, the investigator adjusted the frequency and loudness of pure tones led by the responses of the patient. For the iPod-based procedure, we used a recursive twointerval forced-choice test that required no interaction with an investigator. Both procedures included loudness matching and testing for octave confusion. Results: The iPod-based procedure resulted in lower pitch matches as compared with the conventional audiometry. Psychometric qualities such as test-retest reliability of both methods were comparable. Par ticipants rated the iPod-based procedure as easier to perform and more comfortable to use. Conclusions: In conclusion, we find that the use of self-administered tinnitus pitch-matching procedures on a mobile device is feasible and easier in practice without any loss of reliability and validity. A major advantage is the possibility of repeated measurements without expensive equipment and experienced staff. Repeated measurements of tinnitus pitch can provide more information about the stability of the tinnitus perception and may improve the ability of participants to match their tinnitus. Key Words: Tinnitus, tinnitus pitch, pitch matching, reliability, iPod, smartphone Abbreviations: CAT = conventional audiometric test; ENT = ear, nose, and throat; RIFT = recursive two-interval forced-choice test; SD = standard deviation; SPL = sound pressure level; TMNMT = tailormade notched music training
‘ Institute for Biomagnetism and Biosignalanalysis, University Hospital of Munster, Munster, Germany; fUniversity of Western Australia, Crawley Perth, Western Australia, Australia; ^Department of Otolaryngology, University Hospital of Munster, Munster, Germany Univ.-Prof. Dr. Christo Pantev, Institute for Biomagnetism and Biosignal Analysis, University Hospital of Munster, Malmedyweg 15 D-48149 Munster, Germany; Tel: 0049 251 83 56885; Fax: 0049 251 83 56874; E-mail: [email protected] This research was supported by the Interdisziplinares Zentrum fur klinische Forschung Munster, Medical Faculty, University of Munster, Germany (Project No. CRA05). This work was presented at the TRI 2013 - 7th International TRI Conference on Tinnitus, May 15-18, 2013, Valencia, Spain.
205
Journal o f the American Academy o f Audiology/Volume 26, Number 2, 2015
INTRODUCTION innitus is a phenomenon that draws a large amount of attention in current neuroscientific research. It refers to the perception of a sound that is unrelated to any external source (Eggermont and Roberts, 2004). Although tinnitus may disappear in some cases, for 10-15% of the general population, tin nitus is an unremitting condition (Heller, 2003). An estimated 1-3% of those affected with tinnitus experi ence a significant loss in quality of life, involving sleep disturbance and even depression (Dobie, 2003). Although many questions cannot be answered yet, our understanding of the underlying neurophysiologi cal mechanisms of tinnitus is improving. Initially, tin nitus may be triggered by damage to the inner ear hair cells, but the neural generators of noise-induced tinnitus appear to be located in the central regions of the brain (Eggermont and Roberts, 2004; Norena, 2011). From the individual perspective, tinnitus is described as a whistling or beeping sound, but because there is no physical sound, it is not possible to measure its physical qualities. However, it has perceptual qualities such as pitch, loudness, and bandwidth or pulsing described by patients, which are comparable to real physical sounds. The measurement of these qualities relies on subjective information given by the individual with tinnitus. As knowledge about the neurophysiology of tinnitus grows, recent research has also brought some new per spectives on tinnitus treatment. Many of these treat ments try to reverse changes of maladaptive plasticity. The tailor-made notched music training (TMNMT), which has been applied in tonal tinnitus, aims at sup pressing hyperactive neurons by enhancing lateral inhibition from frequencies lower and higher than the tinnitus (Okamoto et al, 2010; Teismann et al, 2011; Pantev et al, 2012). Pathological neuronal syn chrony is the target of acoustic coordinated reset neuro modulation (Tass et al, 2012; Tass and Popovych, 2012). However, both types of training have not been evaluated in randomized clinical trials with large samples. Yet both sound therapies mentioned above share a common approach that addresses specific regions of the cochlea and the auditory cortex. Often these regions are related to the perceptual qualities of the tinnitus. Theoretically, the TMNMT can only work if the tinnitus frequency is covered by the notch filter in the music. The coordinated reset neuromodulation also focuses on frequencies in a specific distance to the tinnitus frequency. Conse quently, in order to be effective, such training requires reliable estimates of the tinnitus pitch. Different approaches have been made to assess the tinnitus pitch. However, up to now no procedure has been established as a clinical standard, although stan dardization is requested (Vernon and Fenwick, 1984).
