Abstract
Purpose
This study was designed to investigate methods to help patients suffering from unilateral tinnitus synthesizing an auditory replica of their tinnitus.
Materials and methods
Two semi-automatic methods (A and B) derived from the auditory threshold of the patient and a method (C) combining a pure tone and a narrow band-pass noise centred on an adjustable frequency were devised and rated on their likeness over two test sessions. A third test evaluated the stability over time of the synthesized tinnitus replica built with method C, and its proneness to merge with the patient’s tinnitus. Patients were then asked to try and control the lateralisation of this single percept through the adjustment of the tinnitus replica level.
Results
The first two tests showed that seven out of ten patients chose the tinnitus replica built with method C as their preferred one. The third test, performed on twelve patients, revealed pitch tuning was rather stable over a week interval. It showed that eight patients were able to consistently match the central frequency of the synthesized tinnitus (presented to the contralateral ear) to their own tinnitus, which leaded to a unique tinnitus percept. The lateralisation displacement was consistent across patients and revealed an average range of 29dB to obtain a full lateral shift from the ipsilateral to the contralateral side.
Conclusions
Although spectrally simpler than the semi-automatic methods, method C could replicate patients’ tinnitus, to some extent. When a unique percept between synthesized tinnitus and patients’ tinnitus arose, lateralisation of this percept was achieved.
1
Introduction
Subjective tinnitus (i.e. perception of a sound without any external or internal auditory stimulation ) is a highly prevalent otoneurological symptom for which there is currently no causally oriented treatment . In the USA up to 16 % of the general population has been found to suffer from frequent tinnitus . Although most people can become accustomed to a tinnitus percept, clinical subjective tinnitus can significantly alter one’s quality of life . In the most recent pathophysiological models, tinnitus was no longer considered to originate in the cochlea but was rather seen to be a consequence of neuroplastic changes in the central auditory and non auditory pathways occurring as a consequence of sensory deprivation after cochlear damage . Indeed, the clinical patterns of tinnitus were similar to those observed in chronic pain following amputation , where conditioning techniques using virtual reality (VR) have been shown both to be theoretically interesting and effectively useful . Hypothetically, if the patients manage to transfer their subjective auditory perception to a real sound similar to their tinnitus percept, it could allow them to take control of the tinnitus-like percept they manipulate, see and hear.
The aim of the presented study was to build a tinnitus avatar that would be further exploited in a therapeutic protocol, where the patient would be invited to manipulate his/her tinnitus avatar in a virtual reality environment . This implies creating a “virtual tinnitus” that could be manipulated in the virtual environment. In order to reproduce a realistic sensation, the virtual tinnitus needs to “imitate” the patient’s tinnitus behaviour. Although tinnitus can be perceived inside or outside the head, the majority of tinnitus is unilateral (with a predominance for the left ear) or bilateral . To limit the number of different factors, only patients with an unilateral tinnitus were recruited in the presented study.
In headphone listening conditions, a sound source played over headphones can be externalized using head related transfer function (HRTF) but can also move “inside the head” by an action of source lateralisation. In normal headphone listening conditions, diotic signals give rise to a unique auditory event located in the centre of the head. Increasing the gain of one of the signals “pushes” the sound image towards the corresponding ear. If dichotic signals, differing in pitch, are presented, then the fusion does not occur any more and two separate auditory events are heard respectively in the right and left ear. Lackner conducted an experiment on six patients that were suffering from tinnitus resulting from cortical and peripheral damages. The central frequency of their tinnitus was matched with an external sound played in the contralateral ear (opposite ear of the ear where the tinnitus is located – stated as ipsilateral ear). For four of them, the internal and the external signals fused and therefore resulted in a unique percept. Changing the gain of the external signal changed the perceived position of this percept. This experiment was done using only one pure frequency tone that reproduced part of the tinnitus spectrum.
