INDIVIDUAL PROFILING OF PERCEIVED TINNITUS BY DEVELOPING TINNITUS ANALYZER SOFTWARE. Baishali Chaudhury. Thesis Submitted to the Faculty of

Transcription

1 INDIVIDUAL PROFILING OF PERCEIVED TINNITUS BY DEVELOPING TINNITUS ANALYZER SOFTWARE by Baishali Chaudhury Thesis Submitted to the Faculty of The College of Engineering and Computer Science in Partial Fulfillment of the Requirements for the Degree of Master of Science Florida Atlantic University Boca Raton, Florida May 2010

2

3 ACKNOWLEDGEMENTS I would like to thank Dr. Abhijit Pandya, Professor of Computer & Electrical Engineering and Computer Science, Florida Atlantic University, for his assistance and patient corrections in the preparation of this thesis. I also wish to thank Dr. Mirjana Pavlovic for providing me with enough resources to understand the subject well. I am extremely grateful to my professors who have encouraged me during the entire time of my research work. I am also thankful to Dr. Ali Danesh from the Audiology Laboratory, Department of Communication Sciences & Disorder for providing valuable information regarding clinical studies for tinnitus and giving me great advices for improving my work. I would specially like to thank Dr. Pandya for helping me get an internship with Motorola which helped me a lot financially to support my entire Master studies. I am grateful to all faculty as well as student members of the TCN lab who have assisted and advised me time and again through group meetings and various insightful discussions which ultimately helped to improve my thesis. I would like to thank the department and the graduate college for their assistance and guidance to write this thesis. Last but not the least I would like to thank my family and especially my fiancé Laki for his immense help, encouragement and moral support that helped me maintain my calm throughout the research work and the writing of this thesis manuscript. iii

4 ABSTRACT Author: Title: Baishali Chaudhury Individual Profiling Of Perceived Tinnitus by Developing Tinnitus Analyzer Software Institution: Thesis Advisor: Degree: Florida Atlantic University Dr. Abhijit Pandya Master of Science Year: 2010 Tinnitus is a conscious perception of phantom sounds in the absence of external acoustic stimuli, and masking is one of the popular ways to treat it. Due to the variation in the perceived tinnitus sound from patient to patient, the usefulness of masking therapy cannot be generalized. Thus, it is important to first determine the feasibility of masking therapy on a particular patient, by quantifying the tinnitus sound, and then generate an appropriate masking signal. This paper aims to achieve this kind of individual profiling by developing interactive software Tinnitus Analyzer, based on clinical approach. The developed software has been proposed to be used in place of traditional clinical methods and this software (as a part of the future work) will be implemented in the practical scenario involving real tinnitus patients. iv

5 TABLE OF CONTENTS LIST OF FIGURES... vii CHAPTER 1: INTRODUCTION... 1 Definitions... 1 Historical Aspects... 2 Epidemiology... 3 Problem Statement... 6 Contribution... 7 Thesis Organization... 9 CHAPTER 2: BACKGROUND Hearing Mechanism & Anatomy of Ear Causes & Mechanisms of Tinnitus Prevalent Treatments- A Literature Survey Characterization of tinnitus Sound CHAPTER 3: SYSTEM ARCHITECTURE Proposed Scheme Audiogram Generator v

6 Tinnitus Signal Generator Masker Analyzer CHAPTER 4: APPLIED EXPERIMENTATION CHAPTER 5: FUTURE WORK & CONCLUSION BIBLIOGRAPHY vi

7 LIST OF FIGURES Figure 2.1. Anatomy of outer ear Figure 2.2. Anatomy of middle ear Figure 2.3. Anatomy of inner ear Figure 2.4. Hair cells Figure 3.1. Block Diagram of the proposed software tool Figure 3.2. Flow chart of the audiogram generator Figure 3.3. Flow chart of pitch test Figure 3.4. Flow chart of loudness test Figure 3.5. Block diagram for sound localization Figure 3.6. Schematic representation of STFT Figure 3.7. STFT, manipulation of spectral frames and IFFT of the frames Figure 4.1. Welcome Screen Figure 4.2. Audiogram generator Figure 4.3. Audiogram Plotted Figure 4.4. Tinnitus sound Generator Figure 4.5. Masker Analyzer vii

8 CHAPTER 1 INTRODUCTION 1.1 Definitions The word tinnitus comes from the Latin verb tinnire which means to ring, though in common English usage tinnitus primarily means a ringing in the ears. The first logged use of the word dates back to 1693 (Andersson et al. 2005), in Blanchard s Physician Dictionary, where tinnitus has been defined as Tinnitus Aurium, a certain buzzing or tingling in the ears proceeding from obstruction or something that irritates the ear, whereby the air that is shut up is continually moved the beating of arteries, and the drume of the ear is lightly verberated, whences arises a buzzing and a noife Certain languages describe the tinnitus phenomenon with a variety of words. The French have assigned five words for regular use, each describes a particular timbre or quality of sound. While the Swedish termed the word oronsus for tinnitus which means air breeze, since many tinnitus sufferers would describe their noise as an air breeze. However a more accurate definition has been given by McFadden ( Andersson et al. 2005) who describes it as: Tinnitus is the conscious expression of a sound that originates in an involuntary manner in the head of its owner, or may appear to him to do so 1

9 Hence Tinnitus can be defined as the conscious perception of sound that cannot be attributed to an external acoustic stimulation (Henry and Meikle 2000; Lokenberg; Han et al. 2009). It is sometimes referred to as phantom auditory experience. This percept can be tonal, hissing, ringing, whistling or cricket-like sound. Tinnitus must be distinguished from auditory hallucinations (Hearing and Balance 1997). Unlike tinnitus the latter are phantom experiences of hearing meaningful sounds, such as people speaking or music. 1.2 Historical Aspects The experience of the perception of phantom sounds has appeared in many historical medical records which date back all the way to Babylonians, and have been reviewed by Stephens and Feldmann as reported by Tyler (2000). Historical descriptions of tinnitus have depended highly on cultural factors. While the ancient Oriental mysticism regarded tinnitus as sensitivity to the divine, the Egyptians and Romans believed that tinnitus occur from a bewitched ear. There are also several references to tinnitus in texts within Islamic medical tradition (Tyler 2000; Andersson et al. 2005) The first advances in understanding Tinnitus occurred around 17 th century. This led to a shift from the belief that tinnitus occurred from trapped air in the ear, towards a concept more familiar to the modern researchers- which links occurrence of tinnitus to disorders in ears and brain. Another landmark in tinnitus advancement occurred in 1821 with a publication by Itard as reported by Andersson et al (2005). Major contributions of Itard included the distinction between Tinnitus arising from sound (objective tinnitus), and tinnitus arising from false sounds (subjective tinnitus); and the negative effect of tinnitus 2

10 on an individual. In the late 19 th century Joseph Tonybee and William Wilde were identified as the pre-eminent in this area, whose research work on tinnitus and ear disease, established a relationship between the two. MacNaughton produced the first book in English on Tinnitus in 1891, which deals with the classification of tinnitus based on the site of origin and a review of contemporary treatments (Andersson et al. 2005) Twentieth century, saw the rise of electronic instruments to measure hearing accurately using audiometry, which helped to measure the hearing sensitivity in tinnitus sufferers. Fowler is credited with the first attempts to determine the matching and masking characteristics of tinnitus (as reported by Koster et al. 2004). Fowler developed a protocol for the examination of the tinnitus patients in terms of quality of the perceived sound, level of distress and ontological health of the patient. From the above presented brief historical review on Tinnitus, we can conclude that otological conditions of the tinnitus patients as well as the severity levels as experienced by them should be taken into account while treating them. But the most important point to be considered here is that understanding and treatment of tinnitus is an ever evolving Science. The most contemporary research treatments for tinnitus, have been discussed in chapter Epidemiology The epidemiology of tinnitus has been a part of many investigations for years but the quantitative results differ in terms of the definition and duration of tinnitus (Meikle et al. 2007; Herraiz et al. 2007; Bo and Ambrosetti 2007; Zhang et al., Lockwood et al. 2002; Guitton and Dudai 2007; Pridmore et al.; Tyler et al. 2008). However an overall 3

11 approximation indicates that tinnitus prevails in % of the population. Most of us have experienced ringing in our ears in the absence of an external sound. Typically the sensation is associated with a reversible cause such as listening to loud music, fever, use of aspirin or quinine, or transient perturbations of the middle ear and subsides over a period of time ranging from a few seconds to a few days. In this context tinnitus can almost be termed as a universal experience. A famous experiment conducted by Heller and Bergman showed that among a group of subjects kept in a sound proof room, 94 % of them experience tinnitus like perception, though similar experiments conducted by Newby and Levine have not conformed quantitatively to the previous results (reported by Andersson et al. 2005). Regardless of the figurative results, the fact that normal-hearing persons perceive tinnitus like sound when placed in silent room, still prevails. However, in 5 15% of the general population, the tinnitus sensation is unremitting (Andersson et al. 2005; Tyler 2000; Chan 2009; Axelsson and Ringdahl 1989). The number of residents in the U.S. affected by tinnitus is estimated at 37 million to 40 million. Given that the U.S. population is approaching 296 million, about 12% to 14%, or one in seven to eight Americans, is affected with tinnitus. This condition may affect 30% of adults. Experts also estimate that 15% of Americans have experienced tinnitus that lasts longer than five minutes, that 155 million (over 50% of the population) have sought medical care for tinnitus, and that 6% of sufferers report being incapacitated by tinnitus. Several studies have been conducted to study the population distribution for tinnitus. But these tests do not include people with transient tinnitus. One of the most robust and comprehensive study was taken by UK medical research council Institute of Hearing 4

12 research and reported by Davis and El Rafaie (2000) (mentioned by Chan 2009). In a longitudinal study of hearing, 10.1 % of the adults were reported to have prolonged tinnitus in which 5 % reported the tinnitus sound as severely annoying and 0.5 % said that their tinnitus negatively affected their daily activities (Axelsson and Ringdahl 1989). Andersson et al. (2005) did an investigation to make an attempt to distinguish current, 12- month and lifetime tinnitus in a random sample. The figures were 25.4 % for lifetime, 21.5 % for 12- month and a 17.8 % point prevalence. Approximately 60 % of the point prevalence group had tinnitus often or always. Annoyance was rated as average or severe in 1/3 of the participating subjects. Only a minority reported a severe affect on their lives which negatively affected their ability to lead a normal life. Tinnitus is often accompanied with hearing loss. In (Palmer et al. 2002, Phoon et al. 1983) it has been shown that workers (in noisy environment) with tinnitus had consistently higher hearing thresholds at both high and low frequencies than those with no tinnitus. This finding remained even after adjusting for differences in sex, age and ethnic group composition. An early study by Hinchcliffe (as mentioned by Andersson et al ) of 800 subjects showed that tinnitus showed a tendency to increase with age from 21 %(18-24 years), 27 %(25-34 years), 24 %( years), 27%(45 54 years), 39 %( years) and 37 %( years). Therefore, approximately the increase is around 30 %. In Leskes s study, as mentioned by Andersson et al. (2005), from a survey data of , a linear relationship between age and presence of both mild and severe tinnitus was shown and additionally severity of tinnitus was found to be proportional to hearing impairment. 5

13 Gender relationship to the prevalence of tinnitus was shown in the study by Axelsson and Ringdahl (1981) who showed that tinnitus was more common in males than females. Men may have a higher risk of developing tinnitus, but the gender variation is not profound (6.6% for males vs. 5.6% in females in one study) (Tyler et al. 2008; Andersson et al. 2005).In men, the prevalence of this outcome rose steeply with age, from below 1% in those aged years to 8% in those aged The pattern was similar in women, but severe hearing loss was only about half as prevalent in the oldest age band. Whites experience tinnitus in greater numbers than blacks (9% in whites, 5.5% in blacks). Family members with household annual incomes less than $10,000 are affected with tinnitus at a rate of 12.8%, whereas those with an income over $35,000 exhibit a rate of only 7.6%. According to various studies on the localization or the site of Tinnitus, it can be said that bilateral cases are more common followed by unilateral in left ear and then unilateral in right ear (Tinnitus Formula). In some case tinnitus is localized in head while in others it is perceived as an external sound. Cause for an imbalance between left and right ear exists because overall somatic symptoms tend to be laterized to the left rather than to the right (Min and Lee). However Tyler said that unilateral tinnitus does not prove that tinnitus originated in that ear. An interesting finding by Erlandsson showed that multiple localizations of tinnitus were associated with more distress. 1.4 Problem statement The perceived tinnitus sound varies in terms of loudness and pitch from patient to patient (Henry and Meikle 2000; Lokenberg; Davis et al.) and so does its severity, which is not always proportional to the physical parameters of the tinnitus sound (Island Hearing 6