T
206
The reason is the poor reliability within and between sessions of most procedures even in a controlled envi ronment. Standardized methods such as the forcedchoice double-staircase procedure have been shown to cause less within-session variability than methods of manual adjustment (Jesteadt, 1980; Penner and Bilger, 1992; Penner and Klafter, 1992; Henry et al, 2013). With a wider use of computer-automated procedures, it has also become easier to test different adaptive procedures. Among those, the two-alternative forcedchoice method provided the best reliability (Henry et al, 2001, 2004). The recursive two-interval forcedchoice test (RIFT) procedure uses a complex algorithm to bracket the pitch match to smaller intervals until only one frequency remains (Diesch et al, 2004). How ever, its psychometric qualities have not yet been com pared with those of other methods. It has also been proposed that a frequency spectrum measured with a likeness rating could be used to describe the tinnitus percept (Norena et al, 2002). The resulting curve often mirrors the hearing loss and is stable across time (Moffat et al, 2009), although it does not always provide a clear peak that can be interpreted as the prominent tinnitus pitch. Furthermore, patients are biased to rate higher frequencies generally as more similar to their tinnitus (Norena et al, 2002). To summarize, currently no proce dure is available that provides tinnitus pitch matching with high within- and between-session reliability. Besides the different procedures that are used for pitch matching, there is also no common standard to decide which ear should be tested. Although some authors rec ommend using the contralateral ear (Henry et al, 2004), others use the ipsilateral ear (Norena et al, 2002), claim that it does not make a difference, or recommend testing both ears separately (Vernon and Fenwick, 1984). Un fortunately, there is no study comparing pitch-matching reliability systematically for these options. The valida tion of tinnitus pitch-matching procedures is also gen erally difficult. Up to now, there is no objective way to measure tinnitus frequency; therefore, it is hard to define a validation criterion with which to compare pitch-matching results. Traditionally, tinnitus pitch matching is performed with audiometric equipment as part of an audiological evaluation, including measuring hearing thresholds, loudness matching, and minimal masking level. How ever, the procedure is time consuming and repeating measurements is often not practicable in clinical rou tine. Therefore, using different platforms for tinnitus pitch matching that are independent of clinical equip ment could be beneficial. Certainly, these platforms have to meet clinical standards for this kind of measure ment. Recently, smartphones have been increasingly used as a support for professionals and patients in the health care context (Mosa et al, 2012). All kinds of specific applications (apps) can be installed. These
Evaluation of iPod-Based Automated Tinnitus P itch Matching/Wunderlich et al
apps can help with disease diagnosis, management of chronic conditions, clinical communication and decision making, and standardized diagnostics (Boulos et al, 2011; Field et al, 2014). Apps for automated audiometry have already been tested. It was shown th at 96% of the thresholds measured with an iPod-based app were within 10 dB of the threshold measured with conven tional audiometry, depending on whether or not mea surem ents were performed in a sound booth (Foulad et al, 2013). The downloadable UHear app tends to over estimate hearing loss, but the authors claim th at it can be used to screen for hearing loss (Szudek et al, 2012). On a subjective level, most participants preferred iPodbased audiometry because they found it simpler, more comfortable, less intimidating, more convenient, and more controllable (Foulad et al, 2013). For pitch matching, the calibration of loudness is not as important as for measuring audiometric thresholds because loudness is not the main outcome variable. However, it might be possible th at the use of a mobile device has certain advantages beyond the user’s subjec tive preference. Because all methods of pitch matching cause variability, more repetitions can lead to more pre cise estimates of the “true” tinnitus frequency and also give information about the reliability of a single partic ipant (Tyler and Conrad-Armes, 1983). With a proce dure based on a mobile device, it would become much easier to obtain repeated measurements because indi viduals can perform testing in their own environment without much special equipment. Although there is a great potential of this new platform for improving tin nitus diagnostics and treatm ent, detailed evaluation is still needed. The goal of this study was to develop and evaluate an application for automated iPod-based tinnitus pitch matching and to compare its reliability and validity with a conventional audiometric test (CAT) that is based on a forced-choice multiple staircase procedure. The procedure chosen for the iPod-based pitch matching was the RIFT. We hypothesized higher test-retest reliability with RIFT compared with the CAT. Furthermore, we expected that RIFT would be less time consuming and would allow more repetitions during the same time.