The tinnitus sensation is complex to characterize because only the patient suffering from tinnitus can recreate a similar sound with synthetic signals that do not necessarily “sound” like his/her sensation. The tools can only help him/her in this recreation. Nageris et al. reported judiciously that the variability of the measured tinnitus pitch was dependent on different factors such as: test procedure, homogeneity of the population sample and origin of the tinnitus. Indeed, different signals and methods have been carried out to characterize tinnitus dominant pitch. Although the repeated within-session pitch matching was shown to be comparable with that of external tones matching, tinnitus pitch matching was shown to vary significantly across sessions . In most studies, tinnitus characterisation would consist of adjusting a pure tone frequency to match the dominant pitch of the perceived tinnitus . Different procedures have been used to help the patient to create a replica of their tinnitus, avoid octave error and work on the reliability of their results. In , patients used a method of adjustment where the frequency of the external tone was swept by the experimenter. Tyler and Conrad-Armes advised to use either method of adjustment or adaptive method by “bracketing” the tinnitus pitch. Mitchell et al. and Penner and Klafter worked on pure tone matching using a frequency adaptive forced-choice-double-staircase procedure. Moore et al. used a bracketing method centred on a previously defined frequency. Most of these methods are time-consuming therefore tiring for the patient (Tyler suggested repeating the procedure at least seven times to have a final result) and, most of the time, needs an external examiner to manage the frequency change. Furthermore, the reliability of tinnitus matching with a single pure tone might be inappropriate to represent the timbre of tinnitus . In order to better portray the complex tinnitus sensation, Norena then Roberts proposed a method to estimate the various components contributing to the global tinnitus sensation. In , ten subjects were asked to rate successively the similarity of their tinnitus against a collection of pure tones ranging from 500 Hz to 14 kHz. A combination of these tones weighted by their respective ratings forms what the authors named “the internal spectrum of perceived tinnitus”. It was observed that this estimated internal spectrum roughly followed the frequency profile of the auditory threshold with the main components falling into the hearing loss area. Using a similar approach, Roberts et al. proposed to characterize the tinnitus spectrum using eleven sound clips with a central frequency ranging from 0.5 kHz to 12 kHz. For each subject, the bandwidth of these eleven sound clips (pure tone, 5 % of the centre frequency (CF) at − 10 dB, 15 % of CF at − 10 dB) was chosen according to a self-described tinnitus timbre as being respectively “tonal”, “ringing” or “hissing”. The tinnitus spectrum was modelled through the individual rating of these eleven sounds with the tinnitus as the comparison using the Borg CR100 scale . Conducted with 32 subjects, the study also revealed that the tinnitus spectrum spanned the region of the hearing loss . In these two methods the patients had to adjust manually defined frequency tones. Conclusions from Norena et al. and Roberts et al. suggest that an automatic or semi-automatic tinnitus synthesis method could be derived from the measurement of the subject’s auditory threshold shift. The tinnitus matching process could then be simplified and reliable, if the tinnitus sensation is stable.
In the presented study, three methods of creating a credible auditory image (tinnitus avatar) of the patient’s perceived tinnitus were investigated. The reliability was then investigated by a test-retest of the synthesized tinnitus built with one of the methods. Finally, if the combination of synthesized tinnitus and the patient’s tinnitus gave rise to a unique percept, then the apparent lateralisation of this percept was investigated.
2
Materials and methods
In the following section, the three methods as well as the three subjective test procedures are presented. The study received the approval of the local ethical committee (Comité de Protections des Personnes Ile de France VIII, Boulogne Billancourt, France).
2.1
Tinnitus synthesis methods
Three adaptive subject-driven methods have been devised to create an acoustic replica of the patient’s tinnitus. The devised methods were based on a combination of two complex stimuli respectively composed of a sinusoidal component and a filtered noise component. This combination was meant to allow the patient to tune between different tinnitus sensations, often reported in the literature as sounding “tonal” or “ringing”, “hissing”, “whooshing”, etc.
All tinnitus synthesis methods presented below used the contralateral ear to drive the tinnitus matching. This choice was in line with the CIBA recommendation that advocates the use of the contralateral ear for pitch matching if the tinnitus is unilateral .
2.1.1
Auditory threshold measurement
The tinnitus synthesis methods A and B described below were based on the estimated auditory threshold curve of the patient. Consequently, audiograms of the patients have been systematically measured for both ears before inviting them to tune their respective avatar with the different methods.
An audiogram program has been designed to estimate the hearing curve of the patient using an ascendant method. Sequences of one-second fixed-frequency pure tone separated by one second of silence were presented to the patient with a step increase of 3 dB until he/she indicated hearing the sound by pressing the space-bar of the keyboard. The stimulated ear and the frequency were randomly chosen. Once the sound was detected, another tone was presented to the patient until all the measurements were done. The hearing threshold was measured from 125 Hz to 16 kHz. Measured frequencies corresponded to the frequency distribution defined in ISO 389–1 complemented with the frequency distribution defined in ISO 389–5 norms in order to cover the range from 125 Hz to 16 kHz. From 125 Hz to 8 kHz, frequencies were roughly distributed in 1/3 octave band. From 8 kHz to 16 kHz, frequencies were distributed in 1/6 octave band.
