Journal of the International Telemedicine Academy

Interdisciplinary approach to the problem of tinnitus and electronic support for its diagnosis and therapy

Andrzej CZYŻEWSKI^*

^* Gdansk University of Technology, Multimedia Systems Department,
Narutowicza 11/12, 80-952 Gdansk, Poland
and Center of Excellence PROXIM (International Center for Hearing and Speech)
e-mail: ac@pg.gda.pl

Abstract

Tinnitus generating process explanation based on signal quantization theory was proposed. Current work was presented concerning an ultrasound tinnitus device employing dither noise as a masker. The device is being engineered at the Gdansk University of Technology in a close co-operation with the Institute of Physiology and Pathology of Hearing.

---

Introduction

Tinnitus is usually defined as perceived simple sounds for which there is no acoustic stimuli. It is assumed that this discomfort, by which approximately 15% of the total population may be affected, is not a disease but only a symptom that most commonly occurs in elderly age. Tinnitus is qualified as objective or subjective. Objective ear noises occur in the case of circulatory system disorder, e.g. intracranial hypertension can cause pulsatile tinnitus, in cases such as vascular abnormalities e.g. middle ear motility due to reduced blood supply caused by narrowed blood vessels, in addition in myoclonies of the palate muscles, in pathologies of the temporo-mandibular joint, in disorders of the eustachian tube [1]. Thus, the diagnostics and treatment of objective ear noises is the domain of medical specialists. Subjective ear noises are the phantom perception of neuronal signals, and they originate from entirely nonacoustic causes, i.e. without the stimulation of the ear cochlea. Subjective ear noises, as opposed to objective, are of special interest to acoustics and other specialists of digital signal processing.

According to the definition proposed by Jastreboff [2], ear noises (tinnitus) result from the improper activity of the auditory system nerves, that is not caused by any combination of external sounds. Thus, it is an already mentioned phantom perception of a neural signal that comes to the cortex from the auditory pathways but is created in other processes than a normal stimulation of the ear and auditory system caused by external sounds.

Newest tinnitus hypotheses concentrate on the explanation of this phenomenon based on unequal activity of I and II type afferent fibers, the influence of the efferent system, gate-control theory, mutual spontaneous activity of the auditory nerve fibers, influence of the sympathetic nervous system, etc. [3-7].

Mechanisms of Ear Noise Generation and Attempts to Interpret them on Interdisciplinary Grounds

In most cases the basis of ear noises is related to the ear cochlea pathology, i.e. the disorder of the outer auditory cells mechanics. This hypothesis is based on the increase of the electromotility response of hair cells. The disorder of the electromechanics causes the malfunctioning of the ionic canals, the improper concentration of calcium ions inside the external auditory cells, and the disorder of the biosynthesis of proteins. There exists a hypothesis according to which the origin of ear noises is the nonuniform degeneration of the outer ear auditory cells. When the outer ear hair cells are damaged within the area of the basilar membrane, and at the same time the inner ear hair cells stay inviolable, the unbalanced activity of these two populations of cells is responsible for starting a series of processes which result in the perception of ear noises. This is a so-called hypothesis of disharmonic damage of inner and outer hair cells [2,3,7].