MATERIALS AND METHODS Participants A total of 17 participants with tinnitus took part in two sessions of pitch-matching tasks. Participants were recruited via advertisements in local newspapers, our homepage, and distribution of flyers to local ear, nose, and throat (ENT) practitioners. Potential participants were asked to fill out an anamnestic questionnaire th a t we developed for this purpose. The questionnaire covered questions about the characteristics, duration,
onset, possible cause, and treatm en t of the tin n itu s as well as preexisting illnesses, hyperacusis, current medication, and alcohol and drug consumption. The participants were then scheduled for the study. Partici pants had to match the following criteria: The tinnitus had to be present for at least 3 mo. We selected only par ticipants who reported to have a tonal tinnitus (whistling or beeping sound) th at was stable in pitch over time. Hearing loss had to be less than 70 dB HL for all fre quencies between 125 Hz and 8 kHz. No participants with ENT, neurological, or psychiatric disorders were included. The sample included 5 females and 12 males (mean age = 49.44 yr, age range, 26-63 yr, standard deviation [SD] = 11.45 yr). Three participants had musical training in the past, without being professional musicians. The study was conducted according to the Declaration of Helsinki, and the study protocol was approved by the ethics committee of the Medical Faculty of the University of M unster. Each participant signed an approved informed-consent form before study enrollm ent.
Procedures The first session started with an interview to explore the occurrence, development, and history of tinnitus treatm ent as well as the criteria already specified in the questionnaire described above. Each participant received an initial ENT evaluation th at included an ear examination performed by an otorhinolaryngologist from the ENT Department. The purpose of this exami nation was to remove earwax from the participants’ ear th a t otherwise might have influenced hearing thresh olds. We also wanted to check for any illnesses of the ear canal or the tympanic membrane. Audiometric test ing was conducted in a double-walled sound-attenuated chamber (ISO 8253-1). Hearing thresholds were mea sured manually by an audiologist for the frequency range of 125 Hz to 16 kHz with an appropriate audiometer (MADSEN Astera, Denmark) and calibrated headphones (Sennheiser HDA 200). Two runs (each method once) of tinnitus pitch matching were obtained with a half-hour break for the participant to relax outside of the sound booth. A second session took place 1 wk after the first ses sion. At the end of the second session, participants were additionally tested with a tinnitus pitch likeness rating, which was not included in the analysis because of techni cal problems. The order in which the two pitch-matching m ethods were used in each session was random ly assigned. All pitch-matching procedures were performed with the ear ipsilateral to the tinnitus perception. With bilateral tinnitus, participants were asked in which ear they perceived their tinnitus as more dominant. If they were not able to name one ear as more dominant, the ear with less average hearing loss was chosen for testing. All testing sounds were pure tones. Each session lasted approximately 2.5 hr.
50 7
Journal of the American Academy of Audiology/Volume 26, Number 2, 2015
Participants were asked to rate the two procedures on four different visual analog scales (0-100) concerning comprehensibility, ease of use, certainty about the result, and comfort level of the procedure.
CAT The CAT was conducted with the same audiometric equipment used for measuring hearing thresholds, and the participants were informed about the different steps of the procedure. We used seven frequencies as starting tones in the following order: 4, 12.5, 1, 10, 2, 8, and 6 kHz. The participants were presented with the starting tone and were asked if the tinnitus pitch was higher or lower. After the participants provided their answers, the frequency of the tone was adjusted. Adjusting steps were initially half an octave until the participant changed the direction of the answer or rated the presented tone as equal to the tinnitus. The proce dure continued in one-twelfth octave steps until the participant indicated th at the presented tone was again similar to the tinnitus. This procedure was repeated for the seven different starting tones. The highest possible frequency presented to participants was 16 kHz. Ini tially, all tones were presented 10 dB above the individ ual hearing threshold. The maximal presentation level was 90 dB SPL. For the adjusted frequency, we matched loudness to the individual tinnitus loudness in 1 dB steps by asking the participant whether the tinnitus was softer or louder. The resulting seven frequencies were then con tinually entered into a two-alternative forced-choice test where the participant was to choose which frequency was more similar to the tinnitus, starting with the lowest two frequencies until only one frequency remained. To avoid octave confusion, the resulting frequency was afterward tested in a two-alternative forced-choice fashion against its octaves. Beforehand, the octaves were also matched to the loudness of the tinnitus (Vernon and Fenwick, 1984).