2.1.2
Semi automatic methods A and B
The assumption behind the design of method A and B was based on Norena’s conclusion that the tinnitus’ internal spectrum mirrors the frequency profile of the auditory loss of the ipsilateral ear, i.e. the ear where the tinnitus was located. For both methods, the synthetic tinnitus was a combination of two complex stimuli. The first stimulus was composed of a series of sinusoids and the second was a broadband filtered noise. The frequency content of the tinnitus was automatically built and the only task of the patient was to determine the relative gain of the sinusoid aggregate stimulus and of the noise stimulus (one slider controlling the balance between both components). A global gain slider allowed for the matching of the overall signal level to the tinnitus loudness.
2.1.2.1
Method A
The two components were sculpted in frequency to follow the auditory threshold curve (in dB SPL) of the ipsilateral ear. The frequencies of the sinusoids were those used in the auditory threshold measurement procedure. Their relative gains fitted the corresponding auditory thresholds. The frequency envelope of the filter applied to the noise stimulus component was derived from the same auditory threshold curve using linear interpolation between the different measured frequencies. Method A took the ipsilateral auditory threshold curve as the basis of the tinnitus reproduction. Different typologies of hearing loss were equally considered. Indeed, method A weighted equivalently a unilateral notch present on the ipsilateral ear, and a symmetrical hearing impairment.
2.1.2.2
Method B
In method B, it was hypothesized that the analysis of the auditory threshold asymmetry between the two ears may be used as an indicator of the tinnitus’ internal spectrum. Accordingly, the tinnitus avatar spectrum built with method B aimed at mirroring this asymmetry. The relative weight of the sinusoids and the noise filter envelope were built upon the auditory thresholds difference (in dB) between the ipsilateral and the contralateral ears. As for method A, the frequency profile was fixed and the patients were only asked to adjust a slider controlling the level balance between the two components and a global gain slider to match the overall signal level to the loudness of the tinnitus.
2.1.3
“Manual method” C
Method C was close to single pitch matching. Pitch matching has often been based on adaptive procedure or bracketing, with fixed discretized frequency, sometimes managed by instructor. In this case, the patients were asked to span the frequency range to find the matching frequency. The synthetic tinnitus was composed of two stimuli: a pure tone and a 1/6 octave band noise centred on a single and adjustable frequency. The band-pass filter was either implemented using a bi-quadratic parametric filter or in the Fourier domain, using a 1/6 octave band rectangular window. Patients were invited to synthesize the tinnitus avatar by adjusting the central frequency common to both components, their relative gains and a global level.
2.2
Test session 1
This first test session was intended to gather preliminary observations about the different tinnitus synthesis methods. Objective data was collected about the different control parameters tuned by the patients together with their spontaneous comments about the task, the overall similarity of the created tinnitus avatars and the preferred tinnitus avatar (i.e. the one that sounds the most like their tinnitus). The whole session lasted from forty-five minutes to one hour.
2.2.1
Patients
Ten patients (two females, eight males) aged between 26 and 60 (mean 41) and reporting tonal and stable unilateral tinnitus (five left and five right) were involved in a tinnitus avatar synthesis session where they could create their tinnitus avatar using the three methods A, B and C successively.
2.2.2
Apparatus
Both the audiogram and tinnitus matching procedures were performed on a Mac Book Pro equipped with an external RME Fireface 400 sound card set at 44.1 kHz sampling rate. The methods were implemented using MAX/MSP software. Audiogram measurements and test session 1 were performed with Koss R80 headphones. The total electro-acoustic chain comprising the computer, the sound card and the headphones was calibrated using a B&K artificial ear in order to derive the dB FS to dB SPL conversion curve.
For each synthesis method, a simple graphical user interface (GUI) was designed to help the patient to tune the different method parameters. The GUI consisted of different virtual sliders represented on the screen, which the patient controlled with the mouse.
Once the patient was satisfied with the tinnitus matching, an excerpt of 10 seconds was saved in an audio sound file in order to allow for its subjective evaluation in the different test sessions. Sound files were encoded with 24 bits dynamic range and 44.1 kHz sampling rate. In order to better exploit the dynamic of the sound files, signals were amplified before being recorded so as to approach an RMS level of − 10 dB FS. The corresponding gain was saved and compensated for when playing back the sound file, in order to restore the tinnitus avatar to its original level.