Tinnitus often occurs when the threshold of hearing is increased because of the loss in hearing caused by the diseases of the inner ear. As described in the previous paragraph, this may be caused by hair cells degradation, which results in the elevation of the level of signals that activate neurons. However, before we observed such an increased threshold - e.g. as the result of a disease or otosclerosis - signals had been been perceived as auditory stimuli at higher parts of the auditory pathway. In consequence, an additional mechanism of weak acoustic signals quantization - which is caused by the increased threshold of neural auditory cells activity - is intruded. The theories on this phenomenon that exist in audiology do not directly take signal quantization mechanisms into account. The mentioned quantization occurs due to the existence of the threshold characteristic in the transmission pathway. Such interpretation becomes possible only when we take advantage of the knowledge on electric signals processing developed in other scientific fields, e.g. on the ground of digital signal processing. Below we will present the interpretation of the ear noises phenomenon based on the theory of the phonic signal quantization. Moreover, based on digital signal processing, a special methodology that eliminates noise resulting from the process of the threshold quantization has been worked out. This methodology is known under the term of 'dithering'. Shortly speaking, the technique means adding some noise to useful signals of low level. As the result, the process of spontaneous noise generation caused by the threshold characteristic is stopped. It is easy to notice, that this technique is very similar to the ways ear noises are coped with in audiology where masking noise provided by a special device called a masker is used. The efficiency of such elimination techniques developed for ear as well as quantization noise arising spontaneously in electronic circuits is widely known, and may be a good justification for the interpretation that defines ear noises as a direct consequence of weak phonic signals quantization in threshold circuits. The above mentioned issues are thoroughly discussed in the monograph by Czyzewski A., Kostek B., Skarzynski H., "Application of computer technology to audiology, phoniatry and speech therapy" published by Academic Press, Warsaw [8]. The conducted analysis shows how the interpretation of noise origin in quantizing circuits and its elimination through the means of additional masking noise (dither) may be used to explain the phenomena related to ear noises.

Typical transformation functions of a quantizer are described by the following dependencies [9]:

(1)

or:

(2)

where: x - value of a sample before quantization (at input), D - value of the quantization step, [ ] - operator returning the integer closest to a given real number.

In the case of complex input signals with large amplitudes, successive errors are uncorrelated and the spectrum of error power density resembles the spectrum of white noise. The error signal is also not correlated with the input signal. The distribution of error probability for a quantizer whose transformation function is defined by the formula (3) shown below is a rectangular window function.

(3)

For complex input signals a maximum error is equal to the least meaning bit (LSB), and the samples of a quantization error d_n - assuming that the approximation is accurate - may be treated as independent from input signals. For such input signals, homogenous quantization is a good means for creating a model based on adding white noise to input signals. However, for input signals of low level the additive white noise model is not appropriate. In such cases the error is greatly dependant on input. Signals that fall within the range of (-D/2, D/2] are assigned the value of '0' by the converter, and are not transmitted through the pathway (this is called 'digital deafness'). In such a situation, there is no output and the error is equal to the input signal of opposite sign. This type of error is noticeable when listening, and as such is a side-effect of quantization.

The objective of dither techniques is to modify the statistical characteristics of the total error. In quantization systems that do not base on dithering, the instantaneous error is defined by the function of an input signal. If the input signal is not complex and its amplitude is comparable to the value of the quantization step, the error becomes highly dependant on the input signal and causes modulation noise as well as deformations that can be heard. Using dithering signals of well-formed statistical characteristics may result in deformations similar to a stable white noise.

In today's digital auditory pathways, dithering with a triangle function of probability density and the value between peaks equal to 2 LSB is used. Dither noise is thus an additive noise introduced into the signal, most often prior to quantization. An averaged response occurring at the output of the converter, and defined as the function of an input signal may be expressed by the following formula:

(4)

where: p_u(u) - density of noise probability distribution defined for a rectangular distribution noise as:

(5)

where: x - value between peaks of dither noise voltage.

Fig. 1 illustrates a basic phenomenon taking place when the input signal of an analog-to-digital converter has the amplitude comparable to the threshold of quantization.

Figure 1. Effects of low amplitude signal quantization and the influence of dither noise: (a) "digital deafness", (b) "binary quantization", (c) dither eliminates the range of converter insensitivity, (d) "smoothing the response" in the case of binary quantization.

Figure 2. Effect of linearization of a conversion characteristic upon applying dither of different levels corresponding to fractions of quantization step.

A closer look at the way dither noise influences quantization within the initial part of the quantizer characteristics may lead to a conclusion that if a constantly present dither noise was associated with the quantizer characteristic (see Fig. 2), and a proper choice of noise level was made, the quantization steps would smooth and the quantization characteristic would become a straight line thus causing the quantization error and the related noise to decline.

