The Transmission of Sound to the Cochlea in Normal and Pathological Human Middle Ears

Abstract

This article reviews basic and clinical research on the human middle ear. The topics include the use of temporal bones as models of the human middle ear; the integration of mechanical measurements, clinical results and middle-ear models to direct lines of inquiry; hearing with no tympanic membrane or ossicular chain; hearing after tympanic membrane replacement; the function of the tympanic membrane; and sound conduction through the ossicular chain.

Overview

Over the last 100 years, there have been numerous scientific studies performed by research groups around the world to describe sound conduction through normal and pathological human middle ears. Due to space limitations, this review summarizes work on several, but far from all, seminal topics performed in that time span. While it does include the work of many research groups in the Americas, Europe, and Asia over that longer time period, it concentrates on the human middle-ear work at the Massachusetts Eye and Ear Infirmary in the last 40 years. The purpose of this review is to compare basic and clinical functional outcomes and not to describe the anatomy of the middle ear; there are already multiple sources that describe the anatomy in varying detail [1–4].

As depicted in Fig. 1, the general philosophy codified by Saumil Merchant, Bill Peake and myself at the Massachusetts Eye and Ear in the 1990 s combines clinical outcomes (including surgical results and audiograms) in normal and pathological ears, predictive models, and multiple functional measurements in live and cadaveric human ears to address basic and clinical questions about middle-ear function [5]. Each of the three components of Fig. 1 can either drive or respond to inputs from the others. For example, a clinical hearing result can drive new model analyses that drive new mechanical studies, which in turn suggest new surgical procedures. Inherent in this scheme are the development of functional middle-ear models and the use of cadaveric human middle ears to allow measurements of physical variables that are not readily made in live human ears.

Fig. 1.

The interaction of clinical outcomes, including audiograms, acoustical and mechanical measurements in live or cadaveric ears, and model predictions that we use to raise and answer questions concerning the workings of the normal and pathological human middle ear. The questions or the answers can arise from any of the three apexes of the triangle

The Simple Transformer Model of the Middle Ear

To understand the middle ear, you need to start with some simple concepts. The common heuristic model of the middle ear is a transformer that converts sound energy in the air-filled ear canal outside the tympanic membrane, P_EC, to sound energy at the oval window, P_OW, at the entrance of the fluid-filled cochlea (Fig. 2). This transformation depends greatly on two ratios: the ratio of the areas of the tympanic membrane (TM) and the footplate of the stapes, and the ratio of the lengths of the malleus and incus “lever arms” [6]. The “ideal” version of this transformer model assumes the TM is a massless rigid plate, which, along with massless and rigid ossicles that rotate around a ligament-determined axis, is supported by infinitely flexible braces. Because of these assumptions, the ideal transformer model cannot predict the frequency dependence of middle-ear sound transfer, nor the sensitivity of the inner ear to sound. Such predictions require taking the mass and stiffness of the middle ear, the cochlear load, and cochlear sensitivity into account [1, 7, 8].

Fig. 2.

The middle ear as a transformer (a variation of Fig. 2.5 in [6]). A A schematic of the tympanic membrane and the ossicular chain. The malleus is labeled M, the incus I, and the stapes S. The three labeled anatomical features of the malleus are its anterior process a, manubrium or handle m, and umbo u at the inferior tip of the manubrium and the center of the TM. The acoustic input to the middle ear is the sound pressure just lateral to the TM P_EC, and the output is the pressure acting on the annular ligament within the oval window P_OW. The axis of ossicular rotation is orthogonal to the figure’s plane and runs through the red circle. B The ideal transformer model with massless rigid pistons of the areas of the TM A_TM and the stapes footplate A_FP, that are coupled together by a rotating massless and rigid malleus-incus lever with lever arm lengths l_M and l_I. C The relationship between P_OW and P_EC predicted by the ideal transformer model

The Use of Cadaveric Ears

A prominent tool in studies of middle-ear sound transmission in humans has been functional measurements in cadaveric ears. Reasons for this include: (i) many of the acoustic and mechanical measurements needed to investigate middle-ear function are invasive and not readily performed in live humans; (ii) cadaveric ears (usually the isolated temporal bone of the skull [3]) have the same form and structure as live ears, making it easy to investigate the effect of structural changes that mimic pathology, and test new prosthetic devices; (iii) as the mechanics of the middle ear are generally passive, unfixed cadaveric ears that are kept cool and moist (to slow the breakdown of soft tissues) function similarly to live ears [9–11].

The history of the use of cadaveric material for auditory mechanics is long and not without controversy. Prior to the middle of the 1930 s, most, if not all, auditory mechanical studies were performed in cadaveric ears [12]. In the 1930 s and 1940 s, Metz [13] and others began measuring the acoustic input impedance of the middle ear in live human ears, a technique that evolved into tympanometry [14], reflectometry [15–17], and modern wide-band absorbance [18–20].

An early comparison of middle-ear input impedance in live and cadaver ears was that of Zwislocki and Feldman [21], who concluded that cadaver middle ears were significantly stiffer (more impedance at low frequencies) than live ears. This finding was challenged by Rosowski et al. [9], who demonstrated similar middle-ear input impedance over the 0.2 to 4 kHz range in live humans and unfixed cadaveric temporal bones that had been kept cool and moist with aerated middle ears. They suggested the stiff-ear measurements of Zwislocki and Feldman resulted from the build-up of static pressure in the intact cadaveric ears they studied. Later, Goode and coworkers [10] and others [22, 23] measured the sound-driven velocity in the center of the TM (the location of the umbo of the malleus, Fig. 2) in normal live ears and cadaveric temporal bones and demonstrated great similarity of the velocities in the two preparations.