2.2.3
Procedure
The session started with the measurement of the auditory threshold curve of the patient on both ears using the procedure described in Section 2.1.1 . Immediately after, the patient proceeded with the creation of his/her tinnitus avatars using the three tinnitus synthesis methods A, B and C (described in Section 2.1.2 and Section 2.1.3 ) successively. For all patients, the order of the methods was kept the same. Before starting a given method, the patient received explanations about the GUI and the significance of each tuning parameter (virtual sliders controlled with the mouse). Once the tinnitus matching was accomplished, a short sound file excerpt of the tinnitus avatar was automatically recorded. The patient was invited to have a short debriefing with the experimenter, to give oral feedback or any free comments about the task itself or about the global or detailed resemblance between his/her tinnitus and the recently created avatar.
As a last step, after having accomplished the matching using the three creation methods, the patient was invited to compare the three tinnitus avatars and to indicate which was the most similar to his/her tinnitus. To do so, the patient listened alternatively to the three avatar sound files, and was free to switch from one to the other. As during the matching process, the avatars were played back on the contralateral ear. However the patient could not change the playback level, which was set according to what he/she had tuned during the matching step. Once the most resembling tinnitus avatar was chosen, the patient was asked to comment on his/her choice.
2.3
Test session 2
The time interval between the first and second sessions was between two and five months.
The second session was dedicated to the subjective assessment of the tinnitus avatars synthesized during the first session. As exposed in Section 2.1 , the different synthesis methods were devised to portray individual specificities of the tinnitus spectrum. In order to verify whether these methods actually succeeded in their goals, the patients were invited to compare the likeness of their own tinnitus avatars with that of non-individual avatars such as those that were created by the other patients. Although necessarily based on subjective judgements, and despite a small number of patients, the goal was to obtain some objective quantification of the suitability of the different proposed tinnitus synthesis methods.
2.3.1
Patients
This session involved a subset of five patients who participated in the first session (two females, three males, mean age = 46, age range 34–60) reporting tonal and stable unilateral tinnitus (three left and two right).
2.3.2
Apparatus
The equipment was the same that the one used in test session 1, described in 2.2.2 .
2.3.3
Procedure
Two different tests were organised within the same session, the total duration of which was about thirty minutes. For both tests, twelve stimuli including individual and non-individual tinnitus avatars were presented to each patient. They consisted of a subset of the tinnitus sound files collected during the first session. Three of these stimuli were the individual avatars that the patient had built during the first session with methods A, B and C. The nine others were non-individual, and corresponded to the tinnitus avatars that were respectively built and selected as their best by the other nine patients of the first session. It should be noted that the stimuli corpus presented was different for each patient. Ten stimuli were common to all (the ten best avatars including the patient’s own best). Accordingly, they did not necessarily originate from the same synthesis method. Two stimuli were patient specific, i.e. they corresponded to the two other individual avatars that the patient built during the first session but did not select as his/her best.
2.3.3.1
Visual Analogue scale (VAS)
Each stimulus was played back in loop in the contralateral ear and the patient was free to listen to it as long as he/she wanted. The patient was first asked to adjust the level of the tinnitus avatar in accordance to his/her tinnitus. The level was initialized to − 150 dB and the patient could adjust it with a vertical slider represented on a GUI. In order to avoid sound distortion of the avatar and hearing injury, the level was constrained to stay below an upper limit of 0 dB FS (107 dB SPL at 1 kHz) otherwise the tinnitus avatar was simply muted. This precaution was especially important when a patient was adjusting the level of a non-individual tinnitus avatar as his/her hearing threshold may differ from that of the patient who designed it. After matching its level, the patient was asked to rate the level of resemblance of the tinnitus avatar with his/her tinnitus on a VAS (from 0 = completely different to 100 = identical). The adjustment of the signal level and the similarity judgment were made using two vertical faders represented on a GUI. The twelve tinnitus avatars were successively scored using the same procedure. The three avatars created by the patient were presented first followed by the non-individual tinnitus avatars. The patient was not aware of the identity and origin of the tinnitus avatars.
2.3.3.2
Pairwise Comparison test
The twelve stimuli were used in a forced choice pairwise comparison. All stimuli were played back in loop in the contralateral ear using the level tuned by the patient during the VAS test. For each pair the patient could switch freely at any time between both stimuli, but was no longer allowed to modify the level. The patient was invited to select which stimulus from the pair was the most similar to his/her tinnitus. As a full pairwise comparison method with twelve stimuli would have led to a rather tedious task (66 pairs), we restricted the number of pairs. For each patient the three individual avatars were compared to each other (3 pairs), and the three individual avatars were also compared to the other nine non-individual tinnitus avatars (27 pairs), which led to a total of 30 pairs. For each patient the global “preference” score of the different stimuli was obtained by adding the number of times it was chosen, divided by the number of times it was involved in a pairwise comparison (11 times for individual avatars and 3 times for the non-individual ones).