Fig. 3a shows the result of a sinusoidal signal quantization obtained without dither noise. Fig. 3b depicts the result of the same signal quantization, however in this case dither noise is introduced. Figs. 3c and 3d show that averaging the representation may result in retrieving a nearly original signal from the quantized one. It is worth noticing that hearing has a distinct ability of integrating, so one may expect that similar processes take place in the auditory pathway. Fig. 4 illustrates how adding dither noise influences the reduction of harmonic deformations.

Figure 3. Effects of quantizing the signal of the amplitude corresponding to the quantization step: (a) harmonic signal directly after quantization, (b) quantization with dither, (c) signal from previous figure averaged over the time of 32-periods, (d) result of averaging over the time of 960 periods.

Figure 4. Spectrum of a quantized harmonic signal of the amplitude corresponding to the threshold of quantization (a) the same signal spectrum when dither noise is applied at the a/d converter input (b).

The power of noise at output when the input signal is static may be defined as:

(6)

When dither of the Gauss distribution defined by formula (7) is used, noise modulation does not occur.

(7)

Applying the Gauss noise reduces quantization errors and is relatively simple to implement.

(8)

(9)

Introducing a 'masking' dither noise brings desirable effects related to the elimination of a converter insensitivity range and the reduction of deformations, that occur for the quantization of very low amplitude signals. The audibility of masking noise may be reduced by forming its spectrum so that the energy of noise increases within the rage of high frequencies. The same rules apply to masking ear noises, which shows a direct similarity of the phenomena taking place in electronic and biological systems of signal transmission.

Electronic Methods for Ear Noise Reduction

The problem of ear noise and hypersensitivity to sounds as well as ways to improve the situation of patients suffering from these disorders are well described in rich literature and patent descriptions. In general, such methods base on adding external noise to affected ears in order to mask internal noise or get the patients habituated (accustomed) to subjectively perceived noise. In addition to rich literature on the discussed subject, one can also find technical solutions implemented in devices used for tinnitus reduction in many patents.

The oldest American patent (No. 4,034,741) dates from the year of 1977. It concerns a noise generator and transmitter with a circuit which can be switched in order to produce a wave of various shape from an active noise source supplied by an integrated circuit amplifier. The invention rationalizes previously mentioned applications and was used to help to invoke natural sleep. The technical solution proposed in this patent is now out of date. A much newer invention has been described in patent No. 4,222,393. External sounds of different pitch are presented one after another to tinnitus patients, so that they can identify the particular sound of the same pitch as the subjective ear noise. Then a power operated sound generator of frequencies extending above and below the range of perceived pitch is provided to the patient. The generator's role is to mask the tinnitus sound. The sound generator may be combined with a hearing aid if necessary, and it may be carried in the same manner as a hearing aid. The initial energy of the generator falls within the range of frequencies from 1000 Hz, 5000 Hz to 10000 Hz. Some of the concepts of this invention seem to be justified, e.g. the idea of adapting the shape of the masking sound spectrum to the patient's individual needs. However, electronic solutions basing mostly on discrete elements are rather obsolete from the technological point of view.

There are also European patents concerning ear noises masking devices. One of them is described in patent No. 449 860 B1. The device comprises an electronic circuit producing sound spectrum designed to mask ear noises. The main idea is that the spectrum generated by the receiver includes the linear spectrum of the basic tone, and the basic frequency is adjustable. The solution proposed in this patent is too simplified and as such cannot be treated as universal or applicable to all patients' needs.