A complication in this story arose when Huber et al. [24] published the first intraoperative measurements of sound-driven stapes motion in live human subjects, and, as pointed out by Ruggero and Temchin [25], observed factor of 3 (10 dB) smaller sound-driven stapes motions in live compared to cadaveric ears at frequencies below 1 kHz. Chien et al. [26] pointed out that the differences in low-frequency behavior observed by Huber et al. [24] could be attributed to differences in methodology in the live and cadaveric ears: A limited exposure of the stapes in live ears (due to the presence of an intact facial nerve, which is often removed in temporal-bone preparations) led to a larger angle between the ‘piston-like’ translations of the stapes that are prominent at low frequencies and the laser beam in the live ear measurements. The larger observation angles resulted in an underestimation of stapes motion at low frequencies. A later report of Chien et al. [11] of their own live human stapes motion measurements, when corrected for the difference in laser angulation, suggests a great similarity between stapes motion in live and fresh cadaveric ears across the measured 0.2 to 7 kHz frequency range (Fig. 3).

Fig. 3.

A comparison of measurements of the magnitude of the sound-induced stapes velocity V_S, plotted as a transfer function of V_S/P_EC, in live and cadaveric human ears. The data are replotted from Fig. 9 of [11]

Hearing Loss with No Tympanic Membrane or Ossicular Chain

Forty years ago, a common question of otologists to those studying middle-ear function concerned the limited 50 to 60 dB conductive hearing loss in patients with no tympanic membrane or ossicles and normal inner-ear function (Fig. 4). Humans can hear sound levels over a range of more than 120 dB. If all sound reached the cochlea via the TM and ossicles, why didn’t loss of these structures produce larger (> 120 dB) hearing losses? This question was often followed by requests to explain how some patients with no ossicles had much less severe (30 dB) hearing losses than the more commonly observed 50 to 60 dB loss.

Fig. 4.

Mean audiograms from cases with interrupted or missing ossicular chains. The blue curve is the mean of 8 patients with an intact TM and complete interruption of the incudo-stapedial joint (ISJ) joint [27]. The red curve is the mean audiogram of five patients without a TM, malleus, and incus [12]. The green curve is the mean of four patients without TM, malleus, or incus that were treated with a type IV tympanoplasty [38] as described in the text

As has been known for many years, e.g., [12], the answer to both of these questions is that multiple pathways sound follows to stimulate the inner ear. Figure 5A illustrates three of these: (i) In the ossicular pathway (in red) P_EC induces motions of the tympanic membrane (TM) and ossicles that produce a sound pressure acting on the stapes in the oval window (P_OW) of the inner ear [12 (pp. 104–126)]. (ii) In the direct acoustic pathway (in blue) P_EC, either directly or via motion of the TM, produces a sound pressure in the middle-ear cavity (P_ME) that stimulates the lateral surfaces of both the oval and round windows of the cochlea (P_OW and P_RW) [12, 27, 28]. (iii) Sound stimulating the head induces vibrations of the head and skull that are conducted to the inner ear via multiple bone-conduction paths (in green) [12 (pp. 128–163), 28, 29].

Fig. 5.

The pathways that couple sound to the cochlea. A A schematic describing the three paths for coupling environmental sounds to the cochlea: The ossicular pathway, the direct acoustic pathway, and the multiple bone conduction pathways. A variation of Fig. 2 in [5]. B A comparison of the stimulus gain produced at the oval window (OW), or the equivalent, produced by the three pathways of A. Included as a reference is the effect of direct acoustic stimulation of the isolated OW (in purple) as occurs experimentally, e.g., [30, 31] or after type IV tympanoplasty. The ossicular path gain (in red) results from the normal TM and ossicular chain, as described by measurements of P_OW/P_EC [33]. The Direct acoustic path (in blue) is from measurements of $(P_{\text{OW}}-P_{\text{RW}})/P_{\text{EC}}$ with and without the TM [31]. The estimate of the equivalent stimulus gain at the OW produced by bone conduction comes from [35]. (Modified from Fig. 3 of [5])

The mechanism of cochlear stimulation in the ossicular and direct-acoustic paths is the generation of a difference in sound pressure at the two cochlea windows (the window pressure difference, $P_{\text{OW}}-P_{\text{RW}}$) that induces motion of the cochlear fluids between the windows [30, 31]. This fluid motion then initiates the mechanical stimulus to the inner ear [27, 32]. The bone conduction pathways that use the external and middle ear to pass energy to the cochlea also use the window pressure difference mechanism, but other bone paths (notably cochlear compression) do not.

The Ossicular Pathway

The ossicular pathway is dominant in normal hearing function and has been quantified by Kurakawa and Goode [33] as the dB ratio of stapes motions produced by P_EC in the normal middle ear, to that produced by stimulation of the isolated stapes footplate (the TM, malleus, incus and stapes superstructure are removed) within the oval window. As noted above, the ossicular path only drives P_OW and keeps P_RW small, with a resulting window pressure difference ($P_{\text{OW}}-P_{\text{RW}}$) approximated by P_OW alone. The ratio of the window pressure difference to the sound pressure in the ear canal (($P_{\text{OW}}-P_{\text{RW}}$)/P_EC) is the middle-ear gain [30]. In human temporal bones, the normal ossicularly conducted middle-ear gain varies with frequency between 26 and 0 dB (Fig. 5B) [33].