RIFT Testing was performed with an iPod Touch fourthgeneration Model A1367 8GB, running a custom appli cation with all necessary features specifically designed for tinnitus pitch matching. The headphone used was a Sennheiser HD 201. As a safety constraint, the ampli tude of the sounds was fixed. All sound stimuli had a maximal sound pressure level (SPL) between 75 and 90 dB SPL, measured with a 2 cm3 ear simulator (Brad & Kjaer, model 4157). Participants were first instructed how to use the application with a short manual of two pages. Additionally, pop-up windows guided the par ticipant through each step. The testing itself did not require interaction with an examiner. In the first step, participants had to match the loudness of every quarter octave from 1—16 kHz to the perceived
SO B
tinnitus loudness with a slider. Loudness values were stored as default loudness to be used later in the testing. The participant still had the opportunity to change the loudness of the tones during the actual testing. The loudness values for the frequencies between the quarter octaves were interpolated. The frequency range was lim ited to the highest frequency th at the participant was able to hear. The actual procedure was a two-alternative forced-choice test in which participants were instructed to select the frequency th at was more similar to their tinnitus (see Figure 1). For the comparisons, the fre quency range from 1-16 kHz was bisected into two equally large subintervals. Depending on the decisions, bisection and two-interval forced-choice testing was reapplied to the low subinterval, the high subinterval, or a new middle interval that was bound by the mid points of the low and the high subinterval. For example, if the participant was presented with 1 versus 4 kHz (2 octaves) as the first interval and 4 versus 16 kHz (2 octaves) as the second interval and chose 1 kHz as the first decision and 4 kHz as the second decision, the next intervals would have been 1 versus 2 kHz (1 octave) and 2 versus 4 kHz (1 octave). This procedure was repeated until one final pitch was selected. The smallest step was a twelfth octave. The testing was com pleted with an octave confusion test. If the participant’s choices were contradictory twice in a row, the test was aborted and the procedure was restarted. The RIFT was repeated three times or until the participant completed one pitch matching. For the analysis, we used the first valid pitch match and, additionally, the mean across all valid runs.
Statistical Analysis In preparation of the analyses, all frequencies were transformed into a logarithmic scale to calculate means and SDs and retransformed for better interpretation. Differences of the mean pitch matches of both methods were analyzed with a repeated-measures analysis of
iPod
10:08 M essung
Back
m
\
i
Welcher der beiden Tone ist Ihrem Tinnitus in der Tonhohe ahnlicher?
w J it,
*
fit
w . *
V Ton A
OK
Ton B
Figure 1. Graphical user interface of the RIFT. English transla tions: “Measurement. Which of the two tones is closer to your tin nitus in pitch? Tone A, Tone B.”
Evaluation of iPod-Based Automated Tinnitus Pitch Matching/Wunderlich et al
variance with two within-participant factors session (Session 1 versus Session 2) and method (CAT versus RIFT first pitch match versus RIFT mean of all runs). Test-retest reliability was calculated with the Pearson product-moment correlation coefficient between Session 1 and Session 2 for each method. Reliability of the methods was further analyzed with Cronbach’s a. Furthermore, we calculated f-tests on the means of the time duration and subjective evaluation.
RESULTS earing thresholds of the sample are displayed in Figure 2. There was no significant difference in the hearing thresholds of the right and left ears across all participants. Nine participants reported to have uni lateral left-sided or left-dom inant tinnitus, and six reported unilateral right-sided or right-dom inant tin nitus. Two participants reported th eir tin n itu s as bilateral. The mean pitch matches measured with CAT and RIFT are shown in Table 1. The analysis of variance revealed a significant main effect for the method. As the assumption of sphericity was violated, revealed via the Mauchly test (p < 0.000), the degrees of freedom were Greenhouse-Geisser corrected for the withinfactor method resulting in F a 156;18 496) = 6.132, p < 0.05. The session did not have a significant effect on the pitch matches [F(1>16) = 0.995, p > 0.5]. Planned simple contrasts revealed th at CAT and the first pitch match of RIFT provided significantly different pitch matches [F(1j16) = 7.375, p < 0.05], and CAT was differ ent from the average over three runs of RIFT [F(116) = 5.656, p < 0.05]. Averaged absolute differences were cal culated to illustrate the variability of the different methods (Table 1).