14 Services 2001). Therefore the efficiency of the sound masking therapy completely depends on the patient. The main challenge for a doctor is to - judge the feasibility of the sound masking therapy for a particular tinnitus patient and then generate a customized masker signal. It is also important to measure the subjective experiences in order to recreate the tinnitus sound to be heard by the doctor and the patient s family, to provide a psychological help to the patient, and also document the patient's status prior to the start of any treatment for future reference (Lokenberg; Snall 2004). Hence individual profiling of tinnitus sound, as perceived by the patient is extremely essential to provide a structured and an accurate treatment. The proposed research work addresses this problem by the development of Tinnitus Analyzer software to conduct individual profiling of perceived tinnitus. The main challenge is to design a system such that it incorporates the structured nature of traditional clinical treatment and blend it with the modernity of computer software to generate an automated clinical treatment of tinnitus through auditory masking. 1.5 Contribution In this research, a software system has been developed to accurately recreate tinnitus acoustical parameters as perceived by a patient in order to improve the efficiency of masking therapy on that particular patient. This software helps in individual profiling of tinnitus and assists better in providing customized masker signal to the patients. The approach followed for the development of this tool can be better described as a 7

15 multidisciplinary approach. Compared to the previously developed automated clinical tools for tinnitus assessment, the developed system is more methodical and structured. Unlike other similar tools, this tool starts with the generation of the audiogram for the patient to better analyze their case. The development of an audiogram generator needs a strong understanding of the frequency response of the human ear. The second module of this tool is to recreate tinnitus sound. Digital signal processing has been used to recreate tinnitus sound. Study of the characterization of tinnitus sound reveals that it has both noise and tone component. IIR bandpass filter has been designed to generate colored white noise. Sine and distorted sine ( harmonics of sine) have been used for the tone component. Similar to other tools and traditional clinical methods, this module performs the pitch matching test by varying the center frequency of the signal. But in this case, front end GUI makes it possible to choose any intermediate integer value frequency within an interval instead of only a few values. This provides better and more choices for the patient and the accuracy of recreating the sound will be increased. Additionally this module also allows to mix and match the noise and tone components to increase the probability of accurate sound recreation. This tool has an additional module to generate masker signal, and this makes it different from it s counterparts. Through various cases it has been shown that water sounds prove to be good masker signals for tinnitus patients. Hence nature sounds like waterfall, ocean waves, rainfall have been placed as masker choices for patients. The tool gives the patient the flexibility to choose the type of masker signal and also vary its frequency and loudness. In this case, for pitch shifting, phase vocoder is used. At first the short time fourier transform is performed on the 8

16 signal. The length of the duration of the STFT frames are then varied proportional to the factor by which the pitch has to be shifted and then they are converted back to time domain. The signal is then re-sampled by the same factor by which the pitch has to be shifted. Once the chosen signal successfully masks the perceived tinnitus, the tool presents 2 more important tests residual inhibition and minimal masking level; in order to judge the effectiveness of the masker signal. The main novelty of the proposed software lies in the fact that the results from the 4 tests pitch matching, loudness matching, residual inhibition and minimal masking level, are used for judging if masking therapy is feasible for a particular patient. It is important to understand that though this software generates masker signal similar to the tinnitus maskers available online, yet the contribution of the developed software unlike the online maskers, is that it provides a structured clinical treatment which the online maskers lack, decreasing their validity. The tool has a friendly user interface making it easy to use and understand. 1.6 Thesis Organization The Master s thesis is organized in the following way: Section 2 discusses the background of tinnitus which includes the anatomy of ear, causes and mechanisms of tinnitus, literature survey on the prevalent treatments and characterization of tinnitus sound. Section 3 involves the description of the system architecture which includes the overview of the proposed scheme and the algorithm of the developed software tool, 9

17 Section 4 describes an applied experimentation test case used for the validation of the proposed software. Lastly the conclusion and future work are presented in section 5. 10

18 CHAPTER 2 BACKGROUND 2.1 Hearing mechanism and anatomy of ear 2.11 Hearing Mechanism The ears are paired sensory organs comprising the auditory system, involved in the detection of sound, and the vestibular system, involved with maintaining body balance/ equilibrium. The ear can be divided anatomically and functionally into three regions: the external ear, the middle ear, and the inner ear. All three regions are involved in hearing. Only the inner ear functions in the vestibular system. - Outer ear - Middle ear - Inner ear Sound waves enter the ear from external source and travel through the auditory canal which set up vibrations in the eardrum. The vibrations of the eardrum cause the bones in 11

19 the middle ear to move back and forth like tiny levers. This lever action converts the large motions of the eardrum into the shorter, more forceful motions of the stapes. The footplate at the inner end of the stapes moves in and out of the oval window at the same rate that the eardrum is vibrating. The movement of the footplate sets up motions in the fluid that fills the cochlea. The movement of the fluid causes the hair immersed in the fluid to move. The movement stimulates the attached cell to send a tiny impulse along the fibers of the auditory nerve to the brain. In the brain the impulse is translated into the sensation which is perceived as sound Anatomy of ear (Bernsee 1999; America Speech Language Hearing association ;Anatomy of inner ear;patts 2001;Oghalai 2006) A) Outer ear: The main role of the outer ear is to collect the sound waves from the external air and reflect it towards the ear canal or further processing. Different sound frequency components are reflected in different ways, so even though they originate with the same intensity their relative amplitude changes after reflection. The external/outer ear consists of the expanded portion named the auricula or pinna, and the external acoustic meatus. The former projects from the side of the head and serves to collect the vibrations of the air by which sound is produced; the latter leads inward from the bottom of the auricula and conducts the vibrations to the tympanic cavity/membrane. The external ear has no bones. The configuration of the external ear is such that it allows to selectively boost the 12

20 sound pressure 30 to 100 fold for frequencies around 3 khz. Fig 2.1 shows the anatomy of the outer ear. Pinna, or auricle The visible part is called the pinna and functions to collect and focus sound waves. It is of an ovoid form, with its larger end directed upwards. Its lateral surface is irregularly concave, directed slightly forward, and presents numerous eminences and depressions. It is composed of a thin plate of yellow elastic cartilage, covered with integument, and connected to the surrounding parts by ligaments and muscles; and to the commencement of the external acoustic meatus by fibrous tissue. Ear canal, or external auditory meatus From the pinna the sound pressure waves move into the ear canal, a simple tube of length 2.5 cm (when measured from the bottom of concha) running through the middle ear. This tube leads inward from the bottom of the auricula and conducts the vibrations to the tympanic cavity and amplifies frequencies in the range 3 khz to 12 khz. 13

21 Fig 2.1 : Anatomy of outer ear B) Middle ear The middle ear (ME) and contents serve to transmit external sound energy to the components of the inner ear. It is an irregular, laterally compressed space within the temporal bone. It is filled with air, which is conveyed to it from the nasal part of the pharynx through the auditory/eustachian tube, which is also used to maintain the pressure balance. It contains a chain of movable bones -ossicles, which connect its lateral to its medial wall, and these bones are responsible to convey the vibrations communicated to the tympanic membrane across the cavity to the internal ear through oval window. Thus the middle ear acts as a mechanical lever and serves to magnify the sound intensity at the entrance to the cochlea. The inner ear is where motion is translated to neural signals. Fig 2.2 shows the anatomy of the middle ear. 14

22 Ossicles The middle ear contains three tiny bones known as the ossicles malleus (hammer), incus (anvil) and stapes (stirrup). The ossicles directly couple sound energy from the ear drum to the oval window of the cochlea. The ossicles are classically supposed to Fig 2.2 : Anatomy of the middle ear mechanically convert the vibrations of the eardrum, into amplified pressure waves in the fluid of the cochlea with a lever arm factor of 1.3. Since the area of the eardrum is about 17 times larger than that of the oval window, the sound pressure is concentrated, leading to a pressure gain of at least 22. It is the vibration of the stapes which is finally passed as pressure waves in the inner ear. The middle ear efficiency peaks at a frequency of around 15

23 1 khz. The combined transfer function of the outer ear and middle ear gives humans a peak sensitivity to frequencies between 1 khz and 3 khz. C) Inner ear It is the innermost part of the ear and is responsible for auditory transduction by which the mechanical vibrations which cause sound are converted into electromechanical energy. The inner ear contains the sensory systems of balance and hearing. The main function of the inner ear is to inform the brain the amount of energy in the collected environmental sound and the frequency components of the sound. The inner ear is divided into 2 fluid filled chambers one inside the other the perilymph and endolymph. The difference in the chemical composition of these fluid compositions is maintained by specialized cells. The generated potential difference provides chemical energy to the sensory cells (hair cells). The inner ear consists of membranous labyrinth encased in osseous labyrinth. The bony labyrinth is lined with the membranous labyrinth. There is a layer of perilymph between them. The three parts of the bony labyrinth are the vestibule of the ear, the semicircular canals, and the cochlea. The vestibule and semicircular canals are concerned with vestibular function (balance); the cochlea is concerned with hearing. The cochlea is a coiled tube and the oval window opens into the vestibule, at the base of the cochlea. Fig 2.3 shows the anatomy of the inner ear. 16

24 Fig 2.3: Anatomy of inner ear Cochlea The cochlea is the auditory portion of the inner ear. Its core component is the Organ of Corti, the sensory organ of hearing, which is distributed along the partition separating fluid chambers in the coiled tapered tube of the cochlea. The cochlea consists of two and three quarter turns (approximately 30 mm in length). The fluid-filled spaces of the cochlea are comprised of three parallel canals: an outer scala 17

25 vestibuli (ascending spiral), an inner scala tympani (descending spiral), and the central cochlear duct (scala media). The cochlear duct is separated from the scala vestibuli by the vestibular (Reissner s) membrane and from the scala tympani by the basilar membrane. The organ of Corti resides within the cochlear duct on the basilar membrane. Movement of the stapes results in transmission of fluid waves into the scala vestibuli via the cochlear recess, which lies on the medial wall of the vestibule. As these sound waves enter the perilymph of the scala vestibuli, they are transmitted through the vestibular membrane into the endolymph of the cochlear duct, causing displacement of the basilar membrane, which stimulates the hair cell receptors of the organ of Corti. It is the movement of hair cells that generates the electric potentials that are converted into action potentials in the auditory nerve fibers. The basilar membrane varies in width and tension from base to apex. As a result, different portions of the membrane respond to different auditory frequencies. These perilymphatic waves are transmitted via the apex of the cochlea (helicotrema) to the scala tympani and eventually dissipated at the round window. The flexible nature of the round window diaphragm is necessary for fluid propagation. Hair Cells Coiling around the inside of the cochlea, the organ of Corti contains the cells responsible for hearing, the hair cells and also support cells. They are called mechanotransducers as they convert mechanical stimuli into neural information which is transmitted to the brain. They have synapses which provide communication between neuronal cells. One side of synapse is called presynaptic and the other side is called postsynaptic. Inside the 18

26 cochlea, sound waves cause the basilar membrane to vibrate up and down. This creates a shearing force between the basilar membrane and the tectorial membrane, causing the hair cell stereocilia to bend back and forth. A chemical known as neurotransmitter is released from the presynaptic cell which changes the membrane potential o the postsynaptic cell. This release is proportional to the change in the membrane potential of the hair cell due to the bending of the stereocilia bundle. There are 2 kinds of synapses afferent and efferent. The afferent synapses convey information into the central nervous system by exciting action potentials in the afferent nerve fibers that enter the brain. Efferent synapses modulate the membrane potential of the hair cell on the release of neurotransmitter from presynaptic element (terminal of the nerve fiber which arises deep in the brainstem). The neural signals innervated into hair cells through efferent fiber changes the sensitivity of the hair cells. Each hair cell codes the direction and the degree of stereocilia bundle bending by either increasing or decreasing the firing rate of postsynaptic afferent fiber in proportion to the magnitude of the bend. The tectorial membrane is loosely coupled to the reticular lamina. There are 4 rows of hair cells, one on the inner (modiolar) side of the tunnel formed by the pillar cells-- these are the inner hair cells; and 3 one the outer side of the Tunnel of Corti, these are the outer hair cells. The Deiter s cells support the Outer hair cells at their base, but the outer hair cell walls are surrounded by fluid. The inner hair cell is surrounded by the support cells. The reticular lamina is a solid surface at the tops of the hair cells, so the tops of the hair cells are in endolymph and the bottom of the hair cells are in perilymph. Deiter s cell 19