The Direct Acoustic Pathway

The direct acoustic path stimulates the cochlea by differential stimulation of the two cochlear windows [12 (pp. 115–126), 27]. The sound pressure in the ear canal moves the TM and creates a sound pressure in the middle-ear cavity, P_ME, that results in sound pressure at the two windows. Because of the small distance (2–3 mm) between the windows P_OW and P_RW are not perfectly equal, and small phase and magnitude differences between the two result in ($P_{\text{OW}}-P_{\text{RW}}$) of about 2 to 20% of P_ME in the 0.1 to 6 kHz range [12 (pp. 115–126), 27, 34]. The relative size of the middle-ear sound pressure and window pressure difference is best seen in Fig. 5B where the TM is removed, and P_EC approximates P_ME (the dotted blue line).

The Bone Conduction Pathways

Bone conduction results from the sum of multiple stimulus paths, including the generation of ear canal sound pressures that are coupled to the inner ear by ossicular and direct acoustic coupling. Other paths are direct stimulation of the inner ear via motion and compressions of the cochlea fluids and capsule [29]. These latter two BC mechanisms do not rely on the window-pressure difference as a stimulus. The summed effect of these multiple pathways on hearing function when the head is immersed in a sound field was quantified by Zwislocki [35], who measured sound-field hearing thresholds in normal hearing subjects before and after placing the combination of well-fit earplugs and sound-attenuating ear-muff hearing protectors. The BC contribution to hearing function plotted in Fig. 5B is the difference between Zwislocki’s measured thresholds with “maximum” hearing protection and the middle-ear gain in normal ears, where the latter describes the input to the inner ear with a normal middle ear. This difference suggests that sound stimulation of the whole head produces a BC stimulus to the inner ear equivalent to a ($P_{\text{OW}}-P_{\text{RW}}$)/P_Head that varies between − 50 and $-20$ dB. When compared with the normal middle-ear gain, this suggests hearing levels of 40 to 60 dB.

The Relative Strengths of the Three Cochlear Stimulus Paths

Figure 5B compares the strength of the three stimulus paths. The ossicular pathway is the drive for normal hearing function and produces a significantly larger ($P_{\text{OW}}-P_{\text{RW}}$)/P_EC than the others at all frequencies, while sound-driven bone-conduction [29, 35] competes with direct acoustic stimulation [27] to limit the most severe conductive hearing losses. The contribution of the direct acoustic path to hearing function has been described and quantified using a combination of models [12 (pp. 115–126), 27] and functional measurements [12 (pp. 104–126), 30, 34]. While all the evidence points out little role for the direct acoustic path or sound-driven bone conduction in the normal ear, when pathology reduces TM-ossicular sound conduction (e.g., after large TM perforations and/or interruption of the ossicular chain [12 (pp. 104–126), 27, 36] these two paths can effectively stimulate the inner ear, and the difference between the stimulus gain of the ossicular pathway and either the BC or the direct acoustic pathways explains the 50 to 60 dB hearing loss without TM or ossicular chain. Whereas some treatments for middle-ear disease use hearing aids that produce bone-conducted vibrations [37], the direct-acoustic path is of interest here in that it can be increased by simple surgical manipulations.

Figure 5B illustrates two measurements of ($P_{\text{OW}}-P_{\text{RW}}$)/P_EC due to the direct acoustic path [34]. The window pressure difference varies between $-30$ and $-60$ dB when the TM is intact, and is 20 dB larger, between $-10$ and $-40$ dB, when the TM is removed. The increased window-pressure difference in the latter case results from an increase in P_ME/P_EC after TM removal. This and other observations point out that the direct acoustic stimulus can be manipulated by changes in ear structure. This conclusion led to multiple surgical procedures to maximize the window pressure difference in ears with no TM or ossicles [38–40]. The common surgical manipulation in these procedures (types IV and V tympanoplasties) is the placement of stiff sheets of graft material in the middle ear to shield the round window from P_EC, thereby reducing P_RW, while equating P_EC and P_OW, with a resultant significant increase in window pressure difference The best of these procedures reduces the deficit in conductive hearing from 40 to 60 dB to 20 to 30 dB (Fig. 4) [38–40].

The Tympanic Membrane and Middle-Ear Function

This next section starts out with the clinical result of hearing measurements after a simple TM replacement surgery, and then describes measurements of the function and material properties of the normal TM. The section closes with a discussion of past and existing TM models, and notes attempts to use this information to manufacture new TM grafts.

As discussed in the “The Simple Transformer Model of the Middle Ear” section, the TM captures the sound energy in the external ear and transforms it into mechanical energy within the ossicular chain for transmission to the cochlea. This function depends on the area of the TM, its shape and mechanical properties, and how well the TM is coupled to the ossicular chain [2, 41]. Experiments that investigate the contribution of shape and material properties to TM function include measurements of the effect of manipulations on overall TM function [36, 42], detailed measurements of the motion of the surface of the TM [43–48], and model analyses [41, 49–51].

Hearing Results after Simple TM Replacement Surgery

My colleagues Heidi Nakajima, Aaron Remenschneider, and their coworkers [52] addressed the contribution of TM structure to hearing by measuring hearing and middle-ear function in 11 patients who had subtotal perforations of the TM with intact ossicular chains (Fig. 6). The TMs of all these patients were reconstructed using a thin sheet of muscle fascia that restored the original area, approximated the normal shape of the TM, but did not restore the normal mechanical properties of the membrane. The surgery employed a traditional ‘lateral graft’ technique with fascia used to replace the entire TM [53]. The fascia was placed lateral to the bony annulus and medial to the umbo of the malleus. This technique approximates the conical shape of the normal TM but relies upon healing of the graft to the malleus to establish an ossicular connection. The resultant air-bone gaps (ABG: a measure of conductive hearing loss) computed from the post-surgical audiograms are illustrated in Fig. 6A. The median post-surgical conductive hearing loss varied between 5 and 20 dB at different frequencies, but there was a 40 dB range of results with one outlier subject (#11) with a better than normal hearing result (negative ABG) at most frequencies.