H
All test-retest correlations were significant, as shown in Table 1. The difference in the size of the Pearson r’s was not significant between the different methods. To calculate the internal consistency for the CAT, we ana lyzed the first seven matched frequencies. Cronbach’s a was 0.912 for the CAT for Session 1 and 0.859 for Ses sion 2. Because a single run of the RIFT did not provide any intermediate pitch matches, Cronbach’s a was only calculated for the three runs. The RIFT reached a Cron bach’s a of 0.864 during the first session and 0.796 dur ing the second session. It has to be noted th at for the RIFT, five participants provided only one or two valid pitch matches during Session 1. In Session 2, one par ticipant provided only one pitch match. These partic ipants had to be excluded from the calculation of Cronbach’s a, resulting in an overall sample size of 12 for the first session and 16 for the second session. The mean range of pitch matches provided by a par ticipant across both methods (CAT and RIFT) and both sessions was almost 1 octave (mean = 1935 Hz, SD = 1500 Hz). However, there was no difference between female and male participants [t(15) = 0.851, p > 0.05]. In the first session, the participants needed 20.06 min (SD = 3.09 min) to complete the pitch matching with the CAT. In the second session, the matching took 17.20 min (SD = 2.70 min). For the RIFT, participants needed 20.69 min (SD = 4.64 min) in the first session and 20.13 min (SD = 5.78 min) in the second session. Differences between the two methods were only significant for the second session (t = 2.067, p < 0.05). On a subjective level, participants rated both m eth ods as good to comprehend (Table 2). The certainty th at the procedure yielded the correct pitch match did not differ between the two methods. Participants also rated both methods as highly comprehensible but favored the RIFT concerning simplicity (t = 3.05, p < 0.01) and com fort (t = 3.18, p < 0.01).
DISCUSSION wo procedures for pitch matching were investi gated in this study. Both methods had comparable reliability. According to the criteria of Gliner et al (2009), a test-retest reliability above 0.8 can be consid ered as high, which applies only for the RIFT with three repetitions. The CAT demanded a final decision in each session by means of two-alternative forced-choice test ing. However, like any other previous decision, this decision was also confounded by a measurement error that could not be minimized in this procedure. Averag ing over all runs of the RIFT provided a better chance of canceling out this error. The internal consistency for the CAT was good, whereas the consistency for the RIFT was only acceptable. It should be taken into account th at the low number of items th at were used for the calculation of Cronbach’s a, especially in the case
T
Figure 2. M e a n h e a r i n g t h r e s h o l d s (in dB S PL ) of all 17 p articip an ts.
209
J o u rn a l o f th e A m erica n A cad em y o f A udiology/Volum e 26, N um ber 2, 2015
Table 1. Summary of Pitch-Matching Results Mean Pitch Match Pitch-Matching Method
Session 1 (SD)
Session 2 (SD)
Average Absolute Difference
Pearson r
CAT RIFT, First Pitch Match RIFT, Mean of All Runs
7,692 (1,702) 5,816 (1,504) 6,126 (1,484)
8,287 (1,632) 6,262 (1,550) 6,140 (1,479)
468.31 530.89 339.82
0.649* 0.561f 0.825*
Note: Means and SDs are expressed in Hz. *p < 0.05, one-tailed. f p < 0.01, one-tailed.
of the RIFT, deflates the value of Cronbach’s a (Cortina, 1993). Surprisingly, the RIFT resulted in lower pitch matches. A reason for this could be th at the first two decisions of this tw o-alternative forced-choice test always presented very distant frequencies. Errors in these decisions can usually result in a restriction of the frequency spectrum th at does not allow returning to lower or higher frequencies. In particular, the upper boundary of the first high interval might have caused a certain bias because it was always near the edge of the individual’s hearing. On the other hand, there might have been a bias toward high frequencies in the CAT. In the second part of this procedure, when frequencies were entered into a two-alternative forced-choice test, higher frequencies were presented later than lower fre quencies. Therefore, lower frequencies had to pass more decisions than higher frequencies. Additional variance might also have been introduced because of the different headphones used for both pro cedures. For the RIFT, we wanted to increase ecological validity by using a headphone that was not solely for audio metry. We chose to use the Sennheiser HD201 because of its flat frequency spectrum, which does not attenuate specific frequencies. Furtherm ore, we w anted to use a circum aural headphone to shield from surrounding noise. The Sennheiser HD201 is also not too expensive (approximately 21€ in Germany). Nevertheless, the dif ferent equipm ent used for both procedures was a lim itation of this study. Because two different methods and two different setups of equipment were used, it is possible th a t the equipm ent also had an effect on the results. However, small differences in the loudness output should not have influences on the test-retest reliability. Another source of variance responsible for reduced reliability, besides the method itself, was the partici pant’s general ability to perform pitch matches. Pitch matching is not a simple task for most participants.