27 processes fill in the gaps between the tops of the outer hair cells to form the reticular lamina. The endolymph is a potassium-rich extracellular fluid whose standing potential is about 80 mv more positive than perilymph, a fluid similar, and continuous with, cerebrospinal fluid that fills the spiral canal and surrounds the basolateral membrane of the hair cells. The intracellular potential of inner hair cells, referred as perilymph, is about -45 mv, whereas that of the outer hair cells is about -70 mv. Thus, the driving force for current is a potential difference comprised between 125 mv and 150 mv. 20

28 Fig 2.4 : Hair Cells (a) Transverse section through a middle turn of the cochlea, showing the organ of Corti, an assembly of intricately shaped supporting cells and inner and outer hair cells supported by a flexible basilar membrane. (b) Upward displacement of the basilar membrane stimulates the hair cells by bending their stereociliary bundles against the acellular tectorial membrane. Because of the point about which the basilar membrane hinges, the inner hair cells must be stimulated mainly by motion of the tectorial membrane. Signals from each inner hair cell are relayed to the brain via 10 to 20 afferent fibres 21

29 Inner hair Cells (IHC) Studies reveal that it is only the inner hair cells that carry the acoustic information to the brain. The deflection of the hair-cell stereocilia opens mechanically gated ion channels that allow any small, positively charged ions (primarily potassium and calcium) to enter the cell. Unlike many other electrically active cells, the hair cell itself does not fire an action potential. Instead, the influx of positive ions from the endolymph in Scala media depolarizes the cell, resulting in a receptor potential. This receptor potential opens voltage gated calcium channels; calcium ions then enter the cell and trigger the release of neurotransmitters at the basal end of the cell. The neurotransmitters diffuse across the narrow space between the hair cell and a nerve terminal, where they then bind to receptors and thus trigger action potentials in the nerve. In this way, the mechanical sound signal is converted into an electrical nerve signal. The repolarization in the hair cell is done in a special manner. The perilymph in Scala tympani has a very low concentration of positive ions. The electrochemical gradient makes the positive ions flow through channels to the perilymph. Outer hair cell (OHC) The role of outer hair cells in hearing is both sensory and mechanical. They are shaped cylindrically, like a can, and have stereocilia at the top of the cell, and a nucleus at the bottom. When the organ of corti begin to vibrate in response to the incoming sound, each OHC will sense the vibration through the bending of its stereocilia. This bending causes a change in the OHC s internal electrical potential, which drives the electromotility. If the 22

30 resulting mechanical force is at the natural frequency of that portion of the cochlea then the magnitude of the vibration will increase else it will decrease. The OHCs thus increase the frequency selectivity and sensitivity of hearing. The reined mechanical vibrations of the organ of corti are transmitted to IHCs. 2.2 Causes and Mechanisms of Tinnitus Tinnitus does not represent a disease itself but instead is a symptom of a variety of underlying diseases (Han et al. 2009, Henry et al 2005). Medical investigations have shown that there are many pathophysiological causes for the occurrence of tinnitus. Otologic causes include noise-induced hearing loss, presbycusis, otosclerosis, otitis, impacted cerumen, sudden deafness, Meniere s disease, and other causes of hearing loss. Neurologic causes include head injury, whiplash, multiple sclerosis, vestibular schwannoma (commonly called an acoustic neuroma), and other cerebellopontine-angle tumors. Infectious causes include otitis media and sequelae of Lyme disease, meningitis, syphilis, and other infectious or inflammatory processes that affect hearing. Tinnitus is also a side effect of some oral medications, such as salicylates, nonsteroidal antiinflammatory drugs, aminoglycoside antibiotics, loop diuretics, and chemotherapy agents (e.g., platins and vincristine). Temporomandibular-joint dysfunction and other dental disorders can also cause tinnitus. However, in many cases no underlying physical cause is identifiable (Andersson et al.). For many years, hearing loss has been understood to be the most common cause of tinnitus. There have been many cases that prove that the tinnitus activity can occur at an anatomical site away from the initial pathology. 23

31 Therefore the knowledge of the pathological event that triggered tinnitus is limited in its scope to accurately detect the root cause of tinnitus (Andersson et al. 2005) Nevertheless it is still essential to examine the various pathological triggering factors as it may provide the path for methodical assessment. Most often, two or more triggering factors act together to produce symptomatic tinnitus. Hence Tinnitus is a symptom for diverse pathologies and interestingly all the levels of nervous system are responsible, through various degrees, for the perception of tinnitus (Han et al. 2009; Lanting et al. 2009; Adjamian et al. 2009; Eggermont and Roberts 2004; Nelson 2003). Peripheral auditory system The peripheral auditory system consists of outer, middle and inner ear, explained in details in section 2.1. The mechanisms involved with the peripheral auditory system which can cause tinnitus are explained below: Spontaneous otoacoustic emissions In mammalian ear, active mechanical processes contribute to cochlear micromechanics. The mechanical expansion and contraction of cochlear outer hair cells contribute to this mechanics. This mechanics lead to some acoustic emissions. Hence in addition to just being the receiver of sound, the cochlea also produces low-intensity sounds called Otoacoustic emissions (AOEs). The presence of cochlear emissions was hypothesized in the 1940s on the basis of mathematical models of cochlear nonlinearity. However, OAEs could not be measured until the late 1970s, when technology created the extremely sensitive low-noise microphones needed to record these responses. 24

32 The 4 types of otoacoustic emissions are as follows: Spontaneous otoacoustic emissions (SOAEs) - Sounds emitted without an acoustic stimulus (ie, spontaneously) Transient otoacoustic emissions (TOAEs) or transient evoked otoacoustic emissions (TEOAEs) - Sounds emitted in response to an acoustic stimuli of very short duration; usually clicks but can be tone-bursts Distortion product otoacoustic emissions (DPOAEs) - Sounds emitted in response to 2 simultaneous tones of different frequencies Sustained-frequency otoacoustic emissions (SFOAEs) - Sounds emitted in response to a continuous tone Spontaneous otoacoustic emissions (SOAEs), first discovered by Kemp, are small acoustic signals presumed to be generated by the electromotile activity of the OHCs of the cochlea and propagated into the external auditory canal. In various studies it has been shown that SOAEs produced by the cochlea can be perceived as tinnitus. SOAEs are usually inaudible, but they can become audible due to instability. An additional criterion favoring a common patho-mechanism of SOAEs and tinnitus resulted from masking effects. Tinnitus and SOAEs from both sides were suppressed by the same tone, which indicates that the involved pathways are identical (Plinkert et al. 1990). These atypical SOAEs are much more prevalent in the higher frequency range and can appear at sound pressure levels up to 55 db SPL in the ear canal. Tinnitus due to SOAEs is mild and is more common in subjects with normal hearing and in those with only middle ear disorders. SOAEs decrease as hearing loss progresses, and hence these otoacoustic 25

33 emissions are not likely to cause tinnitus when a hearing loss of 35 db or more is present. However, SOAEs cannot fully explain the mechanism of tinnitus since in many cases it has been observed that aspirin largely abolishes SOAEs without improving tinnitus (Campbell 2009). In general, SOAEs occur in only 40-50% of individuals who have normal hearing. SOAEs generally are not found in individuals with hearing thresholds more than 30 db HL. Therefore, the presence of SOAEs usually is considered a sign of cochlear health, though it s absence is not necessarily a sign of abnormality. When present in humans, SOAEs usually occur in the 1000 to 2000Hz region; amplitudes are between -5 and 15 db SPL. Some individuals have multi frequency SOAEs over a broader frequency range. SOAEs typically are bilateral rather than unilateral. If unilateral, they are more likely to be present in the right rather than in the left ear. SOAEs occur more often in females than in males (across all ages). In high-level SOAEs the emissions can be heard by others and are more common in children than in adults. Edge theory Edge theory, also known as contrast theory, proposes that tinnitus is induced by increased spontaneous activity in the edge area, which represents a transition from OHCs in the organ of Corti with relatively normal morphology and function on the apical (i.e., lowfrequency) side of a lesion to OHCs toward the basal side that are missing or have a pathologic appearance and poor functionality. Edge theory can be explained by discordant theory. 26

34 Discordant theory Discordant dysfunction theory postulates that tinnitus is induced by the discordant dysfunction of damaged OHCs and intact inner hair cells (IHCs) of the organ of Corti located in the inner ear. Intense noise and ototoxic agents initially damage OHCs in the basal turn of the cochlea, and subsequently, if continued or repeated, affect IHCs this is due to IHCs being more resistant to such damage. IHCs are the receptor cells for sound transduction, and almost all afferent fibers in the auditory nerve (95%) innervate IHCs. In contrast, OHCs work as mechanical amplifiers, enhancing weak sounds by providing up to 50 db, which can be evaluated by measuring otoacoustic emissions. In almost all situations OHCs are damaged more than IHCs, which results in the uninhibition of neurons in the dorsal cochlear nuclei (DCNs). Spontaneous activity is increased when neurons in the DCN receive excitation from IHCs but not from the damaged OHCs, and this is perceived as tinnitus. Normally there is a small gap between the top of the cilia of the IHCs and the bottom of the tectorial membrane, but in the area in which OHCs are affected but IHCs are intact, the tectorial membrane might touch the IHC cilia, thus causing the IHCs to depolarize. The OHCs normally recover with a few days, but this can be delayed for up to a few months. Therefore, it is hypothesized that tinnitus represents a consequence of a central gain adaptation mechanism when the auditory system is confronted with a hearing loss. Discordant theory explains why many individuals with tinnitus have normal hearing if there is only partial damage to OHCs, since up to 30% of OHCs can be damaged without inducing hearing loss. OHCs die at a rate of approximately 0.5% per year beginning during the first years of life, and OHC-induced hearing loss does not usually appear before the end of the fifth decade of life. 27

35 Discordance is absent in totally deaf individuals with complete damage to both OHCs and IHCs, and hence tinnitus is not induced. If there is increased gain within the Central Nervous System, tinnitus is present even in deaf subjects. Similarly, noise-induced tinnitus is caused by discordant damage between OHCs and IHCs. Two types of noiseinduced tinnitus have been identified: tonal and complex. Tonal tinnitus results from discordant dysfunction of OHCs and IHCs manifesting in a single area, whereas complex tinnitus results from multiple areas of discordance. When patients clearly have the central type of tinnitus, such as after transaction of the auditory nerve, the OHC concept is not applicable and alternative mechanisms need to be considered. Role of Calcium and N- methyl-d aspartate (NMDA) receptors: Calcium plays a vital role in the normal functioning of the cochlea, as discussed in section Literature shows that abnormal calcium handling can lead to malfunctioning of cochlea. Jasterboff (1990) suggested the hypothesis that calcium concentrations might directly affect neurotransmitter release. Raised levels increase the auditory nerve activity in response to cochlear mechanical mechanisms while reducing spontaneous activity. Reduction of calcium concentration conversely causes reduced evoked activity and increases spontaneous activity and thus increasing the possibility of tinnitus. There have been studies to show that tinnitus occurs also due to stress. Dynorphins, endogeneous peptides are released into the synaptic regions between inner hair cells and Type I auditory fibers by lateral efferent neurons in response to stressful situations. Glutamate is an excitatory neurotransmitter released by inner hair cells in response to mechanical stimulations, and it interacts with NMDA receptors. It was suggested by 28

36 Sahley and Nodar that dynorphins increases the effect of glutamate on NMDA receptors and concluded that chronic exposure could result in abnormal excitement in the auditory nerve which can lead to tinnitus. Central auditory system Primary auditory fibers enter the brain stem and immediately make connections with secondary neurons in the cochlear nucleus. Thereon, the auditory information goes to a number of places in the central nervous system, and this system overall is referred to as central auditory system. Central auditory processing can be described as "what we do with what we hear." In other words, it is the ability of the brain (i.e., the central nervous system) to process incoming auditory signals. Central auditory processes are the auditory system mechanisms and processes responsible for the following behavioral phenomena. 1. Sound localization and lateralization 2. Auditory discrimination 3. Temporal aspects of audition including: temporal resolution, temporal masking, temporal integration and temporal ordering. 4. Auditory performance with competing acoustic signals 5. Auditory performance with degraded signals It is now widely assumed that peripheral damage triggers changes in the central auditory system, which are then interpreted as tinnitus by the higher processing stages in the brain (Jastreboff, 1990). 29