Fig. 6.

Audiometric results and TM mechanics after muscle fascia tympanoplasty. The data are from Figs. 3 and 6 of [52]. A A summary of the post-surgery air-bone gaps measured in 11 patients who underwent muscle fascia tympanoplasty of the TM. The median, Interquartile Range and total range of 10 of the cases are shown along with a plot of one outlier, patient #11. B A comparison of the median and inter-quartile range (IQR) of wideband absorbance measured in a normal population (in red) and the 11 post-tympanoplasty ears (in blue). The absorbance A level is plotted in dB, where the dB value of power absorbance equals $10\text{log}10(A)$. The absorbance of patient 11 is within the IQR of the other ears

Wide-band absorbance (WBA) was also measured in the 11 post-surgical ears. Figure 6B compares the median and inter-quartile range (IQR) of the dB value of those measurements with the median and IQR of a population of 56 normal ears [18]. The normal and post-surgical absorbance measurements overlap at frequencies below 1 kHz, even though the median measured ABG in that range is near 20 dB. At higher frequencies, there is a clear decrease in the absorbance of the post-tympanoplasty group at higher frequencies, which suggests that in the 1- to 4-kHz frequency range, the reconstructed TMs absorb about half (– 3 dB) to a quarter (− 6 dB) of the incident sound energy as the normal TMs. While this decrease is highly significant (p < 0.0001), it is much smaller than the median 5 to 20 dB conductive hearing loss observed in these patients. These quantitative differences between the changes in WBA and ABG at low and high frequencies suggest that a significant fraction of the sound power absorbed at the TM after tympanoplasty at any frequency is not transmitted to the inner ear in these ears, and that the excess power absorption may occur within the graft and its imperfect coupling to the manubrium. As the areas of the reconstructed TMs are similar to the normal TM, the significant ABGs likely result from finer differences in the shape and structures of the graft TM and the restored ossicular connection. While others manipulated the structure of the TM to evaluate the effect on TM function, e.g., [36, 41, 42], we used temporal bone preparations to describe the sound-driven motion of the TM to help define the material properties that govern TM function.

Sound-Driven Motions of the Surface of Cadaveric TMs: Sinusoidal Drive

Motions Orthogonal to the Tympanic Annulus

Most measurements of sound-driven TM motion are made so that the measured motion is orthogonal to the bony annular rim that supports the TM. As Tonndorf and Khanna [43, 44] pointed out, the motions they observed with time-averaged holography can be described as the sum of multiple pressure-driven modal displacements [54, 55]. The displacement pattern associated with each mode is determined by its unique modal frequency as well as the membrane’s material properties and support constraints. The magnitude and relative phases of the modal displacements are related to the damping within the membrane and the difference between the frequency of stimulation and the modal frequency. Time-averaged holography measurements provided no phase or timing information [43, 44, 55]. The development of digital holography, high-speed digital cameras, and high-frequency stroboscopic laser illumination allowed the measurement of phasic variations in the motion of the entire TM surface over the frequency range of human hearing [46, 56, 57].

Some of the best examples of sinusoidally driven patterns of the sound-driven motion components orthogonal to the TM’s annular rim [45] are presented in Fig. 7. The figure contains twelve pseudo-3D images of sinusoidal sound-driven displacements of the surface of the TM arranged in three rows of varied tonal frequencies (0.5, 4, and 10 kHz) and four columns of varied stimulus phases (the instant closest to peak outward motion, + 0.25, + 0.5, and + 0.75 cycles), where the measurement protocol gathered 16 points per cycle. The displacements at each frequency are scaled as the fraction of the largest displacement on the surface during all four phases. Since the stimulus levels were chosen to optimize the holographic displacement measurements, a displacement value of 1 corresponds to about 0.5 μm at the three frequencies. The stimulus levels of the tones varied between 80 and 110 dB SPL [45].

Fig. 7.

The time-varying amplitudes of sinusoidally-driven displacements of the entire surface of a normal TM (data from [45]). Thirteen pseudo-3D instantaneous displacement maps are illustrated. At the top center is a schematic of the membrane at rest that shows the outline of the TM’s annular support ring and the manubrium attached to the TM. The planar dimensions of each 3D-plot are 10 mm ($-5$ to + 5 mm). In each plot, superior is at 12 o’clock, and anterior is at 9 o’clock. The next three rows picture the instantaneous displacements produced by constant frequency stimuli (0.5, 4, and 10 kHz). The four columns within each row show the displacements at different stimulus phases (the time at peak displacement, + 0.25 cycles, + 0.5 cycles, and + 0.75 cycles). The displacements are normalized to the largest displacements measured at each frequency, which were ~ 0.5 μm. The stimulus levels are noted. The green arrows in the first row point to the location of the umbo, which is visible as an indentation or protrusion in the lowest frequency responses. Inward displacements are coded from orange to yellow, while outward displacements are coded from orange to red