2.4
Test session 3
A third test session investigated, on the one hand, the reproducibility and stability of the tinnitus avatar over a one-week interval and, on the other hand, whether the patient’s tinnitus and the tinnitus avatar merged together and gave rise to a unique percept. Based on the observations and results of the first two test sessions (see section 3.1 and 3.2 for details), the preferred method was chosen to be method C. Only this method was used in this third test session.
2.4.1
Patients
A new set of twelve patients (six females and six males) aged from 24 to 69 (mean 45 years) were involved in this third test session. They all reported unilateral tinnitus (six right and six left).
2.4.2
Apparatus
The equipment was similar to the test sessions 1 and 2, apart from the headphones. Test session 3 was performed using BeyerDynamic 990 open headphones which were used during the therapeutic application . The calibration was adapted accordingly.
2.4.3
Procedure
The procedure was conducted in two phases, separated by a one-week interval.
2.4.3.1
First phase
The auditory threshold curve of the patients was first measured on both ears using the audiogram method described in Section 2.1.1 . The patients were then invited to a training task where they had to match two sinusoidal tones in pitch and loudness. A reference tone of fixed frequency was played in the ipsilateral ear. A variable frequency tone was played in the contralateral ear. The task of the patients was to match the pitch and loudness of the variable tone to that of the reference tone by moving sliders integrated in a graphical user interface situated in front of them. The training was repeated for five tones with increasing frequency (500 Hz, 1000 Hz, 2000 Hz, 4000 Hz and 8000 Hz). Once the training was completed, the patients were invited to create a tinnitus avatar (hereafter called C1) with method C as described in Section 2.1.3 .
2.4.3.2
Second phase
One week later, the patients were asked to create a new version (hereafter called C2) of their tinnitus avatar following the same procedure as in the first phase. This was also preceded by a frequency-matching training phase. After creating the new tinnitus avatar, the patients were invited to evaluate the similarity of the C1 and C2 tinnitus avatar versions with their tinnitus. Each avatar was presented to the contralateral ear, and the patients were not allowed to modify its level. The evaluation was performed with two methods.
In the first evaluation method, the patients had to evaluate the similarity of each tinnitus avatar version with their tinnitus using a VAS from 0 (“very different”) to 100 (“identical”).
In the second evaluation method, the two avatars were compared to each other. The patient could freely switch from one tinnitus avatar version to the other as many times as they wanted. The two versions were denoted “A” and “B” and the patients were not aware of their respective identity. Using a 7-point horizontal scale (from left to right: ”A much more similar to the tinnitus than B”, “A more similar to the tinnitus than B”, “A and B are equivalent”, “B more similar to the tinnitus than A”, “B much more similar to the tinnitus than A”), the patients were asked to judge the similarity according to three criteria: “global similarity”, “pitch similarity”, “loudness similarity”.
A final test was performed in which the patients were asked to choose their preferred tinnitus avatar version (C1 or C2) and check if the combination of the tinnitus avatar with their tinnitus was merging into a unique percept. The patients were then successively invited to tune the tinnitus avatar gain in order to localize the tinnitus image at five different lateral directions (an apparent direction towards the contralateral side, an apparent position in between the median plane and the contralateral ear – approximately on the same sagittal plane as the eye, an apparent direction in the median plane, an apparent position in between the median plane and the ipsilateral ear and, an apparent direction towards the ipsilateral side). Each direction tuning was repeated twice.
2
Materials and methods
In the following section, the three methods as well as the three subjective test procedures are presented. The study received the approval of the local ethical committee (Comité de Protections des Personnes Ile de France VIII, Boulogne Billancourt, France).
2.1
Tinnitus synthesis methods
Three adaptive subject-driven methods have been devised to create an acoustic replica of the patient’s tinnitus. The devised methods were based on a combination of two complex stimuli respectively composed of a sinusoidal component and a filtered noise component. This combination was meant to allow the patient to tune between different tinnitus sensations, often reported in the literature as sounding “tonal” or “ringing”, “hissing”, “whooshing”, etc.
All tinnitus synthesis methods presented below used the contralateral ear to drive the tinnitus matching. This choice was in line with the CIBA recommendation that advocates the use of the contralateral ear for pitch matching if the tinnitus is unilateral .