37 The dorsal cochlear nucleus The DCN has been implicated as a possible site for the generation of tinnitus-related signals owing to its tendency to become hyperactive following exposure to tinnitusinducing agents such as intense sound and cisplatin. OHC damage triggers plastic readjustments in the DCN, resulting in DCN hyperactivity. It is hypothesized that a reduction in auditory nerve input leads to uninhibition of the DCN and an increase in spontaneous activity in the central auditory system, which manifests as tinnitus. This mechanism could explain the temporary ringing sensation that can follow exposure to loud sound. The plastic readjustments in the DCN are slow and lead to tinnitus with a delayed onset. IHC damage prevents hyperactivity in the DCN. Auditory plasticity theory Plasticity can be defined as the modification of nerve cells to adapt better to immediate environmental influences; this alteration is often associated with behavioral change. Auditory plasticity (AP), refers to plasticity in the auditory system. AP plays a vital role in auditory training. The peripheral auditory system does not have plasticity but central auditory system has, since the brain has plasticity. According to auditory plasticity theory, damage to the cochlea enhances neural activity in the central auditory pathway. Auditory plasticity emerges as a consequence of the aberrant pathway (Museik). Tinnitus might be generated in the temporal lobe in the auditory association cortex and inferior colliculus. The ability of some individuals to modulate tinnitus by performing voluntary somatosensory or motor actions can be 30

38 attributed to plastic changes involving the development of aberrant connections between the auditory and sensory-motor systems in the brains of these patients. Crosstalk theory According to crosstalk theory, when auditory nerve fibers are intact and some other cranial nerves are damaged, artificial synapses (called crosstalk) can develop between individual auditory nerve fibers, resulting in the phase-locking of the spontaneous activity of groups of auditory neurons. In the absence of external sounds, this creates a neural pattern that resembles patterns evoked by actual sounds. These cranial nerves are sensitive to compression at the root entry zone, where they are covered by myelin. Nerve compression causes crosstalk between nerve fibers, and the breakdown of the myelin insulation of the nerve fibers establishes ephaptic coupling between them. This notion is applied to the cochlear vestibular nerve, which is covered by central myelin for most of its length and hence is vulnerable to compression from blood vessels or tumors impinging upon the nerve (e.g., vestibular schwannoma). Such compression and consequent ephaptic coupling might lead to tinnitus if synchronization of the stochastic firing in the cochlear nerve is perceived as sound. Somatosensory system The activity of the DCN is also influenced by stimulation of nonauditory structures; however, the somatosensory system is the only nonauditory sensory system that appears to be related to tinnitus (e.g., in temporomandibular-joint syndrome and whiplash). It is further hypothesized that somatic tinnitus is due to central crosstalk within the brain, 31

39 because certain head and neck nerves enter the brain near regions known to be involved in hearing. Limbic and autonomic nervous systems It has been observed that if the first perception of tinnitus induces high levels of annoyance or anxiety by association with unpleasant stimuli or with periods of stress and anxiety, tinnitus might lead to high levels of annoyance or anxiety. At the unconscious level, tinnitus can increase progressively without the patient being aware, resulting in enhanced activity in the limbic and autonomic nervous systems. In such situations, tinnitus emerges as a clinically significant problem. Toxic substances Salicylate, aspirin, antibiotics, cisplatin, quinine, furosemide, hydroxychloroquine, ethacrynic acid, bumetanide, amphotericin B, heavy metals such as mercury, antidepressants such as Wellbutrin (Zyban), and possibly caffeine can cause tinnitus. The inclusion of caffeine in this list is controversial. Otosclerosis Otosclerosis is the abnormal growth of bone of the middle ear. This bone prevents structures within the ear from working properly and causes hearing loss. For some people with otosclerosis, the hearing loss may become severe. Otosclerosis is one of the causes of Tinnitus, some of the supporting theories are: - Conductive hearing losses reduce the masking effect of environmental sound 32

40 - The new grown bone has a rich blood supply causing pulsatile tinnitus - The abnormal connections between arteries and veins result in pulsatile tinnitus - Toxic enzymes produced by the otosclerotic bone, damage the cochlea and cochlear blood supply which causes cochlear tinnitus Dental problems Dental problems are another frequent cause of tinnitus. Tooth abscesses or impacted wisdom teeth can cause tinnitus. In such cases, further dental work will often cure the problem. Injury of the nerves during extraction of a wisdom tooth has also been known to cause tinnitus. Muscle spasm Muscle spasm in head or neck is probably the most common cause of tinnitus, accounting for 80% of patients. When the noise is made better or worse by changes in bodily posture, or arm or neck movements, the patient has "somatic tinnitus." Somatic tinnitus is usually unilateral. In its earliest stages, it may be caused by hearing trauma, an injury, or a muscle contraction (such as by grinding one's teeth) that compresses some part of the auditory system. Later, cross-talk occurs between the signals the muscles send to the brain and the signals from the ear. 2.3 Prevalent treatments A literature survey The available treatments for the management of tinnitus are several and diverse. Tinnitus is only a symptom that might be the manifestation of different underlying pathologies and 33

41 has several etiologies. This heterogeneity within tinnitus patients leads to the unavailability of any standardized protocol for treating tinnitus. Broadly, tinnitus treatments can be divided into two categories (Han et al ) those aimed at directly reducing the intensity of tinnitus and 2) those aimed at reducing the severity associated with tinnitus as perceived by a particular patient. The former include pharmacotherapy and electrical suppression, and the latter include pharmacotherapy, cognitive and behavioral therapy, sound therapy, habituation therapy, massage and stretching, and hearing aids. It is important to distinguish between the severity and intensity of the perceived tinnitus and understand that severity does not depend on the intensity of perceived tinnitus sound. The former depends on the perception of tinnitus sound where as the later depends on the tinnitus sound characteristic. In spite of the huge number of options available for treating tinnitus, the need for complete solution for tinnitus management still remains unmet. This section deals with the literature review of the tinnitus treatments prevalent in the past and present. Pharmacotherapy Though there is a huge need of drugs for tinnitus treatments, yet not a single FDAapproved drug is available in the market (Han et al. 2009; Henry et al. 2005; Elgoyhen and Langguth 2009). Extensive reviews of randomized clinical trials have revealed that only nortriptyline, amitriptyline, alprazolam (Vernon and Meikle 2003), clonazepam, acamprosate, and oxazepam are more beneficial than placebo while most of the drugs used for drug therapy have a transient effect and on discontinuing the drugs tinnitus recurred to its prior level or even worse level (Han et al. 2009). The only medication 34

42 providing a reliable reduction of tinnitus is intravenous lidocaine, and there is a close association between the effects of lidocaine and oral carbamazepine. Unfortunately, lidocaine cannot be used clinically because it must be injected, its effects are of short duration, and it frequently produces adverse side effects (Elgoyhen and Langguth 2009). Tinnitus due to SOAEs can be diminished by aspirin. Antidepressants are commonly used in pharmacological protocols for the management of tinnitus. The analgesic effects of tricyclic group (amitriptyline, trimipramine and nortriptyline, nortriptyline) make it one of the commonly used antidepressant (Robinson et al.). Severe tinnitus can be an extremely stressful condition, heavily influencing every aspect of the patient s life. Thus, anxiolytics such as benzodiazepines have been used extensively to help patients cope with their tinnitus. Anticonvulsants have also been used in tinnitus patients. The lack of complete understanding of underlying pathology of tinnitus limits the scope of formulating the most appropriate drugs for tinnitus treatment. Cognitive and behavioral therapy Due to the diversity in the reasons behind the occurrence of tinnitus, most clinicians agree that tinnitus needs to be managed within a multidisciplinary setting. It has been observed that counseling is probably more helpful than medical treatments in treating tinnitus. Several studies have confirmed an association between psychological factors, such as anxiety, depression etc, and severity of tinnitus as experienced by the patient 35

43 (Andersson). With the 2 popular models - neurophysiological and psychological (Andersson et al.), in mind, tinnitus patients have been thus successfully targeted with cognitive behavioral therapy (CBT). The potential of CBT in tinnitus management has been reviewed in (Andersson). Cognitive therapy focuses on how one thinks about tinnitus and on the avoidance of negative ideation, whereas behavioral therapy uses the systematic desensitization approach that is applied to many phobias (Han et al 2009). Cognitive therapy involves teaching patients to cope with their tinnitus by replacing negative thinking with more positive thinking. Cognitive therapy includes counseling and cognitive restructuring. Counseling should include 1) informing patients that it is unlikely that their annoyance with tinnitus will improve dramatically, 2) informing patients about the usefulness of tinnitus self-help groups, 3) helping patients to minimize the time devoted to activities and/or conditions in which the tinnitus intensity is increased and to maximize the time devoted to activities and/or conditions in which the tinnitus intensity is decreased, and 4) stressing the avoidance of noise exposure, since noise-induced hearing loss and tinnitus are related. Cognitive restructuring involves changing thoughts associated with tinnitus and retraining the brain in order to not pay much attention to the tinnitus sound. Behavioral therapy focuses on positive imagery, attention control, and relaxation training. Positive imagery involves focusing thoughts on something pleasant, thereby diverting thoughts from tinnitus. Patients begin with pleasant visual images and auditory images. Attention control involves moving attention away from the tinnitus when it is bothersome. Relaxation training uses a guided protocol to teach participants to apply progressive muscle relaxation, involving tensing and relaxing the arms, face, neck, shoulders, abdomen, legs, and feet. 36

44 Sound therapy The main goal of sound therapy is to reprogram the patient s brain to pay attention to different sounds and retrain their ears to different sounds. They try to stretch the patient s range in registering sound and tuning out the constant tinnitus sound that the patient might be perceiving (Jasterboff 2007). Sound therapy uses continuous neutral sounds (which is - not annoying, nonintrusive, not meaningful and doesnot attracts attention) found in natural settings, including those associated with streams, rain, waterfalls, and wind in order to decrease the strength of the tinnitus-related neuronal activity within the auditory system by assisting in the rehabilitation of the ears as well as helping to stimulate the brain (Andersson; Jasterboff 2007). The actual Sound Therapy then stimulates the whole ear by sending it constantly alternating sounds that are high and low in their tone and the result is that this exercises the middle ear muscles which then stimulate the various receptor cells in the inner ear. As the ear becomes more receptive to high frequency sounds they are then passed directly onwards towards the brain. The sound intensity should be at or below the level at which the patient can perceive the tinnitus and the external sound separately. The sound must be applied bilaterally to avoid asymmetrical stimulation of the auditory system, since stimulating only one side in unilateral tinnitus frequently results in a shift of the perceived location of the tinnitus to the opposite side due to strong interactions within the auditory pathways. This process is like recharging the brain using sound which removes tiredness and the need for excessive sleep thereby creating a more balanced state of peace and relaxation. 37

45 Applying sound therapy during the night can be helpful for individuals without sleep problems because the auditory pathways are fully active up to the level of the inferior colliculi during sleep (Andersson). Sound Therapy self help programs can be done with just a CD and player or even a cassette tape and portable tape player and hence is a very easy to undertake. Occlusion with ear plugs should be minimized by using open-ear molds to allow normal access of environmental sounds to the ear. (Jasterboff 2007; Hesser et al. 2009). New methods of treatment using auditory and visual attention training are proposed as a means to augment counseling and sound therapies for tinnitus management. Auditory discrimination training (ADT), a procedure to increase cortical areas responding to trained frequencies and reducing the response to tinnitus frequencies, has been recently used as a sound therapy (Herraiz et al. 2007) which improved tinnitus in 40 % of the 27 patients used for the study. (Searchfield et al. 2007) shows the effectiveness of structured Auditory Object Identification and Localization (AOIL) tasks to train patients to ignore tinnitus. Using such a type of attention training assists the improvement of an individuals ability to attend to relevant sounds while ignoring distracters. Masking therapy for tinnitus The masking therapy uses the fact that a sound can be masked by another sound signal at a nearby frequency but of a higher intensity. Similar theory can be applied to suppress perceived tinnitus sound (Hazell and Wood 1981). The first use of Tinnitus masking dates back to the early 20 th century. But Dr. Vernon (1973) is known as the father of the modern tinnitus masking. The simplest tinnitus masker is white noise, however this is an 38