The first stimulus row shows a response to a 0.5 kHz tone, with two local maxima at the time of peak motion. This response approximates the lowest-order mode of surface motion in which each point on the surface moves with a similar phase, but where the magnitude of displacement at each point is constrained by the TM shape and local mechanical properties as well as by the TM’s attachments to the malleus and the bony annulus supporting the TM. The uniform phase of motion is illustrated by each surface point arriving at its near maximum outward motion in the leftmost column, with near maximum inward motions a half-cycle later, and near minimal motions at + 0.25 and + 0.75 cycles. (As the timing of the strobe illumination was not precisely at the peak of the displacement, the measured motion is neither 0, nor equal at 0.25 and 0.75 cycles.) Note that while the entire surface of the TM moves in-phase, this mode only approximates piston-like motion, as the images show the membrane attached to the malleus and the bony TM annulus moves less than the free-standing surfaces [43, 55]. The location of the manubrium and its umbo tip (green arrows) is easily visible in the displacements at each phase, showing reduced outward displacements relative to the rest of the membrane at the time of peak and 0.25 cycles later, and relatively reduced inward displacements at 0.5 and 0.75 cycles after peak.

With 4 kHz stimulation (the second row of Fig. 7), the number of local displacement maxima on the TM surface increases from two to about eight, consistent with the presence of multiple higher-order motion modes, each generating more than one motion peak [54, 55]. These local maxima vary in amplitude and occur at different phases of stimulation, e.g., the largest (peak) displacement occurs in the posterior region in the left-hand column, but regions superior and anterior to that show smaller local maximum motions 0.25 cycles later. While the motions at the four illustrated phases are of increased complexity compared to the 0.5 kHz tone results, near uniform motions near the membrane rim suggest the continued presence of low-order modes of motion [45]. While not easily visible in this display, because the multiple peaks mask the locations of the manubrium and umbo, the manubrium and umbo still move less than the surrounding TM surface.

With 10 kHz tonal stimulation (third row of Fig. 7), even more displacement modes are excited, and we see an increase in the number of spatially local displacement maxima and the complexity of the overall pattern. However, while the displacement patterns are more complex, they share many of the attributes of the lower-frequency responses including a continued contribution of more unform low-order modes, where the spatially varying higher-order modes ride on top of the low-order modes, with the result that a significant fraction of the TM surface still moves with similar phase. De LaRouchefoucauld and Olson [58] came to similar conclusions for the gerbil TM.

The continued presence of the more spatially uniform lower-order modes can help the TM transfer sound energy into ossicular motion, in a manner different from previous suggestions. Tonndorf and Khanna [54], whose phase-insensitive techniques did not reveal the continued presence of lower-order modes with higher-frequency stimuli, suggested that the more spatially variable higher-order modes on the membrane surface cancelled each other out, which produced a decrease in TM sound transfer at higher frequencies. Another view of high-frequency TM function is that of Fay et al. [41] who hypothesized the closely-spaced-in-frequency higher-order modes summed to improve high-frequency TM function. We believe the stroboscopic holography results of Cheng et al. [45] are most consistent with persistent lower-order modes, with the resultant near in-phase behavior of much of the TM, determining the transfer of TM motion to the ossicles at all frequencies. This result suggests that effective replacement TMs need only reproduce the lower-order modal motion patterns to be functional.

TM Shape and the 3D Motion of Points on the TM Surface

While motions of the TM orthogonal to the TM’s annular rim are easier to measure, the tent-like shape of the TM affects the measured motions, such that motions of surface regions that are not orthogonal to its supporting rim may be underestimated. This phenomenon is illustrated in Fig. 8A and 8B. Figure 8A illustrates the conical shape of a human TM measured using holographic techniques [46]. Figure 8B illustrates the same TM in two different spatial coordinate systems: The more generalized x, y, z coordinates (where y runs in and out of the illustration) can be applied to the entire surface of the TM and is used to measure z-directional motions that are orthogonal to the TM’s annular rim (as in Fig. 7). The $\eta ,\alpha ,\beta$ in Fig. 8B describes a coordinate system (where $\beta$ runs in and out of the illustration) local to a point on the TM surface, where $\eta$ is the normal vector perpendicular to the tangential surface (the $\alpha ,\beta$ plane) at that point. Knowing the 3-D shape and the motion along the x-, y-, and z-directions allows computation of the motion along the $\eta ,\alpha ,\beta$ directions at each point on the surface of the membrane [46].

Fig. 8.

Holographic measurements of the shape and 3D-motion of the surface of a sinusoidally driven normal TM. Adapted with permission from Figs. 4, 5, and 9 of [46]. Copyright 2016, Acoustical Society of America. A A pseudo 3D plot of the TM surface. The supporting ring of the TM was positioned parallel to the x–y plane. The umbo, the most inward position of the membrane, is located at z = 0. The black lines on the TM outer surface illustrate the location of the manubrium. B A rotated view of the same TM with the x–z plane on the surface of the paper, and y is in and out from the surface. The inset shows the definition of a normal vector for one point on the membrane surface. The tangential plane orthogonal to it is defined by the vectors α and β. From the shape measurements, a unique set of parameters was defined for each point on the TM surface. C Color maps of the magnitude of the surface displacements of the TM (in dB re the displacement at the umbo) in the surface normal (η, out-of-plane) and tangential (α and β, in-plane) directions at three stimulus frequencies

Using a 3D-holography system that measured shape and 3D motion at each point on the TM surface nearly simultaneously, Khaleghi et al. [46] demonstrated that motion along the normal surface vector was usually significantly larger than the “in-plane” motions along $\alpha$ and $\beta$, at least across the 0.5 to 8 kHz frequency range of their measurements (Fig. 8C). The exceptions to this rule were small surface regions, usually near the manubrium arm, where the in-plane motion magnitudes were similar or somewhat greater than the normal motions at 4.5 and 8 kHz. These regions had the largest shape gradients and were prone to the largest errors in motion component computations. Another conclusion is that measurements in z are generally reasonable approximations of motion in $\eta$.