The largest differences can be observed when musicians and nonmusicians are compared (Moreti et al, 2012). Playing an instrum ent trains the auditory system, including the auditory cortex, in a very special way that leads to an increased cortical representation and also morphological changes (Pantev et al, 1998; Schneider et al, 2002). Musicians have a lower threshold for dis criminating frequencies than nonmusicians; nonethe less, nonmusicians can be trained to perform better (Micheyl et al, 2006). Unfortunately, our sample did not give us the opportunity to analyze how musical training affected our data. Untrained participants also often confuse frequency differences with loudness dif ferences (Vernon and Meikle, 1988). Special training before pitch-matching tasks might prevent participants from mixing up both dimensions or at least ensure that participants who are not able to do so can be identified and excluded (Henry et al, 2001). Another factor that reduces the ability to match frequencies is hearing loss th at is often present in patients with tinnitus. In the region of hearing loss, the frequency discrimination is considerably reduced (Moore et al, 2001). Significant hearing loss can even result in dead regions, which leads to distorted processing of pure tones (Huss and Moore, 2005). This is even more problematic because a connection between the region of hearing loss and the frequency of tinnitus is assumed (Henry et al, 1999; Norena et al, 2002; Eggerm ont and Roberts, 2004; Konig et al, 2006; Roberts et al, 2006; Moffat et al, 2009; Schaette and Kempter, 2009). Considering these issues, pitch matching is a challenging task for the tested participants. This is reflected in the certainty ratings, which were only moderate. To minimize varia bility introduced by the tinnitus itself, it is better to exclude p a rticip a n ts w ith unstable or even atonal tinnitus and too-pronounced hearing loss. It is also a possibility to include only participants who are known to provide reliable pitch matches because they have
Table 2. Subjective Evaluation of Pitch-Matching Methods (VAS Means and SDs) Pitch-Matching Method CAT RIFT
SIO
Certainty
Comprehensibility
Simplicity
Comfort
70.65 (19.37) 74.41 (19.85)
91.81 (6.85) 92.00 (8.06)
51.94 (29.64) 74.82 (17.36)
74.41 (20.83) 89.29 (9.67)
Evaluation of iPod-Based Automated Tinnitus Pitch MatchingAVunderlich et al
been tested before. Unlike in some other studies (e.g., Henry et al, 2000, 2001), we had no prior knowledge about the ability of the participants to do pitch matches. Although this probably led to lower reliability coeffi cients, it should allow a more realistic view on pitch matching reliability in a clinical sample. However, the reliability of both methods is still comparable to reliabil ities of pitch-m atching procedures found in other studies (Henry et al, 2000; Henry et al, 2001, 2004). The RIFT provides the possibility to carry out repeated measurements to achieve a higher accuracy and at low cost, whereas performing CAT repetitions is almost unrealistic. Additionally, the RIFT could allow ana lyzing within-participant variability more deeply and efficiently. In the end, whether the accuracy of the pitch matching is considered as sufficient depends also on the purpose of the diagnostic. The subjective evaluation of both methods provided clear results. The iPod-based pitch-matching procedure was as well understood and performed as the procedure guided by the examiner. Information in the manual and the pop-up help windows within the RIFT application provided sufficient information for the tinnitus partici pants to perform the RIFT correctly. In addition, partici pants even rated the RIFT as easier to perform. The higher difficulty of the CAT might also have caused the saving of time in the second session that was observed for the CAT but not for the RIFT. The RIFT was probably already very economic considering time effort. Time was not limited for making a decision in both procedures. However, participants were completely independent during the RIFT, which might have left more time to listen to tones again, if a decision could not be made at once. In con trast, the presence of an examiner during the CAT might have created more pressure on the participants. The more comfortable feehng during the RIFT procedure also con firms this interpretation and is in line with research on iPod-based audiometry (Foulad et al, 2013). However, we have to add th at the items we used for the subjective evaluation have not been validated. To further evaluate and compare the usability of pitch-matching procedures, we would recommend using validated measures such as the system usability scale (Brooke, 1996).