46 abrasive sound not pleasant to the ear for long term (Mitchelle 1983; Vernon and Meikle 2003). Though masking therapy is a non-invasive and an easy way to deal with tinnitus, yet it does fail for some patients. The usefulness of masking therapy depends on the characteristics of the perceived tinnitus sound pitch, loudness, minimal masking level and residual inhibition (Hazell and Wood 1981; Vernon et al. 1990; Jasterboff et al. 1994; Roberts 2007). Thus quantification of the tinnitus sensation is mandatory. Once the correct pitch and loudness of the tinnitus sound is known and the psychoacoustic measures (Henry and Meikle 2000) of tinnitus for a particular patient allow masking, then a masker signal can be generated. This masker signal is presented to the ears of tinnitus patients for a specific time interval and after its removal the tinnitus sound disappears for sometime. Residual inhibition, which is the suppression of tinnitus for a period of time after the removal of the masker signal, is helpful in reducing tinnitus by interrupting abnormal synchronous activity among networks of neurons that generate tinnitus (Henry and Meikle 2000; Roberts 2007; Lokenberg). However (Zhang et al.) has shown that variations in individual physical parameters of tinnitus sound donot have significant role in determining the acceptance of tinnitus. The paper (Vernon et al. 1990) proposes the use of masking indicator, obtained by subtracting the loudness match of the tinnitus from the minimum masking level, to determine acceptance of tinnitus or a particular patient. In (Jasterboff et al. 1994), the authors have presented the influence of masking therapy on minimal masking level. They showed that minimal masking level decreased in 73.9% of the people whose tinnitus improved using masking. It has been demonstrated in (Hazell and Wood 1981) that patient management/ counseling is essential for successful tinnitus masking. Authors of (Zhang et al.) show that masking therapy might also work on 39

47 patients with cochlear implants by deliberate introduction of noise into cochlear implant. Findings from an interesting study (Roberts 2007) suggest that tinnitus and its suppression in residual inhibition depend on processes that span the region of hearing impairment and not on mechanisms that enhance cortical representations for sound frequencies at the audiometric edge. There are broadly 3 types of devices that can be used for masking (Han et al. 2009) hearing aids (explained below), tinnitus maskers (with user-adjustable frequency emphasis) and tinnitus instruments/combination devices (high frequency hearing aid and a tinnitus masker). In order to make the most effective use of the hearing aids, adaptation to the hearing aids is necessary through counseling and customization (Bo and Ambrosetti 2007). Since perceived tinnitus sounds are generally high pitched and high frequency hearing losses are most common so majority of the tinnitus maskers give emphasis to the high frequency signals (Bo and Ambrosetti 2007; Kemp and George 1992). UltraQuiet device is such a tinnitus masker which produces high frequency bone conduction masking therapy (Barbara et al. 2005, Madsen and Guinta 2002). The masking stimulation, if properly chosen, may have the capacity of changing how the brain processes tinnitus, which could provide long-lasting benefit (Jasterboff et al. 1994). A novel concept has been introduced by the authors (Winkler 2009) to manipulate the spectrum of masking signal digitally, which proved to provide therapeutic benefit in tinnitus patients by prolonging residual inhibition. A series of digitally synthesized low frequency complex acoustic waveforms (TIPA) have been used to achieve the same. In any kind of tinnitus masking technique, it is important to keep in mind that the masker signal should be at an intensity level which will not be perceived by the patient as too 40

48 loud. If this is the case then the patient is not a good candidate for masking therapy (Henry and Meikle 2000) Hearing aids Tinnitus is often associated with hearing loss, because due to hearing loss the external sounds do not reach the ear and so the tinnitus perception becomes more prevalent. Hearing aids is one of the forms of sound therapy that is specifically beneficial to tinnitus patients with significant hearing loss. Hearing aids are designed in a way to amplify the external sounds (speech and other ambient sounds). Amplification of speech diverts attention away from tinnitus, and amplification of other ambient sounds serves to partially mask tinnitus (Hesser 2009). Hearing aids can prove to have a long term beneficial effect on tinnitus patients since they use the neural plasticity characteristic of the brain to reprogram it to reduce the neural activity, responsible for tinnitus perception (Bo and Ambrosetti 2007;Lanting et al. 2009; Eggermont and Roberts 2004; Jsterboff 1990; Tyler). Hearing aids are not appropriate for those with hearing loss confined to above 6 khz, because most hearing aids have limited high-frequency amplification abilities. Many hearing-impaired patients have normal or near-normal hearing at low frequencies, and common environmental sounds contain a significant amount of energy below 200 Hz, which provides constant sound stimulation and thus helps to prevent difficulties associated with increased gain in the auditory system. The best possible use of hearing aids can be achieved when fitted in both the ears with open ear aid and with wide band amplification (Bo and Ambrosetti 2007; Han et al. 2009). 41

49 Music therapy Music therapy is a desensitization method that utilizes music that has been spectrally modified according to the hearing characteristics of each patient to allow the masking of tinnitus and to facilitate relaxation at a comfortable listening level. Music directly affects the limbic system, bypassing the slower linguistically based processing in the auditory cortex. Hearing thresholds decline substantially above 3 khz among many tinnitus patients, and hence the spectral modification should involve reducing the energy of lower frequency components of the music. Tinnitus retraining therapy As mentioned before brain has the unique property of neural plasticity. This implies that the brain can be retrained for habituation to the tinnitus sound and this is the basic idea behind Tinnitus retraining therapy (TRT). TRT mainly targets nonauditory systems, particularly the limbic and autonomic nervous systems, and is based on the assumption that tinnitus represents a side effect of the normal compensatory mechanisms in the brain. The goal of TRT is to habituate the brain to the physiological reactions to tinnitus sound in order to habituate the patient to the tinnitus sound.(mechanisms of Tinitus 2001; Jasterboff 2007;Scheldrake et al. 1999; Hazell 1999) TRT mainly consists of two components: structured patient counseling (unlike the case of tinnitus masking where counseling is informal (Henry et al. 2006) ), and sound therapy, most frequently with the use of sound generators (which emit low level of broad-band noise), following a specific habituation protocol. Retraining counseling aims to help patients to think of their tinnitus 42

50 as a type of neutral sound (Han et al. 2009) and not associate it with threatening pathology. Tinnitus sound should never be masked in TRT, because habituation the sound cannot be achieved if it cannot be detected. Regardless of the benefits of TRT, it requires atleast 18 months to achieve observable stable effects, and this time-consuming treatment does not achieve satisfactory results in some patients. TRT requires patience and discipline from both the patient and a knowledgeable and experienced professional (Han et al. 2009; Jasterboff 2007). However it has been shown that the effectiveness of TRT increases with time (after 6 months) compared to that of tinnitus masking (Henry et al. 2006). explores a new treatment of tinnitus tinnitus intensive therapy habituation, which has shown a significant clinical effect. A new methodology - Neuromonics Tinnitus Treatment, has been investigated in, which overcomes the limitations of TRT by providing individually customized acoustic stimulus to the ears. The experiments performed by the authors showed a success rate of 86% using the new treatment method. Massage and stretching Massage and stretching of the neck and masticatory muscles have been associated with significant improvement in tinnitus. (Nelson 2003) Patients with somatic tinnitus can have symptoms of cervical spine disorders, including head, neck, and shoulder pain as well as limitations in sideways bending and rotation. Treating jaw and neck disorders has beneficial effects on tinnitus. Injecting lidocaine into jaw muscles, such as the lateral pterygoid, also reduces tinnitus (Nelson 2003). 43

51 Electrical suppression Electrical stimulation of the cochlea with trains of pulses at 5,000 pulses per second can substantially or completely suppress tinnitus with either no perception or only a transient perception of the stimulus. Stimulus with electrical pulses at such a high rate restores spontaneous-like patterns of spike activity in the auditory nerve, which could explain how it suppresses tinnitus. Transcutaneous electrical nerve stimulation of areas of skin close to the ear increases the activation of the DCN via the somatosensory pathway and could augment the inhibitory role played by this nucleus on the CNS, thereby ameliorating tinnitus. Clinical studies : Past and Present From the above literature survey, it is evident that there are diverse ways of treating tinnitus but there is no proven single solution. This is due to the fact that the type and characteristics of perceived tinnitus differs from patient to patient. In this project masking therapy has been considered to be used for the treatment of the patients. Due to the diversity in the characteristics of perceived tinnitus sound the effectiveness of masking therapy also varies from patient to patient. This necessitates the use of a structured clinical approach for psychoacoustic measures of the tinnitus. Clinical measures of tinnitus loudness, pitch, maskability, residual inhibition and patient s audiogram provide important information for the patient assessment and counseling and judging the feasibility of masking therapy on a patient(henry et al. 2004; Henry and Meikle 2000; Lokenberg 2000) A methodical assessment will help to : 44

52 Diagnose the underlying cause(s) of tinnitus better by sorting out the symptoms. Quantify the subjectiveness of the perceived tinnitus by a patient to counsel patient and family and thus provide psychological help Design a treatment that addresses the tinnitus symptoms as well as the psychological effects Make a judgment if tinnitus masking is an appropriate treatment for a particular patient based on the psychoacoustic measures of the perceived tinnitus. Progressive Audiologic Tinnitus Management (PATM) (Myers et al.; Research Highlights 2009), is an example of such a clinical methodology for tinnitus management which involves a series of clinical studies for maximizing the management of the tinnitus by minimizing the impact of tinnitus on the patient s life as efficiently as possible. PATM has 5 hierarchical stages - (1) Triage, (2) Audiologic Evaluation, (3) Group Education, (4) Tinnitus Evaluation, and (5) Individualized Management. Though traditional clinical methods for managing tinnitus have been around for quite some time (Davis et al.; Hanley et al. 2008) yet the methodology varies considerably among clinics and moreover there is limited number of clinics providing the clinical treatment of tinnitus (Henry et al Gilbert et al. 2004). In addition, the use of traditional clinical audiometers limit their scope to be easily accessed and used by a patient. Thus there is a need for a standardized clinical measurement which can be easily used by the patient with or without the guidance of a clinician. A possible solution to this can be the use of an automated clinical measurement with user friendly guidelines. 45

53 (Henry et al. 2004) proposes such a computer-automated tinnitus assessment system. The authors have developed a user friendly system for quantifying tinnitus by measuring the acoustic parameters- loudness, pitch and the patient s hearing threshold. The results obtained by the authors show that such computerized systems are suitable for clinical applications. The same authors present a comparison between automated and manual clinical approach for tinnitus assessment (Gilbert et al. 2004), where they prove the validity and ease in use of the computerized systems. The research work presented here is based on these lines. A tinnitus analyzer software has been developed for computerized clinical treatment of tinnitus. Similar to the software presented in (Snall 2004), the tinnitus analyzer software also uses the patient feedback for pitch and loudness matching. However in the proposed software there has been an addition of a module to generate masker signal. But the main novelty of the proposed software lies in the fact that the results from the 4 tests pitch matching, loudness matching, residual inhibition and minimal masking level, are used for judging if masking therapy is feasible for a particular patient. It is important to understand that though this software generates masker signal similar to the tinnitus maskers available online, yet the difference between them is that the proposed software provides a structured clinical treatment which the online maskers lack which decreases their validity. 46

54 2.4 Characterization of Tinnitus sound (Han et al. 2009; Andersson; Tyler; Chan 2009) The perceived tinnitus can vary from a quiet background noise to a noise more audible than external sounds with average loudness. Tinnitus is generally divided into two categories: objective and subjective. Objective Tinnitus This form of tinnitus is rare and in this case the perception of the tinnitus like sound is not just limited to the sufferer but can also experienced by an examiner using stethoscope. The noises are usually caused by vascular anomalies, repetitive muscle contractions, or inner ear structural defects. The sounds are heard by the sufferer and are generally external to the auditory system. Benign causes, such as noise from TMJ, openings of the Eustachian tubes, or repetitive muscle contractions are some of the causes of objective tinnitus. The sufferer might hear the pulsatile flow of the carotid artery or the continuous hum of normal venous outflow through the jugular vein when in a quiet setting. The sounds may arise from a turbulent flow through compressed venous structures at the base of the brain. Subjective Tinnitus Unlike objective tinnitus, subjective tinnitus is audible only to the patient. Subjective tinnitus is a symptom of a number of different underlying pathophysiologic processes. Causes of subjective tinnitus include otologic, neurologic, infectious, and drug-related (Lockwood et al., 2002). Otologic cause is the most common cause of subjective tinnitus. These include noise-induced hearing loss, presbycusis, otosclerosis, otitis, cerumen 47