Sound-Driven Motions of the Surface of the TM: Clicks and Impulses

The use of continuous pure-tone stimuli makes it difficult to assess the frequency and temporal dependence of TM motions with any detail, as it takes time to measure the response at each frequency. Impulse-like acoustic “click” stimuli represent the sum of multiple frequency terms and can be used to assess the membrane’s frequency dependence more quickly. Such transient stimuli are also more similar to real-world sounds and are better at assessing the temporal relation between the motions of different locations on the membrane surface [59]. (However, the generation of clicks with sufficient low-frequency energy and simple spectral properties at high frequencies can be challenging.) Holographic measurements of the response of the entire membrane’s surface to clicks required the development of very high-speed digital cameras (frame rates > 40,000 frames per second) and accurate high-frequency holographic measurement paradigms [47, 48, 57]. Figure 9 summarizes the response of the TM in an isolated temporal bone to an acoustic click measured by Tang et al. [48]. Figure 9B illustrates the click stimulus, with amplitude of +/– 10 Pa (111 dB peak equivalent SPL), initiated at time 0. After initiation, the absolute value of the transient sound pressure reduces to 30% of the peak in 0.5 ms [48]. Figure 9C illustrates the displacement at five locations on the TM surface (P1 is the umbo) where the frame rate of the high-speed camera was 67,200 frames per second, providing a temporal resolution of 15 μs per frame. These displacement waveforms exhibit location-dependent peak amplitudes and response shapes with slower time courses than the sound pressure stimulus (the time to peak magnitude is about twice that of the pressure stimulus, and the absolute values of the displacements fall to 30% of the peak value in about 1.3 ms). Though the motions at P2 through P5 all begin by about 30 μs after the start of the pressure stimulus and reach their different largest inward motions in about 100 μs, the timing and magnitude of the subsequent motions vary among the positions.

Fig. 9.

Measurements of the surface motion of a TM in response to an acoustic click. Adapted from Fig. 8 of [48]. A A 2D map of the click-induced 3 ms rms displacements at each surface point. B The time waveform of the sound pressure associated with the click. C Time waveforms of the absolute measured displacements at five points on the TM, including the umbo (P1) and the four quadrants of the surface. Inward motions are scaled as positive

Careful observation of the data in Fig. 9C shows that the motion of the umbo (P1, the blue curve) starts about 30 s later than the other positions, reaches its peak near 150 s, and has a smaller motion magnitude. The differences in peak timing, peak magnitudes, and temporal pattern reflect local differences in the mechanics related to TM shape, local TM properties, and location-dependent differences in the effect of the support constraints place by the manubrium and bony annular rim. It is worthwhile noting that the observed differences in delay between the umbo and the other points are well within the temporal accuracy (15 μs/frame) of the high-speed camera.

Figure 9A shows a map of the root-mean-square (rms) displacements averaged over the first 3 ms of the response at more than 40,000 points on the TM surface. Similar to the responses to tones, the larger motion magnitudes occur halfway between the manubrium and the annular rim, where the curvature of the TM is larger (Fig. 8A and 8B), while some of the smallest motion magnitudes occur at the umbo and along the manubrium.

Figure 10 shows the generality of the motion onset delays observed in Fig. 9c by describing the temporal development of the click-induced motion of a different TM’s surface [47], where six successive holographic frames of the surface motion display a time course of 119 μs. (The camera frame rate was 42,000 fps with a time per frame of ~ 24 μs). The first frame (Fig. 10A at time 0) shows the neutral position of the membrane just before the click reaches its surface. The second frame (10B at + 24 μs) shows the developing inward motion surrounding the manubrium and umbo, which appears similar to the rms displacement map of Fig. 9A. The third frame (10 C at + 47 μs) shows the further development of inward motion on the TM surface and the first measurable motion at the umbo. The fourth frame (10D at + 71 μs) shows the maximal inward motion along much of the surface. The fifth (10E at + 95 μs) and sixth (10 F at + 119 μs) show reductions in the inward displacements and the initiation of outward displacements. The timing of these outward movements varies much more in space than the nearly uniform inward motions driven by the click onset. The near-uniform timing of the initial motion of most of the TM surface is inconsistent with suggestions of sound-generating waves on the TM surface that travel from the periphery to the center [102]. Instead, small local variations in timing and magnitude of surface motions likely result from the TM shape and differences in the local mechanical properties of the TM. For example, the extra mass of the bony umbo and manubrium should delay the motion of the TM at those locations [59, 60]. These spatial differences in response to clicks should translate to spatial differences in the response to other acoustic transients, as well as tones and harmonic series.

Fig. 10.

Six color maps of the absolute click-induced displacement of the surface of a TM showing 119 μs of the initial response of the TM to the click (data from [47]). Inward responses are plotted as positive

The temporal variations within the click-driven motions in Figs. 9 and 10, depend in part on the temporal variations within the sound pressure stimulus (Fig. 9B). The effect of these stimulus variations is removed by computing the impulse response (IR) of the motion at each of the more than 40,000 measured points on the TM surface [48]. Definition of the IR allows quantification of multiple stimulus-independent characteristics of the response of the entire membrane to sound (Fig. 11) including (i) the time to peak amplitude, (ii) the exponential decay time—the time it takes the response to decay to an absolute amplitude of 30% of its peak value [48], and (iii) the dominant frequency component of the local response. These characteristics again depend on local and support mechanics. In particular, the damping ratio computed from the ratio of the decay rate (1/decay time) and dominant frequency is an estimate of the local damping within the membrane [48]. Figure 12 plots the damping ratios computed from 40,000 locations on eight cadaveric TMs sorted by local dominant frequencies. The frequency dependence of the measurements in Fig. 12 suggests a mass-governed damping, consistent with the modeling conclusion of Funnell et al. [51]. Others have used a combination of static measurements, point displacement, and time-averaged holographic motion measurements with finite-element (FE) or continuum models for other estimates of damping ratio [50, 51, 61, 62].