2.1.1
Auditory threshold measurement
The tinnitus synthesis methods A and B described below were based on the estimated auditory threshold curve of the patient. Consequently, audiograms of the patients have been systematically measured for both ears before inviting them to tune their respective avatar with the different methods.
An audiogram program has been designed to estimate the hearing curve of the patient using an ascendant method. Sequences of one-second fixed-frequency pure tone separated by one second of silence were presented to the patient with a step increase of 3 dB until he/she indicated hearing the sound by pressing the space-bar of the keyboard. The stimulated ear and the frequency were randomly chosen. Once the sound was detected, another tone was presented to the patient until all the measurements were done. The hearing threshold was measured from 125 Hz to 16 kHz. Measured frequencies corresponded to the frequency distribution defined in ISO 389–1 complemented with the frequency distribution defined in ISO 389–5 norms in order to cover the range from 125 Hz to 16 kHz. From 125 Hz to 8 kHz, frequencies were roughly distributed in 1/3 octave band. From 8 kHz to 16 kHz, frequencies were distributed in 1/6 octave band.
2.1.2
Semi automatic methods A and B
The assumption behind the design of method A and B was based on Norena’s conclusion that the tinnitus’ internal spectrum mirrors the frequency profile of the auditory loss of the ipsilateral ear, i.e. the ear where the tinnitus was located. For both methods, the synthetic tinnitus was a combination of two complex stimuli. The first stimulus was composed of a series of sinusoids and the second was a broadband filtered noise. The frequency content of the tinnitus was automatically built and the only task of the patient was to determine the relative gain of the sinusoid aggregate stimulus and of the noise stimulus (one slider controlling the balance between both components). A global gain slider allowed for the matching of the overall signal level to the tinnitus loudness.
2.1.2.1
Method A
The two components were sculpted in frequency to follow the auditory threshold curve (in dB SPL) of the ipsilateral ear. The frequencies of the sinusoids were those used in the auditory threshold measurement procedure. Their relative gains fitted the corresponding auditory thresholds. The frequency envelope of the filter applied to the noise stimulus component was derived from the same auditory threshold curve using linear interpolation between the different measured frequencies. Method A took the ipsilateral auditory threshold curve as the basis of the tinnitus reproduction. Different typologies of hearing loss were equally considered. Indeed, method A weighted equivalently a unilateral notch present on the ipsilateral ear, and a symmetrical hearing impairment.
2.1.2.2
Method B
In method B, it was hypothesized that the analysis of the auditory threshold asymmetry between the two ears may be used as an indicator of the tinnitus’ internal spectrum. Accordingly, the tinnitus avatar spectrum built with method B aimed at mirroring this asymmetry. The relative weight of the sinusoids and the noise filter envelope were built upon the auditory thresholds difference (in dB) between the ipsilateral and the contralateral ears. As for method A, the frequency profile was fixed and the patients were only asked to adjust a slider controlling the level balance between the two components and a global gain slider to match the overall signal level to the loudness of the tinnitus.
2.1.3
“Manual method” C
Method C was close to single pitch matching. Pitch matching has often been based on adaptive procedure or bracketing, with fixed discretized frequency, sometimes managed by instructor. In this case, the patients were asked to span the frequency range to find the matching frequency. The synthetic tinnitus was composed of two stimuli: a pure tone and a 1/6 octave band noise centred on a single and adjustable frequency. The band-pass filter was either implemented using a bi-quadratic parametric filter or in the Fourier domain, using a 1/6 octave band rectangular window. Patients were invited to synthesize the tinnitus avatar by adjusting the central frequency common to both components, their relative gains and a global level.
2.2
Test session 1
This first test session was intended to gather preliminary observations about the different tinnitus synthesis methods. Objective data was collected about the different control parameters tuned by the patients together with their spontaneous comments about the task, the overall similarity of the created tinnitus avatars and the preferred tinnitus avatar (i.e. the one that sounds the most like their tinnitus). The whole session lasted from forty-five minutes to one hour.
2.2.1
Patients
Ten patients (two females, eight males) aged between 26 and 60 (mean 41) and reporting tonal and stable unilateral tinnitus (five left and five right) were involved in a tinnitus avatar synthesis session where they could create their tinnitus avatar using the three methods A, B and C successively.
2.2.2
Apparatus
Both the audiogram and tinnitus matching procedures were performed on a Mac Book Pro equipped with an external RME Fireface 400 sound card set at 44.1 kHz sampling rate. The methods were implemented using MAX/MSP software. Audiogram measurements and test session 1 were performed with Koss R80 headphones. The total electro-acoustic chain comprising the computer, the sound card and the headphones was calibrated using a B&K artificial ear in order to derive the dB FS to dB SPL conversion curve.