55 impaction, Meniere s disease, and sudden sensorineural hearing loss. Neurologic etiologies include head injury, whiplash, multiple sclerosis, vestibular schwannoma, and other cerebellopontine-angle tumors. Tinnitus may arise as a result of a number of infectious sources such as otitis media, Lyme disease, meningitis, or syphilis. Medications also constitute a common cause of subjective tinnitus. Most commonly implicated drugs include salicylates, non-steroidal anti-inflammatory medication, aminoglycocide antibiotics, loop diuretics, and chemotherapy agents. Many physicians use the term tinnitus to designate subjective tinnitus and the term somatosound to designate objective tinnitus. The sounds associated with most cases of tinnitus have been described as being analogous to cicadas, crickets, winds, falling tap water, grinding steel, escaping steam, fluorescent lights, running engines, and so on. It is believed that these types of perception result from abnormal neuronal activity at a subcortical level of the auditory pathway. The pattern characterizing tinnitus is related to the library of patterns stored in auditory memory and also, via the limbic system, associated with emotional states. The characteristics of tinnitus are generally unrelated to the type or severity of any associated hearing impairment, and thus the latter offers little diagnostic value. Most tinnitus patients match their tinnitus to a pitch above 3 khz. The tinnitus characterizing Meniere s disease, described as roaring, matches a low-frequency tone that is usually from 125 to 250 Hz. However, tinnitus in the advanced burned-out stage of Meniere s disease is often higher in pitch and tonal in quality. Most patients with both tinnitus and hearing loss report that the frequency of the tinnitus correlates with the severity and frequency characteristics of their hearing loss, and that the intensity of the tinnitus is usually less than 10 db above the patient s hearing threshold at that 48

56 frequency. Some patients who have central auditory processing disorders and have difficulties in understanding speech in noise, report experiencing tinnitus even though their pure-tone audiometric thresholds are normal. Less prevalent forms of tinnitus, such as those involving well-known musical tunes or voices without understandable speech, occur among older people with hearing loss and are believed to represent a central type of tinnitus involving reverberatory activity within neural loops at a high level of processing in the auditory cortex. Somatic tinnitus is a type of subjective tinnitus in which the frequency or intensity is altered by body movements such as clenching the jaw, turning the eyes, or applying pressure to the head and neck. Reports that tinnitus is louder upon awakening suggest the involvement of somatic factors, such as bruxism. Reports that tinnitus vanishes during sleep but returns within a few hours further suggest that psychosomatic factors, such as neck muscle contractions occurring in an upright position or jaw clenching, play etiological roles. Because objective tinnitus (which is audible to another person) represents the semantic opposite of subjective tinnitus, a better nosological approach might be to use the term somato sound instead of objective tinnitus irrespective of whether the sounds are audible to others, reserving the term tinnitus for the perception of sound in the absence of any acoustic source. Thus, tinnitus would describe cases previously diagnosed as subjective tinnitus. Objective tinnitus might be vascular or mechanical in origin. Objective tinnitus of vascular origin could be a referred bruit from stenosis in the carotid or vertebrobasilar system. Objective mechanical tinnitus is due to abnormal muscular contraction of the nasopharynx or middle ear, as can occur in palatal myoclonus. Pulsatile tinnitus can also manifest subjectively as an increased awareness of blood flow in the ear. Indeed, the cause of somatosensory pulsatile tinnitus 49

57 syndrome is not vascular, with the syndrome deriving from cardiac-synchronous somatosensory activation of the central auditory pathway or the failure of somatosensoryauditory central nervous system (CNS) interactions to suppress cardiac somatosounds. Pulsatile tinnitus superimposed on steady tinnitus could result from the pulsation of blow flow with the spiral capillary of the basilar membrane. 50

58 CHAPTER 3 SYSTEM ARCHITECTURE 3.1 Proposed Scheme: A Tinnitus analyzer has been developed (in MATLAB) for individual profiling and analysis of the perceived tinnitus sound by a patient. The proposed software tool has 3 main modules Audiogram Generator, tinnitus sound generator and a masker analyzer. A) Hearing Test Module/ Audiogram generator One of the important parameters to measure tinnitus sound is its loudness which is in terms of sensation level i.e. in reference to the actual hearing threshold at that particular frequency. Hence this module has been developed to measure the frequency response of the ear and generate the audiogram. B) Tinnitus sound generator: Prior to any successful treatment of a tinnitus patient, it is essential to quantify the perceived tinnitus sound as accurately as possible. This module has been designed in a way to allow the doctor/ patient to do the following: 51

59 1) Vary frequency and bandwidth of the white colored noise, sinusoidal and distorted sinusoidal 2) Choose a particular ear in case of localized tinnitus 3) Mix and match the noise and the tone components 4) Measure the loudness of the tinnitus sound 5) Save the results in the local drive for future reference. C) Masker Analyzer The main function of this module is to analyze the feasibility of masking therapy for a particular patient depending on the values of minimum masking level and residual inhibition and then generate masker This section discusses the above mentioned functional modules in details. Fig 3.1 shows the proposed scheme through a block diagram. 52

60 Fig 3.1: Block Diagram of the proposed scheme. The 3 modules are Audiogram generator,sound simulation system/tinnitus sound generator and masker analyzer. 3.2 Hearing test module/ Audiogram generator The human ear does not respond to all the frequencies in the same manner. The audible range for a normal human ear is 20Hz to 20 KHz and this varies with age, health, working conditions and so on. As mentioned in section 2.12 ear acts as a filter which prefers some frequencies over others. The intensity of sound required to detect it varies 53

61 with the frequency component of the sound. For example intensity of sound needed at 20 Hz is about 1 watt per square meter whereas at 100 Hz the intensity needed to hear a sound is W/m 2. The hearing threshold decreases with increase in frequency. In other words it can be said that tones of different frequencies if played at the same volume will still seem to be played at different sound volumes. This is due to the fact that ear has a non-flat frequency response. On an average, humans are most sensitive to tones at about 3500 Hz, because these tones require the least gain (-4 db). The frequency response of the ear varies depending on age and health. Measuring Sound Intensity Sound intensity is defined as the sound power per unit area. The basic units are watts/m 2 or watts/cm 2. Many sound intensity measurements are made relative to a standard threshold of hearing intensity I 0 : I o watts / m 10 watts / cm (1) The most common approach to sound intensity measurement is to use the decibel scale: I I( db) 10log [ ] (2) 10 I 0 Since audible sound consists of pressure waves, one of the ways to quantify the sound is to state the amount of pressure variation relative to atmospheric pressure caused by the sound. The standard threshold of hearing can be stated in terms of pressure and the sound intensity in decibels can be expressed in terms of the sound pressure: I P 2 P I( db) 10log 10[ ] 10log 10[ ] 20log 10[ ] (3) I P P

62 Where P 0 = 2x10-5 Newton/m 2 The pressure P here is to be understood as the amplitude of the pressure wave. This module is responsible for measuring the frequency response of one or both the ears. Measurement of the frequency response of a patient is important prior to the start of any formal tinnitus treatment due to the following reasons : An assessment of the audiogram of a patient helps the patient/ and the clinician to find out if the patient has any hearing loss and if so then in which region. This is important for proper diagnosis of tinnitus. In the tinnitus loudness match, which is a part of tinnitus generator module and will be discussed later, the tinnitus loudness will be expressed in terms of sensation level which is the intensity level of the perceived tinnitus measured as the number of levels above the hearing threshold. So for the calculation of the sensation level, the measurement of the hearing threshold is required. Working principal of the module: This module has been developed in MATLAB to measure the frequency response of one/ both the ears and generate an audiogram accordingly. The module has a user friendly GUI which makes it easily understandable to the patients. Both the backend and the front end module have been developed in MATLAB. This module does the following: 1) Interactive hearing test 2) Plot the audiogram 3) Save the results 55

63 1. Interactive hearing test Through onscreen instructions the patient is guided through the hearing test. In this test 29 frequencies from 40 Hz to Hz, each having 25 intensity levels (with a difference of 3 db ) have been considered. Intensity levels for each tone is presented to the patient till that tone is no more audible. The hearing threshold for each tone is measured by counting the number of intensity levels to which the tone is audible, and comparing this count to that of the tone 3500 Hz Hz has been chosen since frequencies approximately closer to this tone have the least hearing threshold. The detailed algorithm has been presented below. Algorithm 1. The sound volume of the computer is set to a level at which the tone of 3500 Hz is inaudible 2. The interactive hearing test begins (user choice) 2.1 A total of 29 frequency tones from 40 Hz to 16000Hz is stored in array freqs 2.2 One frequency is chosen at a time from the array 2.4 Calculating the intensity levels In the beginning of this section, the different equations to measure intensity has been shown. Two of the main equations are : I 10log 10 I 1 I( db) and 2 I P ( 1 db) 10log 10 P2 2 56

64 In this section each tone has 26 different intensity levels. => I( db) 20log 10 P1 P 2 => 1 I( db) => 20log 10 A A 2 3( db) 20log 10 A A 1 2 => => log 10 A A A A 1 => 2 A 2 A A1 => A (4) n 10 3 *( n1) 20 Where A 1 = amplitude of the tone T at an intensity level 1 ( has a value of 1) A 2 = amplitude of the tone T at an intensity level 2 A n = amplitude of the tone T at an intensity level n I = 3 db (since each succeeding intensity level is decremented by 3 db from the previous one so intensity at level 2 is 3 db less than that of level 1 and so on) 2.5 Using equation (4) 26 intensity levels are calculated including the 1 st level which has been chosen to have amplitude of 10 and are stored in an array named amps. 2.6 Amplitude value is chosen one at a time from the amplitude array 2.7 A sound signal is generated by sine waveform with the selected frequency, amplitude and a sampling frequency of 40 khz 57

65 2.8 The frequency is plotted against the sensation level 2.9 If the sound is audible to the patient then the patient presses a button on the front end GUI The presented sound intensity level is stored in a variable Steps 2.6 to 2.10 is repeated for all the stored values of the amplitude till the sound is no more audible to the patient The final intensity level at which the tone is detected is stored as the final sensation level for that tone in the array handles.response 2. Plot the audiogram From the previous step the hearing threshold for each tone was measured in terms of the intensity level till which the tone was audible to the patient. But the hearing threshold has to be denoted in db. For this purpose the threshold for each tone has to be calculated with respect to the intensity level of the tone of frequency 3500 Hz because it has the least hearing threshold which is -4 db. Algorithm 1. Let response_3500 be the minimum intensity level at which the tone 3500 Hz is audible 2 Subtract each value of the array handle.response from response_

66 3 Since the difference between 2 adjacent intensity levels is 3 db so the difference calculated from step 3.2 is multiplied to 3 to get the total decrement and then added to -4 db, which is the hearing threshold of 3500 Hz tone. The complete formula to calculate the hearing threshold in db for a particular tone is : response_db = 3*(response_3500-handles.response)-4; 4 The response_db (on yx-axis) is plotted with respect to values in freqs array (plotted on x-axis) on a semi-logarithmic scale after polynomial curve fitting. This is the required audiogram. 3. Saving the file Depending on the user choice the audiogram can be saved in the local drive as a. mat file. Figure 3.2 shows the complete flow diagram for this module 59

67 Fig 3.2: Flow chart of the audiogram generator. 3.3 Tinnitus Signal generator The biggest challenge for a tinnitus sufferer which makes tinnitus difficult to deal with is the fact that since tinnitus is a sound that is only perceived by a patient in his/her head (in case of subjective tinnitus), so it becomes inherently tough to describe the sound to others. This makes not only the job of a clinician to diagnose the patient difficult, but also the patient feels helpless as he/she cannot share their experience with their closed ones. Tinnitus sound is very complex and since it is just 60

68 perception in the brain so it is not easy to recreate it. However in case tinnitus sound is to be recreated, tinnitus sound can be broadly considered to be composed of noise /and tone component. Therefore tinnitus sound can be recreated with the help of signal processing. Recreating tinnitus sound is of a great importance since it helps to overcome the above discussed challenges and proves to be of a psychological help to the patient. To recreate the tinnitus sound the knowledge of the sound intensity and pitch is mandatory. It is also important to know if tinnitus is localized to a particular ear. Therefore the main functions of this module are: - Pitch matching - Loudness matching - Choose the ear where tinnitus is localized - Mix and match tone and noise components - Save results for future reference 1. Pitch matching As mentioned earlier tinnitus sound has noise and tone components. The noise component is generally colored noise which is actually white noise passed through some kind of digital filter. For this research work the noise component is colored noise (Ellis) generated by passing white noise through IIR bandpass filter. The tone components have been generated using sine and distorted sine. 61