Fig. 11.

Color maps of the “time to first peak” in ms, the “decay time” in ms, and the “dominant frequency” in kHz computed from the impulse responses that describe the motion at each point on the surface of one TM. Adapted from Fig. 10 of [48]

Fig. 12.

An estimate of the damping ratio computed from impulse response data in eight normal TMs plotted vs. local dominant frequency. The range of the values at each point describes the damping ratios from regions with that dominant frequency collected across all measured surface points (40,000 points × 8 specimens), and the symbols describe the median of this collection. Redrawn from Fig. 18 of [48]

Model- and Measurement-Based Investigations of the Effect of TM Shape and Structural Properties on TM Function

While many have performed direct stress and strain measurements of the mechanical properties of the TM [63–68] with static or quasi-static stimuli, others have used a combination of dynamic measurements and the FE and continuum models mentioned above to assess important mechanical features of the TM. These include (i) demonstrations that the curvature of the TM affects the magnitude of the surface motions of the TM that result from sound stimulation [41, 49], and (ii) estimates of the material properties of the TM [69, 70]. Additionally, the models of Rabbitt and Holmes [50] and Fay et al. [41, 69] point out that the anisotropic arrangement of collagen fibers in the membrane, in particular the fibers that run radially from the umbo and manubrium to the TM rim, play a significant role in membrane dynamics.

An outcome of these measurements is the manufacture of artificial membranes that mimic the features and measured material properties of the human TM, including shape and fiber arrangements [71–74]. Such studies have the primary goal of better restoring the TM’s normal function after tympanoplasty. While the results of such attempts are promising, they have not come into clinical use.

The Transmission of TM Vibrations to the Cochlea: The Role of the Ossicular Joints

As reviewed by Merchant [103], reconstructions of the ossicular system generally result in at least mild hearing loss, with greater hearing losses resulting from more extensive reconstructions. This section investigates how the ossicular joints, which are not normally reconstructed in ossiculoplasty procedures, contribute to normal and pathological middle-ear function.

The Ossicular Lever and the Ratio of Stapes to Umbo Motion

Modern laser-Doppler vibrometer comparisons of the sound-driven motions of the umbo (at the input to the ossicular chain) and the stapes (at its output) [75, 76] are inconsistent with the simple transformer described in Fig. 2. Indeed, the ratio of the motion of the stapes to the umbo is not well explained by the anatomical “lever-ratio” provided by the lengths of the incus and malleus (Fig. 13A). The lever-ratio model suggests stapes motion should be 3 dB less (a factor of 0.7) than the motion of the umbo, and the three ossicles should move with the same phase at all frequencies. Measurements (Fig. 13B) show the stapes and the tip of the distal incus move at least 10 dB (a factor of 1/3) less than the umbo at low frequencies, with increasing losses and accumulating phase difference at higher frequencies. While some of the differences between measurement and theory might be explained by differences in the alignment of the laser beams and the primary direction of motion of the three ossicles, such an explanation is inconsistent with the frequency dependence and phase of the motion ratio. Instead, the observed motion differences suggest the existence of motion losses and delays between the malleus and the incus and between the incus and the stapes.

Fig. 13.

Measurements and models of the coupled motion of the malleus, incus, and stapes. A A reconstruction of the human ossicular chain [100]. The low-frequency ossicular mode of the rotation of the malleal and incudal lever arms about an axis defined by the anterior-malleal and posterior-incudal ligaments is drawn in red; red dashes illustrate the parts of the axes that are contained within the bones. Definitions of the lever arm lengths of the malleus and incus (the distances between the umbo and incus and a rotary axis) are included along with their ratio in linear and dB terms. B A graph describing the ratio of the magnitudes of the displacements of the inferior end of the incus and the stapes (above) to the displacement of the umbo, and the differences in their phase of motion (below). The data are from Fig. 3 of [76]. Included in red is the predicted difference in magnitude and phase from rotation about the axis described in A

Flexibility Within the Ossicular Joints

Several studies performed between 2002 and 2005 identified and quantified the ossicular motion differences in terms of flexibility of the incudomallear joint (IMJ) and, to a lesser extent, the incudostapedial joint (ISJ). Willi et al. [77] used scanning-laser vibrometry of a multitude of points around the IMJ and along the malleus lever arm to demonstrate 3 to 6 dB rotational losses across the IMJ due to a flexion of the joint at low frequencies, with increasing loss at higher frequencies. Decraemer and Khanna [78] scanned a multitude of points along the superior and posterior surfaces of the three ossicles of several human specimens, using multiple controlled laser angulations, along with precise micro-CT-based anatomical measurements of the individual ossicular geometry to define the 3-dimensional motion of the ossicles in response to tones. Their reconstructed motions demonstrated flexion of both the IMJ and the ISJ. While flexion of these joints had been known to occur with extremely intense sound stimuli (greater than 130 dB SPL) [12, 79], these were the first studies to suggest joint flexion across a broad range of frequencies with low-to-moderate stimulus levels.