For each synthesis method, a simple graphical user interface (GUI) was designed to help the patient to tune the different method parameters. The GUI consisted of different virtual sliders represented on the screen, which the patient controlled with the mouse.
Once the patient was satisfied with the tinnitus matching, an excerpt of 10 seconds was saved in an audio sound file in order to allow for its subjective evaluation in the different test sessions. Sound files were encoded with 24 bits dynamic range and 44.1 kHz sampling rate. In order to better exploit the dynamic of the sound files, signals were amplified before being recorded so as to approach an RMS level of − 10 dB FS. The corresponding gain was saved and compensated for when playing back the sound file, in order to restore the tinnitus avatar to its original level.
2.2.3
Procedure
The session started with the measurement of the auditory threshold curve of the patient on both ears using the procedure described in Section 2.1.1 . Immediately after, the patient proceeded with the creation of his/her tinnitus avatars using the three tinnitus synthesis methods A, B and C (described in Section 2.1.2 and Section 2.1.3 ) successively. For all patients, the order of the methods was kept the same. Before starting a given method, the patient received explanations about the GUI and the significance of each tuning parameter (virtual sliders controlled with the mouse). Once the tinnitus matching was accomplished, a short sound file excerpt of the tinnitus avatar was automatically recorded. The patient was invited to have a short debriefing with the experimenter, to give oral feedback or any free comments about the task itself or about the global or detailed resemblance between his/her tinnitus and the recently created avatar.
As a last step, after having accomplished the matching using the three creation methods, the patient was invited to compare the three tinnitus avatars and to indicate which was the most similar to his/her tinnitus. To do so, the patient listened alternatively to the three avatar sound files, and was free to switch from one to the other. As during the matching process, the avatars were played back on the contralateral ear. However the patient could not change the playback level, which was set according to what he/she had tuned during the matching step. Once the most resembling tinnitus avatar was chosen, the patient was asked to comment on his/her choice.
2.3
Test session 2
The time interval between the first and second sessions was between two and five months.
The second session was dedicated to the subjective assessment of the tinnitus avatars synthesized during the first session. As exposed in Section 2.1 , the different synthesis methods were devised to portray individual specificities of the tinnitus spectrum. In order to verify whether these methods actually succeeded in their goals, the patients were invited to compare the likeness of their own tinnitus avatars with that of non-individual avatars such as those that were created by the other patients. Although necessarily based on subjective judgements, and despite a small number of patients, the goal was to obtain some objective quantification of the suitability of the different proposed tinnitus synthesis methods.
2.3.1
Patients
This session involved a subset of five patients who participated in the first session (two females, three males, mean age = 46, age range 34–60) reporting tonal and stable unilateral tinnitus (three left and two right).
2.3.2
Apparatus
The equipment was the same that the one used in test session 1, described in 2.2.2 .
2.3.3
Procedure
Two different tests were organised within the same session, the total duration of which was about thirty minutes. For both tests, twelve stimuli including individual and non-individual tinnitus avatars were presented to each patient. They consisted of a subset of the tinnitus sound files collected during the first session. Three of these stimuli were the individual avatars that the patient had built during the first session with methods A, B and C. The nine others were non-individual, and corresponded to the tinnitus avatars that were respectively built and selected as their best by the other nine patients of the first session. It should be noted that the stimuli corpus presented was different for each patient. Ten stimuli were common to all (the ten best avatars including the patient’s own best). Accordingly, they did not necessarily originate from the same synthesis method. Two stimuli were patient specific, i.e. they corresponded to the two other individual avatars that the patient built during the first session but did not select as his/her best.
2.3.3.1
Visual Analogue scale (VAS)
Each stimulus was played back in loop in the contralateral ear and the patient was free to listen to it as long as he/she wanted. The patient was first asked to adjust the level of the tinnitus avatar in accordance to his/her tinnitus. The level was initialized to − 150 dB and the patient could adjust it with a vertical slider represented on a GUI. In order to avoid sound distortion of the avatar and hearing injury, the level was constrained to stay below an upper limit of 0 dB FS (107 dB SPL at 1 kHz) otherwise the tinnitus avatar was simply muted. This precaution was especially important when a patient was adjusting the level of a non-individual tinnitus avatar as his/her hearing threshold may differ from that of the patient who designed it. After matching its level, the patient was asked to rate the level of resemblance of the tinnitus avatar with his/her tinnitus on a VAS (from 0 = completely different to 100 = identical). The adjustment of the signal level and the similarity judgment were made using two vertical faders represented on a GUI. The twelve tinnitus avatars were successively scored using the same procedure. The three avatars created by the patient were presented first followed by the non-individual tinnitus avatars. The patient was not aware of the identity and origin of the tinnitus avatars.