69 Creating noise component: The color names for different types of sounds are derived from a loose analogy between the spectrum of frequencies of sound wave present in the sound and the equivalent spectrum of light wave frequencies. White noise has equal energy at all frequencies which produces a flat frequency spectrum i.e. it carries equal power in any band of a given bandwidth. White noise is a random signal and has no correlation. But for colored noise, the coloredness describes the noise component s correlation. More specifically, white noise has no correlation from one sample to the next, is completely random and thus has a flat spectrum. In contrast, colored noise exhibits some correlation from one sample to the next and has a spectrum that can describe with an equation based on frequency. For example the frequency equation for pink noise is 1/f, for brown noise is 1/f 2 and so on. It should be noted that a noise component s probability distribution function and its coloredness are two independent terms. There are 2 popular methods to generate colored noise. The first is based on Fourier representation, while the second uses the Auto Correlation Function. In the Fourier method the general shape of the spectrum's magnitude is considered, for example 1/f. Given the magnitude in the frequency domain, and after specifying the phase (must lie between ±π/2), the inverse FFT is used to create a time-domain signal which is colored noise -- each different phase waveform generates a different 1/f noise waveform in the time domain. The drawback in generating colored noise using the FFT method is that the more values needed in time-domain waveform, the longer the FFT must run. In fact, sequences of 62

70 moderate length going beyond several thousand values prohibit using this technique. Thus in these cases ACF is used which applies random-signal analysis to the problem. Based on ACF, this alternative scheme produces coefficients for a special type of FIR filter such that when it s fed with a series of random numbers (typically Gaussian with mean = 0 and arbitrary standard deviation), it outputs a sequence whose correlation characteristics can make it colored noise. This technique is possible because the ACF is directly related to the Fourier transform, which allows generating signals with a desired frequency spectrum. This study uses the second method to generate colored noise where the colored noise is generated by passing the white noise through a filter with the desired frequency response. Selection of the type of filter: Here digital filter has been used because it s easy to configure, implement and save values than analog filters. There are two types of digital filters Finite impulse response (FIR) and infinite impulse response (IIR). FIR FIR stands for Finite impulse response. The impulse response is finite because there is no feedback in the FIR. The difference equation that defines the output of an FIR filter in terms of its input is: y n] b x[ n] b x[ n 1]... bn x[ n ] (5) [ 0 1 N where: x[n] is the input signal, 63

71 y[n] is the output signal, b i are the filter coefficients, and N is the filter order an Nth-order filter has (N + 1) terms on the right-hand side; these are commonly referred to as taps. This equation can also be expressed as a convolution of the coefficient sequence b i with the input signal: N y[ n] b x[ n i] (6) i0 i That is, the filter output is a weighted sum of the current and a finite number of previous values of the input. IIR Filter IIR stands for Impulse Infinite Response. In case of IIR filter the impulse response is "infinite" because there is feedback in the filter; if the input is an impulse (a single "1" sample followed by many "0" samples), an infinite number of non-zero values will come out (theoretically). Digitals filters are often described and implemented in terms of the difference equation that defines how the output signal is related to the input signal: 1 y[ n] ( b0 x[ n] b1 x[ n 1]... bp x[ n P] a1 y[ n 1] a2 a 0 y[ n 2]... a Q y[ n Q]) (7) where: is the feedforward filter order 64

72 65 are the feedforward filter coefficients is the feedback filter order are the feedback filter coefficients is the input signal is the output signal. A more condensed form of the difference equation is: P i Q j j i j n y a i n x b a n y ]) [ ] [ ( 1 ] [ (8) which, when rearranged, becomes: P i Q j j i j n y a i n x b 0 0 ] [ ] [ (9) To find the transfer function of the filter, we first take the Z-transform of each side of the above equation, where we use the time-shift property to obtain: P i Q j j j j i z Y z a z X z b 0 0 ) ( ) ( (10) We define the transfer function to be: Q j j j P i i i z a z b z X z Y z H 0 0 ) ( ) ( ) ( (11)

73 Considering that in most IIR filter designs coefficient is 1, the IIR filter transfer function takes the more traditional form: P i bi z i0 H ( z) Q (12) 1 j a z j1 j FIR vs IIR filter Compared to IIR filters, FIR filters offer the following advantages: They can easily be designed to be "linear phase". So they have less phase distortion. They are stable as there is no feedback involved They are simple to implement. However the primary disadvantage of FIR filters is that they often require a much higher filter order than IIR filters to achieve a given level of performance. Correspondingly, the delay of these filters is often much greater than for an equal performing IIR filter. In this project the filter required to generate colored noise should be tunable, since the motive is to create an interactive tool that can be easily used by the patients and clinicians. A FIR filter will have more tunable parameters than an IIR filter. As the aim is to create an easily usable software, so here IIR filter has been used. The bandpass filter used here has just 2 tunable parameters which makes it easy to use. Design of the filter used: 66

74 The design of the filter used has been chosen from the book (Mitra 1998). Hence the filter will be only described briefly here. The transfer function is given by : H 1 (1 ) e ) [1 2 1 (1 ) e j e e 2 j ( BP j 2 j ] (13) where, is inversely proportional to bandwidth is directly proportional to center frequency and are 2 tunable parameters in this tool The squared magnitude of the filter is given by: 2 2 (1 ) (1 cos 2) H BP ( ) (14) [1 (1 ) 2 (1 ) cos 2 cos 2] 2 2 H BP ( ) becomes 0 at 0 and. The maximum value of unity is reached when 0, where 0 is the center frequency of the bandpass filter. This is used to calculate using cos( 0 ). For ensuring stability of the IIR filter, <1. Taking all the above points under consideration the program has been written. The user can change the value of (from to 0.99) and the center frequency 0 (from 0 to 2250 Hz) through the interactive front end GUI. Below is the algorithm which includes how the user entered values are used to generate colored noise. Fig 3.3 shows the flowchart of this algorithm. 67

75 Algorithm 1. User sets using slider with the handle name bandwidth. 2. The value set by the user is extracted using handles.bandwidth and is stored in variable a1 a1 = get(handles.bandwidth,'value'); 3. User sets 0 using slider with the handle name center_freq. 4. The value set by the user is extracted using handles.center_freq and is stored in variable b1 b1 = get(handles.center_freq,'value'); 5. b1 is used to calculate the value of using b1=cos(b1); 6. The numerator and the denominator the IIR filter to be used is set by b=[((a1+1)/2);0;(((-1)-a1)/2)]; a=[1;-(b1*(1-a1));-a1]; 7. The sampling frequency has been chosen to be Hz which is double the maximum hearing threshold for human beings which conforms to Nyquist theorem 1 fsamp 2 f T chosen Where, f samp is the sampling frequency selected by the user using GUI f chosen is the center frequency selected by the user using GUI sampling_t = 1/44100; 8. Since white noise is random signal so it can be generated using randn function t = 0:sampling_t:1; size1=size(t,2)-1; 68

76 x = randn(size1,1); 9. The appropriate filter, with the calculated numerator and denominator, is designed using the command freqz [h,f]=freqz(b,a,512,t) Where, h stores the frequency response of the filter f stores the first 512 frequency points between 0 and T/2 10. Colored noise is generated by passing the white noise through the filter designed in step 9 y= filter(b,a,x); 11. The sound is generated by sound(y,t); Algorithm for pitch matching: 1. The user changes the center frequency and bandwidth through front end GUI. 2. These values are used to generate colored noise component 3. The user can select a pure/ distorted tone component and change its frequency 4. If the pitch of the sound generated doesnot match the perceived tinnitus then the user can vary accordingly. 69

77 Fig 3.3 Flow chart of pitch test 2. Loudness Matching Retrieving the hearing threshold The button threshold, on the GUI is used to retrieve the hearing threshold at the frequency chosen from the pitch matching section. It uses the parameter - threshold passed from function eartest. Fig 3.4 shows the flowchart for the loudness test. Algorithm: 1. The frequency selected during the pitch test is stored in variable b11 is used here and stored in variable b Since the array threshold holds hearing thresholds for only 29 preset frequencies between 40 Hz to 16 Khz so the selected frequency b11 has to be rounded off to the nearest frequency in the array threshold. The code snippet below shows the rounding off rules: 70

78 if (b12>1 & b12<41) b11=40; elseif(b12>=41 & b12<70) b12=60; elseif(b12>=71 & b12<=100) b12=100; elseif(b12>100 & b12<=950) b12=round(b11/100)*100; elseif(b12>950 & b12<=16000) b12=round(b12/1000)*1000; else b12=16000; end 3. The respective hearing threshold for the rounded off frequency is extracted from the array Threshold and displayed (after converting to string) in the text box with a handle name hearing threshold. level1=threshold((find(threshold(:,1) == b12)),2); set(handles.hearingthreshold,'string', int2str(level1)); Loudness check at the chosen frequency The loudness button is used by the patient to check the loudness of the perceived tinnitus sound. A sinosidal tone is played at the hearing threshold, so that the patient can check if the loudness matches that of the perceived tinnitus sound. 71

79 Algorithm 1. A sinosidal tone is generated with frequency b11 t=[0:1/44100:1-1/44100]; y1= sin(2*3.14*b11*t); 2. To get the loudness at a level chosen in the previous algorithm the, intensity factor is calculated using amp_ratio= 10 ^ (level1/10); 3. The required sound intensity is obtained by multiplying the amplitude ratio or the intensity factor with the sinosidal signal in step 1 y1= amp_ratio.*y1; set(handles.loudness,'value',level1); Adjust the loudness If the loudness doesn t sound correct to the patient then he/she can adjust the loudness using the slider in the front end GUI. Algorithm 1. A tone is generated with frequency b11 t=[0:1/44100:1-1/44100]; y1= sin(2*3.14*b11*t); 3. The current slider value (which is the loudness in decibels) is read into a variable named decibels. 4. The value of decibels is displayed in the texttbox with the handle name loudnessdisp 72

80 decibels = get(hobject,'value'); set(handles.loudnessdisp,'string', int2str(decibels)); 5. The factor for intensity change is calculated in a way similar to the previous algorithm (step 3) 6. As mentioned before the loudness of the perceived tinnitus sound as selected /adjusted by the patient is expressed in terms of sensation level which is the difference between the intensity level of hearing threshold and the that of the adjusted loudness (of the perceived tinnitus sound) by the patient. sensationlevel=abs(level1-decibels); 6. The final measured sensation level is displayed in the textbox with the handle name Sensationlevel set(handles.sensationlevel,'string', int2str(sensationlevel)); 73

81 Fig 3.4 : Flowchart of the loudness test 3. Localization of tinnitus There are cases when tinnitus is perceived in only one of the ears. It is therefore important to give the user the option to choose between one of the ears, in case the patient experiences such a type of tinnitus localization. To achieve sound localization the required programming has been done in MATLAB. When we hear sound from a PC, we hear it through the loudspeakers which are located at the extreme right and the left corners serving the right and the left ear respectively. The illusion of an imaginary sound source somewhere between these two physical ones is 74

82 reached by imitating the natural behavior of it. Normally sound-waves from the sound source reach the two ears with a small time delay caused by the position of the ears and an intensity difference caused by various reasons (e.g. damping of the waves by the head). It is possible to create an intensity difference large enough to give the listener an illusion of sound localization. This principle has been used here to localize the sound to the selected ear. When the user makes a choice between left and the right ear by pressing the respective button the function localized.m is called whose aim is to localize the sound based on the principle discussed before. The algorithm of this function is given below and the same is depicted in fig 3.5.: Algorithm 1) The parameters amp_diff and filename are passed to this function from the previous function. Amp_diff is the intensity difference between the sounds heard in left and the right ear, if it is positive then the sound in the right ear is intensified and vice versa. The variable filename is the audio file which stores the last played sound n the previous function. 2) Y= filename 3) The number of rows i.e the size of the array Y is stored in variable len len=size(y,1); 4) Initialize a 2 d stereo matrix S. The first column has sound signal values for left ear and the 2 nd column has sound signal values for right ear S = zeros(len,2); 75

83 5) Using the loudness equation the amplitude ratio is calculated from amp_diff (in db in this case the value is +/- (33 db)) using ratio = 10^(amp_diff/10); The amplitude factor for the right ear is given by rf = ratio / (ratio + 1); and for left ear it is given by lf = 1 / (ratio + 1); 6) The amplitude factor for the left ear is multiplied with the original sound signal Y to obtain the sound signal for left ear. 7) The amplitude factor for the left ear is multiplied with the original sound signal Y to obtain the sound signal for left ear and is stored in the first column of the S array L = lf * Y; S(:,1) = L; 8) The amplitude factor for the right ear is multiplied with the original sound signal Y to obtain the sound signal for right ear and is stored in the second column of the S array R= rf * Y S(:,2) = R; 11) The sound is played using sound(s, FS) and the result is sound localization in the ear depending on the sign of the amp_diff 76