CONCLUSIONS he purpose of this study was to investigate whether it is feasible to perform tinnitus pitch matching using portable devices and to achieve positive effects for the reliability of these procedures. The main result was that, although both methods had methodological weaknesses, pitch matching with the iPod-based RIFT procedure was as reliable as the standardized audio metric procedure in a controlled environment. However, we could not confirm that the iPod-based procedure was more reliable th an the audiom etric procedure. The
T
strongest argument for mobile devices is th at repeated measurements can be obtained easily, and thus, the error of a single measurement can be reduced. The forced-choice double-staircase procedure could also be realized in an iPod-based application and might even provide more reliable results. It is also possible to run more sophisticated probabilistic approaches. At the moment, the benefit of spending extra time on tin nitus pitch matching as a clinician may not be obvious because this information is currently not needed for therapy. However, if training such as TMNMT proves to be effective against tinnitus, this might change. F ur thermore, to receive reliable measurable information about the individual tinnitus could already mean a reduction of distress for some patients. With mobile testing, the entire diagnostic and treatm ent prepara tions could be done with less time and effort needed for both patients and attending clinical personnel. In future research, it should be analyzed if repeated mea surements can also be done completely independently by the patient at home using the same setup as in our study. Data could be sent via e-mail to the clinic, where the data could be analyzed before the training is set up for the patient. Also, it would be possible to con nect pitch-matching apps directly with apps for tinnitus treatm ent. Thus, the accessibility of standardized diag nostics and treatm ents could be largely increased.
Acknowledgments. The authors thank Andreas Wollbrink for his technical support.
REFERENCES Boulos MNK, Wheeler S, Tavares C, Jones R. (2011) How smart phones are changing the face of mobile and participatory health care: an overview, with example from eCAALYX. Biomed Eng Online 10:24. Brooke J. (1996) SUS-A quick and dirty usability scale. In: Jordan PW, Thomas B, McClelland IL, Weerdmeester BA, eds. Usability Evaluation in Industry. London, UK: Taylor and Fran cis, 189-194. Cortina JM. (1993) What is coefficient alpha? An examination of theory and applications. J Appl Psychol 78(1):98-104. Diesch E, Struve M, Rupp A, Ritter S, Hiilse M, Flor H. (2004) Enhancement of steady-state auditory evoked magnetic fields in tinnitus. Eur J Neurosci 19(4):1093-1104. Dobie RA. (2003) Depression and tinnitus. Otolaryngol Clin North Am 36(2):383-388. Eggermont JJ, Roberts LE. (2004) The neuroscience of tinnitus. Trends Neurosci 27(ll):676-682. Field LC, McEvoy MD, Smalley JC, et al. (2014) Use of an elec tronic decision support tool improves management of simulated in-hospital cardiac arrest. Resuscitation 85(1): 138-142. Foulad A, Bui P, Djalilian H. (2013) Automated audiometry using apple iOS-based application technology. Otolaryngol Head Neck Surg 149(5):700-706.
an n
Journal of the American Academy of Audiology/Volume 26, Number 2, 2015
Gliner JA, Morgan GA, Leech N. (2009) Research Methods in Applied Settings: A n Integrated Approach to Design and Analysis. 2nd ed. New York, London: Routledge. Heller AJ. (2003) Classification and epidemiology of tinnitus. Oto laryngol Clin North A m 36(2):239-248. Henry JA, Fausti SA, Flick CL, Helt WJ, Ellingson RM. (2000) Computer-automated clinical technique for tinnitus quantifica tion. A m J Audiol 9(1 j:36—49. Henry JA, Flick CL, Gilbert A, Ellingson RM, Fausti SA. (2001) Comparison of two computer-automated procedures for tinnitus pitch matching. J Rehabil Res Dev 38(5):557-566. Henry JA, Flick CL, Gilbert A, Ellingson RM, Fausti SA. (2004) Comparison of manual and computer-automated procedures for tinnitus pitch-matching. J Rehabil Res Dev 41(2): 121-138. Henry JA, McMillan GP, Thielman EJ, et al. (2013) Evaluating psychoacoustic measures for establishing presence of tinnitus. J Rehabil Res Dev 50(4):573-584. Henry JA, Meikle MB, Gilbert A. (1999) Audiometric correlates of tinnitus pitch: insights from the tinnitus data registry. In: Hazell J, ed. Proceedings o f the Sixth International Tinnitus Semi nar. Cambridge, 51-57. Huss M, Moore BCJ. (2005) Dead regions and noisiness of pure tones. Int J Audiol 44(10):599-611. Jesteadt W. (1980) An adaptive procedure for subjective judg ments. Percept Psychophys 28(l):85-88.