2.3.3.2
Pairwise Comparison test
The twelve stimuli were used in a forced choice pairwise comparison. All stimuli were played back in loop in the contralateral ear using the level tuned by the patient during the VAS test. For each pair the patient could switch freely at any time between both stimuli, but was no longer allowed to modify the level. The patient was invited to select which stimulus from the pair was the most similar to his/her tinnitus. As a full pairwise comparison method with twelve stimuli would have led to a rather tedious task (66 pairs), we restricted the number of pairs. For each patient the three individual avatars were compared to each other (3 pairs), and the three individual avatars were also compared to the other nine non-individual tinnitus avatars (27 pairs), which led to a total of 30 pairs. For each patient the global “preference” score of the different stimuli was obtained by adding the number of times it was chosen, divided by the number of times it was involved in a pairwise comparison (11 times for individual avatars and 3 times for the non-individual ones).
2.4
Test session 3
A third test session investigated, on the one hand, the reproducibility and stability of the tinnitus avatar over a one-week interval and, on the other hand, whether the patient’s tinnitus and the tinnitus avatar merged together and gave rise to a unique percept. Based on the observations and results of the first two test sessions (see section 3.1 and 3.2 for details), the preferred method was chosen to be method C. Only this method was used in this third test session.
2.4.1
Patients
A new set of twelve patients (six females and six males) aged from 24 to 69 (mean 45 years) were involved in this third test session. They all reported unilateral tinnitus (six right and six left).
2.4.2
Apparatus
The equipment was similar to the test sessions 1 and 2, apart from the headphones. Test session 3 was performed using BeyerDynamic 990 open headphones which were used during the therapeutic application . The calibration was adapted accordingly.
2.4.3
Procedure
The procedure was conducted in two phases, separated by a one-week interval.
2.4.3.1
First phase
The auditory threshold curve of the patients was first measured on both ears using the audiogram method described in Section 2.1.1 . The patients were then invited to a training task where they had to match two sinusoidal tones in pitch and loudness. A reference tone of fixed frequency was played in the ipsilateral ear. A variable frequency tone was played in the contralateral ear. The task of the patients was to match the pitch and loudness of the variable tone to that of the reference tone by moving sliders integrated in a graphical user interface situated in front of them. The training was repeated for five tones with increasing frequency (500 Hz, 1000 Hz, 2000 Hz, 4000 Hz and 8000 Hz). Once the training was completed, the patients were invited to create a tinnitus avatar (hereafter called C1) with method C as described in Section 2.1.3 .
2.4.3.2
Second phase
One week later, the patients were asked to create a new version (hereafter called C2) of their tinnitus avatar following the same procedure as in the first phase. This was also preceded by a frequency-matching training phase. After creating the new tinnitus avatar, the patients were invited to evaluate the similarity of the C1 and C2 tinnitus avatar versions with their tinnitus. Each avatar was presented to the contralateral ear, and the patients were not allowed to modify its level. The evaluation was performed with two methods.
In the first evaluation method, the patients had to evaluate the similarity of each tinnitus avatar version with their tinnitus using a VAS from 0 (“very different”) to 100 (“identical”).
In the second evaluation method, the two avatars were compared to each other. The patient could freely switch from one tinnitus avatar version to the other as many times as they wanted. The two versions were denoted “A” and “B” and the patients were not aware of their respective identity. Using a 7-point horizontal scale (from left to right: ”A much more similar to the tinnitus than B”, “A more similar to the tinnitus than B”, “A and B are equivalent”, “B more similar to the tinnitus than A”, “B much more similar to the tinnitus than A”), the patients were asked to judge the similarity according to three criteria: “global similarity”, “pitch similarity”, “loudness similarity”.
A final test was performed in which the patients were asked to choose their preferred tinnitus avatar version (C1 or C2) and check if the combination of the tinnitus avatar with their tinnitus was merging into a unique percept. The patients were then successively invited to tune the tinnitus avatar gain in order to localize the tinnitus image at five different lateral directions (an apparent direction towards the contralateral side, an apparent position in between the median plane and the contralateral ear – approximately on the same sagittal plane as the eye, an apparent direction in the median plane, an apparent position in between the median plane and the ipsilateral ear and, an apparent direction towards the ipsilateral side). Each direction tuning was repeated twice.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