84 Fig 3.5 Block diagram for sound localization 4. Mix and match sounds The user can select from noise or tone components or both to create a sound close to their perceived tinnitus. 5. Save the results The user can save the selected frequency, loudness and the sound signal in the local drive for future reference 3.4 Masker analyzer This module is responsible for the generation of masker signal and then judging whether the masking therapy is viable for a particular patient based on the results from residual inhibition test and minimal masking level test. This module has the following sub modules: 77

85 - Generation of masker signal - Residual Inhibition test - Minimal Masking Level Test - Save the masker signal 1.Generation of masker signal The main function of this sub module is to generate masker signals. Here the patient has the option to choose from the preloaded sounds and vary it s the pitch and loudness to mask the perceived tinnitus. The patient also has the option to load an audio file of their choice from their local drive which they would like to use as a masker signal. There are 4 natural sounds, 4 noise sounds and sinosidal tone which are preloaded in the software. The patient has the option to choose from any of the sounds through an interactive front end GUI and can vary the pitch and loudness for the chosen signal using a slider in a front end GUI. The 2 main functions here are to vary pitch and loudness. Varying pitch The fundamental concept behind achieving the pitch shift of an audio signal is to change the sampling rate of the signal keeping the length of the signal constant. Phase vocoder is a popular method to achieve pitch variation, and has been used in this project to achieve the same (Heller 2003; Managing tinnitus 2009). The fundamentals behind phase vocoder have been briefly described below. 78

86 Phase Vocoder: The phase vocoder is a digital signal processing technique (Sethares) for timescale modification of audio. The phase vocoder can be categorized as an analysis-synthesis technique which takes an input signal and produces an output signal which is either identical to the input signal or a modified version of it. The underlying assumption is that the input signal can be well represented by a model whose parameters are varying with time. The analysis is devoted to determining the values of these parameters for the signal in question, and the synthesis is the output of the model itself. Phase vocoder has 2 mathematically complementary implementations Filter Bank representation and Fourier Transformation. The latter representation has been used in this project. In this implementation the signal is first broken down into small parts and then its FFT is calculated. This is known as STFT short time fourier transform. This is then modified according to the requirements. For example in order to time-expand a sound, the FFT's can simply be spaced further apart than the analysis FFT's. As a result, spectral changes occur more slowly in the synthesized sound than in the original. But this overlooks the details of the magnitude and phase signals in the middle and hence, the frequency of the signal is altered. The solution is to rescale the phase by precisely the same factor by which the sound is being time-expanded. This ensures that the signal in any given filter band has the same frequency variation in the re-synthesis as in the original (though it occurs more slowly). The altered FFT frames can then be re-synthesized back to time domain using inverse FFT. 79

87 Since the phase vocoder can be used to change the temporal evolution of a sound without changing its pitch, it is also possible to do the reverse (i.e., change the pitch without changing the duration). This can be achieved by time scaling the pitch by the desired factor, and then to play the resulting sound back at some other wrong sample rate. For example, to raise the pitch by an octave, the sound is first time-expanded by a factor of two, and the time-expansion is then played at twice the original sample rate. This shrinks the sound back to its original duration while simultaneously doubling all frequencies. To achieve this the MATLAB implementation of phase vocoder by Sethares has been used in this project and the associated algorithm has been described in this section (Fig 3.6 depicts this schematically). A. Computing Short Time Fourier Transform stft.m is the function written in MATLAB which is responsible for computing STFT. In this function the main idea is to take the signal x k (where k is an index representing each successive element of the signal) and do the following. 1. Take a small sub-section of the signal of size M samples 2. Window it by multiplying by a suitable windowing function (hamming, hanning, etc). X 3. Apply a DFT or FFT on the section of the signal to generate a spectra m (where m is an index representing successive elements of the spectra). 4. Store the spectra in an array of different spectra indexed by the integer l. 5. Hop by s samples and repeat the process. 80

88 Fig 3.6 Schematic representation of STFT Below is the algorithm of stft.m based on the above Algorithm 1) Takes input signal (x), fft sample points (1024), window size (1024) and hop size (256) 2) Form the Hann window function halfwin = 0.5 * ( 1 + cos( pi * (0:halflen)/halflen)); where, halflen= w/2 halff = f/2 3) Initialize an array win [1 x1024] 4) Transpose array halfwin to array win so that the size is [1x1024] 5) Initialize the output array d [512x(s-f)/h] where s is the length of x 6) Multiply the win with the input signal to form windows with 1024 points 81

89 u = win.*x((b+1):(b+f)); where, win is computed from the window equation b = 0: h: (s-f) 7) For each of the above short time frames compute the FFT. This is the STFT 8) Store these values in the output array d t = fft(u); d(:,c) = t(1:(1+f/2))'; where, c =1: (s-f)/h 9) Return array d (generates (s-f)/h number of STFT bins/frames each with 512 FFT points) B. Time Compression/ Expansion The next job is to modify the time duration of the STFT frames. The function pvsample.m is used to time expand /compress the STFT frames. Algorithm 1) Takes magnitude scaled STFT (b), t (0: r: cols-2, where r is the time alteration factor and cols is the number of columns of the array d received from the previous algorithm) and hop (256) 2) The output array C is initialized with a size of 512 x length of t 3) Initialize the variable ph with the angle of the first frame in the STFT array b 82

90 ph = angle(b(:,1)); 4) Append a column to the end of b b = [b,zeros(rows,1)]; 5) For every value in the array t (0:r:ols-2), magnitude and phase have to be calculated 6) 2 columns from array b is taken one at a time and is used to calculate magnitude and phase for the new point 7) tf = t - floor(t); bmag = (1-tf)*abs(bcols(:,1)) + tf*(abs(bcols(:,2))); % calculate phase advance dp = angle(bcols(:,2)) - angle(bcols(:,1)) 8) The angle dp is reduced to pi to pi range 9) The total value for the new time frame/point is calculated using the magnitude and the phase values and stored in the output array : c c(:,ocol) = bmag.* exp(j*ph); where, ocol changes from 1: (cols-2)/r 10) The phase is cumulated to be used in the next time frame ph = ph + dp; 11) Returns the array c C. Convert frequency domain back to time domain 83

91 The function istft.m uses overlap save method to achieve the resynthesis of the time domain signal. The algorithm of this is given below. Algorithm 1) Takes parameters d (array c from the previous algorithm), ftsize (1024), hann windowing points w(1024) and hop points h(256) 2) Form the windows equation using the same method as used in step 2 of the algorithm for stft.m 3) Calculate the inverse fft using inbuilt MATLAB function ifft for b = 0:h:(h*(cols-1)) ft = d(:,1+b/h)'; ft = [ft, conj(ft([((ftsize/2)):-1:2]))]; px = real(ifft(ft)); x((b+1):(b+ftsize)) = x((b+1):(b+ftsize))+px.*win; end; D. Main wrapping function - pvoc.m This is the main function which calls the functions stft.m, pvsample.m and istft.m, to time expand/ compress an input audio signal. Algorithm 1) Takes input parameters x (input filename), r (time alteration factor) 2) FFT size is taken as 1024 and hop size is taken to be 25 % i.e 256 3) Function STFT is called to generate divide the signal into small time frames and then compute their FFTs 84

92 4) The output from STFT is scaled by 2/3 5) Function pvsample.m is called to a modified spectrogram array by sampling the original array at a sequence of fractional time values (the fraction is user selected), 6) Interpolating the magnitudes and cumulating the phases with each upcoming frame. 7) Convert the modified spectrogram back into sound using istft.m E. Pitch transposition For pitch shifting the time is first altered by factor which the user selects for pitch shifting. This is done using pvoc.m. The time shifted signal is then resampled i.e the sampling rate is changed using the inbuilt MATLAB function resample. The MATLAB function resample takes 3 parameters input signal to be resampled, p and q (p/q is the fraction by which the rate of sampling is to be changed). Thus by time duration modification in the first step, it is ensured that the length of the signal be constant after changing the sampling rate. The code used for pitch transposition is showed below: function pitch(input_filename,step,output_filename) step1=step/10; [d,sr]=wavread(input_filename); e = pvoc(d, step1); f = resample(e,step,10); % NB: 0.8 = 4/5 sound(f,sr); wavwrite(f,sr,output_filename); 85

93 Fig 3.7: STFT, manipulation of spectral frames and IFFT of the frames. Figure taken from 2. Varying loudness After selecting one of the preloaded sounds and setting the approximate frequency, the user can vary the loudness of the sound using a slider provided in the front end GUI. Algorithm 1. Read the slider value and store it in a variable level 2. Calculate the intensity factor amp_ratio using amp_ratio= 10 ^ (level/10); 3. The required sound intensity is obtained by multiplying the amplitude ratio or the 86

94 intensity factor with the already selected signal y y= amp_ratio.*y; 3. Minimal masking level test It is extremely mandatory to check the minimum loudness of masker signal that can mask the tinnitus sound. The loudness level is used to judge the feasibility of the masking therapy on a patient. If the loudness is more than a certain preset threshold (in this case 80 db has been chosen as the threshold), then the use of masker signal should be avoided as it might have adverse effects on hearing. Once the patient feels satisfied with the masker signal, the loudness value is checked and accordingly feasibility of the masking therapy is judged. 4. Residual inhibition test In this interactive test, the patient first hears the masker signal generated in sub module A, and then notifies through the front end, if the perceived tinnitus completely/partially disappears or remains unchanged. In case there is no change or increase in the perceived tinnitus after the removal of the masker signal then the software notifies the patient that he /she is not a good candidate for masking therapy. 5. Saving the results The patient has the option to save the generated masker signal to use it in future. 87

95 CHAPTER 4 APPLIED EXPERIMENTATION The software tool in this research work has been developed with the intention to be used in the practical clinical scenario. Since it was not possible to test on actual patients (human subjects) due to time constraints, a static and dynamic testing was conducted on the developed software tool. The testing procedure was focused on validating the system and didnot include performance related tests like stress, load etc. The focus was limited to checking the errors in the code and failures during the execution, if any. This section mainly aims to demonstrate the navigation through the tool through a validation test case. 4.1 Assumed test case The test case assumes that the perceived tinnitus has both noise and tone components. The pitch of the noise component is 3969 Hz and that of the tone component is 1110 Hz (for sine) and 1550 Hz (for distorted sine) and loudness of 9 db above the hearing threshold at that frequency. The perceived tinnitus is assumed to be bilateral. This test case is executed through the following steps. 88

96 1.Click the enter button on the welcome screen to proceed. Fig 4.1: Welcome Screen 2. The first module presented is the audiogram generator. The onscreen instructions are to be followed to first adjust the system sound and then begin the test. 89

97 Fig 4.2 : Audiogram generator 3. Once the test is completed the final audiogram is plotted in a logarithmic graph. A mock case has been shown here. The x axis is the frequency and the y axis is the hearing threshold in db. The plot can be saved for future reference. The continue to tinnitus test button is used to go to the tinnitus generator module. 90

98 Fig 4.3: Audiogram plotted 91

99 4. The next module presented is tinnitus generator. Since the test case considered here has both noise and tone components so they are selected through the front end GUI. The bandwidth and center frequency is varied till a close approximation of the perceived tinnitus sound is achieved. The chosen frequency is displayed in a text box to know the exact value selected. Using the play sound button the chosen tinnitus parameters can be checked for their accuracy. Once the pitch of the sound components are selected, the loudness test is to be conducted. The threshold button, is Fig 4.4 : Tinnitus sound generator 92

100 used to get the hearing threshold (-4 db in the case shown below) at the pitch of the noise component (the pitch of the tone components are not considered in this project and will be taken into account as a part of the future work). The loudness can be adjusted through the slider till it sounds ok to the user. Once satisfied (loudness is 5 db in this case), the final sensation level button is used to get the loudness of the perceived tinnitus sound in terms of sensation level (for the particular case the sensation level is -4-5 =9 ). The generate masker signal button is clicked to go to the masker signal analyzer module. 5. In this module, one of the sounds from the panel nature or other sounds, is selected. When one is chosen the other options will be disabled as shown. Alternatively an audio file of user s choice can be loaded to be used as a masker signal. Once the sound is chosen the pitch and loudness can be adjusted using the sliders. Once the tinnitus sound seems to be just masked, the minimal masking test is conducted using the value of the minimal level. If this level is below 80 db (as in this test case) then it will be notified that masking therapy is suggested and vice versa. Then residual inhibition test is conducted interactively.the masker signal needs to be first replayed and after it stops the user has to choose one of the provided options. Accordingly the tool will notify if masking therapy is suggested or not. In this test case there is no change in the perceived tinnitus after the removal of the masker signal. After the completion of all the tests the generated masker signal can be saved in the local drive. 93

101 Fig 4.5 Masker Analyzer 94