Norena A, Micheyl C, Chery-Croze S, Collet L. (2002) Psycho acoustic characterization of the tinnitus spectrum: implications for the underlying mechanisms of tinnitus. Audiol Neurootol 7:(6):358-369. Okamoto H, Stracke H, Stoll W, Pantev C. (2010) Listening to tailormade notched music reduces tinnitus loudness and tinnitus-related auditory cortex activity. Proc Natl Acad Sci USA 107(3):1207-1210. Pantev C, Okamoto H, Teismann H. (2012) Music-induced cortical plasticity and lateral inhibition in the human auditory cortex as foundations for tonal tinnitus treatment. Front Syst Neurosci 6:50. Pantev C, Oostenveld R, Engelien A, Ross B, Roberts LE, Hoke M. (1998) Increased auditory cortical representation in musicians. Nature 392(6678):811-814. Penner MJ, Bilger RC. (1992) Consistent within-session measures of tinnitus. J Speech Hear Res 35(3):694-700. Penner MJ, Klafter EJ. (1992) Measures of tinnitus: step size, matches to imagined tones, and masking patterns. Ear Hear 13:(6):410-416. Roberts LE, Moffat G, Bosnyak DJ. (2006) Residual inhibition functions in relation to tinnitus spectra and auditory threshold shift. Acta Otolaryngol 126:27-33. Schaette R, Kempter R. (2009) Predicting tinnitus pitch from patients’ audiograms with a computational model for the develop ment of neuronal hyperactivity. J Neurophysiol 101(6):3042-3052.
Konig O, Schaette R, Kempter R, Gross M. (2006) Course of hear ing loss and occurrence of tinnitus. Hear Res 221(l-2):59-64.
Schneider P, Scherg M, Dosch HG, Specht HJ, Gutschalk A, Rupp A. (2002) Morphology of Heschl’s gyrus reflects enhanced activation in the auditory cortex of musicians. N at Neurosci 5:(7):688-694.
Micheyl C, Delhommeau K, Perrot X, Oxenham AJ. (2006) Influ ence of musical and psychoacoustical training on pitch discrimina tion. Hear Res 219(l-2):36-47.
Szudek J, Ostevik A, Dziegielewski P, et al. (2012) Can Uhear me now? Validation of an iPod-based hearing loss screening test. J Otolaryngol Head Neck Surg 41(Suppl 1):S78-S84.
Moffat G, Adjout K, Gallego S, Thai-Van H, Collet L, Norena AJ. (2009) Effects of hearing aid fitting on the perceptual character istics of tinnitus. Hear Res 254(1-21:82—91.
Tass PA, Adamchic I, Freund H-J, von Stackelberg T, Hauptmann C. (2012) Counteracting tinnitus by acoustic coordinated reset neuromo dulation. Restor Neurol Neurosci 30(2):137-159.
Moore B, Huss M, Vickers D. (2001) Psychoacoustics of dead regions. In: Breebaart D, Houtsma A, Kohlrausch M, Prijs V, Schoonhoven R, eds. Physiological and Psychophysical Bases o f Auditory Function. Maastricht, Netherlands: Shaker, 419-A25.
Tass PA, Popovych OV. (2012) Unlearning tinnitus-related cere bral synchrony with acoustic coordinated reset stimulation: theo retical concept and modelling. Biol Cybern 106(l):27-36.
Moreti F, Pereira LD, Gielow I. (2012) Pitch-matching scanning: com parison of m usicians and non-m usicians’ perform ance. J Soc B ras Fonoaudiol 24(4):368-373. [Article in English, Portuguese]
Teismann H, Okamoto H, Pantev C. (2011) Short and intense tailor-made notched music training against tinnitus: the tinnitus frequency m atters. PLoS ONE 6(9):e24685. Tyler RS, Conrad-Armes D. (1983) Tinnitus pitch: a comparison of three measurement methods. Br J Audiol 17(2): 101-107.
Mosa ASM, Yoo I, Sheets L. (2012) A systematic review of health care applications for smartphones. BMC Med Inform Decis Mak 12:67.
Vernon J, Fenwick J. (1984) Identification of tinnitus: A plea for standardization. J Laryngol Otol 98(Suppl S9):45-53.
Norena AJ. (2011) An integrative model of tinnitus based on a central gain controlling neural sensitivity. Neurosci Biobehav Rev 35(5):1089-1109.
Vernon J, Meikle MB. (1988) Measurement of tinnitus: an update. In: Kitahara M, ed. Tinnitus. Pathophysiology and Management. Tokyo: Igaku-Shoin, 36-52.
S1S
Copyright of Journal of the American Academy of Audiology is the property of American Academy of Audiology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