Quite recently an American patent has been published (No. 5,403,262) that describes a tinnitus masking device and method for producing a masking signal with a selected center frequency, bandwidth, and volume. A random noise generator, clock circuit, and switched capacitance filter bank are all employed to produce the masking signal of the desired bandwidth and frequency. The masking signal is then provided to a volume control unit where it is amplified before the delivery to a tinnitus sufferer's ears by speakers or headphones. However, the proposed system cannot be used in diagnosing ear noises. The next American patent application dealing with the discussed technical problem is No. 5,788,656. Patients hearing ringing or other sounds are treated by an electronic stimulation system. An important element of this system is an electronically actuated probe. The probe is applied a complex signal whose frequency falls within the auditory range which is done in order to make the probe vibrate in accordance with the applied signal. The signal is produced by two adjustable audio-frequency oscillators. One of the oscillators operates in low frequencies (of about 400 Hz) while the other uses high frequencies (of about 1000). The oscillators have their outputs connected and amplified. The resulting signal is supplied to the probe. The vibrations emitted by the probe must be properly related to the sonic frequencies of the tinnitus sounds perceived by the patient. The probe is placed close to the patient's cochlea whereto the vibrations are transmitted stimulating this organ and bringing relief from a tinnitus disorder. The described solution differs from previous ones in the way acoustic signals are delivered to the patients' ears which in this case is realized through then use of probes. A similar solution, which involves a hearing device with a vibrating direct drive, has been described in patent No. 5,795,287. The vibrating direct drive hearing device stimulates hearing by vibrating the object with which it is coupled. The frequency, intensity and phase of a generated tone may be selected by a user. Additionally, a second tone or a background sound may be selected. Another device for diagnosing and treating hearing disorders including a supersonic transducer which has a resonance frequency in the supersonic range has been introduced in an American patent No. 6,068,590.

A European patent (No. 9611047) published in 1996 is another example. Its American version dates from the year 2000 and was registered under the number of 6,047,074. It concerns a programmable digital hearing aid that may use the mode of operation appropriate for tinnitus therapy carried out in combination with the correction of other hearing disorders. An arrangement for generating a tinnitus therapy signal has been included in the signal processing chain. The tinnitus therapy signal is combined with the useful signal, dependent on the mode of operation which has been selected or set. The patent does not include the description of any coordinated tinnitus diagnosis methods, neither gives it information on how external signals are delivered, formed, recorded and transmitted to the hearing aid acting as a masking device.

The analysis of available literature and patents mentioned above leads to a conclusion that the inventors of electronic devices do not clearly see the analogy between the phenomena accompanying the origin of subjective ear noises and the way an acoustic pathway works. The above fact brought the author to the decision to work out his own solutions in the field of diagnosing and therapy methods, of which some will be presented in the following sections.

Engineered Solutions

Solutions worked out in the Gdansk University of Technology and the Warsaw Institute of Physiology and Pathology of Hearing concern methods of creating new interactive information services, diagnosing, and rehabilitating patients suffering from ear noises and hypersensitivity to sounds. The services utilize computer-based techniques and telecommunication. In particular, a multimedia software that employs the worked out concepts and is installable through a local network or available via the Internet gives the opportunity to carry out tinnitus massive examinations. Due to a special method used to adjust masking signals to an individual patient's needs, it also enables to rehabilitate patients through masking noises or habituation. A rather simple to implement function of this system is a direct delivery of information to the patients suffering from tinnitus. Another group of solutions concerns the miniature tinnitus maskers, and is based on the interpretation of mechanisms of the ear noise origin described in the previous section.

The main idea of the proposed diagnostic system is the automatization of hearing examinations which is based on a computer questionnaire (see the Internet portal http://www.telewelfare.com/) and the presentation of sounds that resemble typical subjective ear noises. The patients must fill in the questionnaire giving detailed information, and then they are asked to point out sounds that most closely resemble the ear noise they perceive. Based on both the questionnaire algorithmic analysis and the selection of sounds, the decision is made whether a patient is free from tinnitus or belongs to a group of risk. The results of the examination, including the causes and the proposed treatment, are then passed to the patient. At this stage it is reasonable to contact a chosen specialist or to take part in a teleconference organized through the means available within the discussed computer system. To simplify further contact with doctors, the system may assign an individual identification number to a patient. The described diagnostic tool is coupled with a rehabilitation tool that employs a programmable tinnitus masker. The patients may choose from different masking noises replayed by the system via a soundboard and earphones, and then load the selected sounds to a private digital recorder in order to use them in further therapy. In such a way a multimedia computer is successfully utilized to inform patients about tinnitus problems, carry out examinations, and organize a proper therapy based on optimized masking of various types of ear noises achieved through externally supplied filtered noise, or synthetic or natural sounds. The spectra of some sounds that have been selected for tinnitus therapy are presented in Fig. 5.

The discussed system may also be applied in the form of a software for multimedia PCs on condition that they have a sound board with earphones, and a communication port capable to send digital samples to miniature digital players. The software used for tinnitus diagnosing and rehabilitation may be installed locally or on a remote network server with which a patient's computer can communicate.

The electronic questionnaire used in the discussed system contains 30 questions designed by the specialists from the Warsaw Institute of Physiology and Pathology of Hearing. The questions concern the results of previous hearing examinations, genetic conditions, history of ear diseases, giddiness, as well as the character, intensity, duration, and frequency of perceived ear noises, patient's exposition to noise in present day life, drug taking, hypertension and other factors. Sound material used for a comparative examination of subjective ear noises comprises compressed sound files, that contain dither noise of a triangular or rectangular distribution of spectral power density, simple tones of various frequencies, harmonic or nonharmonic multitones, and different types of natural and synthetic sounds. The number of sounds available to the patient during the comparative examination is 16, however, the most important are dither noise sounds. As already mentioned, based on the algorithmic analysis of the questionnaire and the patient's selection of sounds, the computer qualifies a person as free from tinnitus or belonging to a group of risk.

Figure 5. Spectra of selected sounds used in the therapy of tinnitus: (a) sea shimmer, (b) birds singing.

The applied decision algorithm grounds on the experts knowledge and is deterministic. It uses weights that scale the questionnaire answers in such a way that their sum adequately illustrates the degree of tinnitus threat. Phonic range sounds selected by patients (possibly after additional consultations with specialists) may be loaded from the "Tinnitus" system to a sound file player equipped with miniature earphones, and then used as maskers or applied in the therapy based on habituation. It is worth noticing, that patients suffering from ear noise should not have ear totally blocked, and for this they should use open earphones. The basic advantage of such a sound generation method is that the downloaded sounds, that are coded in the mp3 format, may be directly used. A modern digital player is light, small, and buttery-supplied; it does not include any mechanical elements in the carrier drive - it is a perfect portable device to replay a good noise masker (see Fig. 6).

Figure 6. Sound files player built into an electronic watch

Based on described concepts an implementation of the Public Diagnostic and Rehabilitation System for Tinnitus and Sound Hypersensitivity Sufferers called "Tinnitus" has been developed in the HTML and Java languages, and is available as a web application (url: www.telewelfare.com/) or on a CD-ROM from which it may be locally installed. The proposed system was subjected to pilot tests carried out by the Warsaw Tinnitus Clinic of the Institute of Physiology and Pathology of Hearing. The application home page is presented in Fig. 7.

Figure 7. Main GUI for "Tinnitus" Web service

Figure 8. Piezoelectric converters, that are the basis of an ultrasound bone stimulator - ear noise masker (a), highly miniaturized signal processors used to generate dither noise at 30 kHz frequency (b)

Currently the Gdansk University of Technology in a close co-operation with the Institute of Physiology and Pathology of Hearing carries out work to construct an ultrasound tinnitus device employing dither noise as a masker. The application of an ultrasound converter (see Fig. 8) enables to obtain advantageous characteristics of the masker because [10]:

1. It is possible to employ bone conductivity for the transmission of masking noise.

2. It is possible to increase the noise level without making it audible and invoking hearing tiredness.

3. An ultrasound converter of small dimensions makes a further significant miniaturization possible.

This research is in progress, thus further results should be expected after having examined a number of tinnitus patients. Some preliminary results are visible in Fig. 9.

Figure 9. Average results of audiometry testing of 3 patients without a masker and with an ultrasound masking signal.

Acknowledgements

Research subsidized by the Polish Ministry of Education and Science, Warsaw, Poland. Project grant No. 3 T11E028293