Dobrev et al. (101) also investigated losses across the IMJ using 3D-laser vibrometry at multiple measurement points along all three ossicles in two temporal bones to reconstruct the motions across the IMJ, and came to different conclusions. In one bone, they saw a loss of rotational motion of the incus compared to the malleus, but only with frequencies above 1.5 kHz. In the second bone, they saw little difference in the sound-driven rotation of the malleus and incus at frequencies between 0.5 and 4 kHz. Some of the differences in results across the different studies may be explained by the small number of bones in some of the studies and individual differences in the compliance of the IMJ, which may be partially related to age [94].

While there is conflicting evidence that losses in the IMJ reduce the sound-driven motion of the incus, particularly at frequencies below 1 kHz, in normal ears, there is significant evidence that pathologies that limit incus and stapes motions result in flexion of the IMJ. Some of the best evidence is that of Nakajima et al. [80] who demonstrated flexibility in the IMJ using single-point laser measurements of the sound-driven motion of the stapes and umbo with moderate stimulus intensities while using cements and gels to reversibly reduce the mobility of the umbo or stapes (Fig. 14). Their demonstration that cementing the stapes footplate in the oval window produced large (– 50 dB) reductions in sound-driven stapes motion, but only mild (– 10 dB) reductions in umbo motion (Fig. 14B) is consistent with a flexible IMJ and ISJ that allow the umbo to move independently of the stapes. This observation is in line with the common observation that stapes fixation can have little effect on clinical tests of TM mobility, including tympanometry and broadband admittance [19, 20, 81, 82]. As a test of their techniques and interpretation, Nakajima et al. [80] also demonstrated that reducing the mobility of the malleus, by graded cementing of the malleus head to its surrounding bone, produced similar mild to moderate (– 20 dB) losses in both sound-driven umbo and stapes motions. Such fixations alter the input to the IMJ and ISJ without changing their mechanical loads.

Fig. 14.

Measurements of the effect of immobilizing (fixing) the head of either the malleus or the stapes footplate on the sound-induced motion of the umbo and stapes. Data are from Fig. 4 of [80]. The solid lines are means; the shaded areas represent the 90% confidence interval (CI) around the mean

Others have manipulated the flexibility of the IMJ in cadaveric preparations. Farahmand et al. [83] increased the flexibility of the IMJ via small cuts in the joint and saw a reduction in the sound-evoked stapes motion at frequencies above 1 kHz. This high frequency decrease mimicked the isolated high frequency conductive-hearing loss seen in some patients [83, 84]. The observations of decreases in high-frequency stapes motion produced by normally flexible IMJs [75, 76] (Fig. 13) are supported by cadaveric measurements of stapes motion showing improved high-frequency response after stiffening the IMJ [85].

There are other consequences of a flexible IMJ and ISJ. Hüttenbrink [86] used cadaveric temporal bones to demonstrate that large static ear-canal pressures (± 30 to ± 500 daPa) induced large motions of the umbo and malleus, but the motions of the stapes were much smaller. The observed 20 dB reductions in stapes motion relative to the umbo can help protect the inner ear from sudden pressure changes produced during diving, flying, and sneezing. At the other end of the frequency spectrum, Gottleib et al. [87] demonstrated that the low-pass filtering introduced by the ossicular joints (Fig. 10) acts to reduce the sharpness of environmental transients (such as thunder or other blasts) that reach the inner ear. Finally, ossicular joint flexibility is useful in cases of stapes replacement, where the flexible IMJ protects the stapes prostheses from large displacements of the TM caused by sneezing and swallowing [88].

Multiple Modes of Ossicular Motion

Some of the frequency dependence observed in measurements of sound-driven stapes motion can result from frequency-dependent variations in the modes of ossicular motion, for just as TM motion can be described by the sum of multiple frequency-dependent modal motions, so can the motion of the ossicular chain [78, 89, 90]. The mode that dominates low-frequency ossicular motion is consistent with the ossicular lever (Fig. 15A) that rotates about an axis that runs from the posterior incudal ligament to the anterior mallear ligament [4]. Because of this rotation, the umbo and stapes move in phase, and the stapes moves less than the umbo. However, at frequencies above 0.6 kHz, other modes of motion appear, such as rotation of the incus and malleus about an axis that runs superior to inferior (Fig. 15B). In this second mode, the stapes and umbo are ½ cycle out-of-phase, and the stapes may move more than the umbo depending on the precise location of the rotational axis. As Decraemer and Khanna [78] demonstrated, the number of ossicular motion modes is large. These frequency-dependent changes in ossicular motion mode introduce changes in the transfer of motion magnitude and phase between the umbo and the stapes that can extend the high-frequency limit of middle-ear sound transfer [89, 91].

Fig. 15.

Two hypothetical modes of ossicular motion that predict different relationships between the relative magnitudes and phases of the umbo and stapes motion. The axes of rotation and lever arms are drawn as in Fig. 13. The volume rendered ossicles are modified after [100]. A The low-frequency ossicular motion mode is an ossicular lever with a rotational axis that runs from the anterior mallear ligament (1) to the posterior incudal ligament (2). The mallear lever arm (m) is about 40% longer than the incudal lever arm. Both arms are on the same side of the rotational axis, so the malleus and incus move inwards and outwards with the same phase, with the stapes moving less than the umbo. B A hypothetical higher-frequency motion mode, with a rotational axis that runs through the IMJ from superior (1) to inferior (2). The relative lengths of the lever arms depend on the precise location of the rotational axis. In this illustration, the incudal lever arm (i) is about twice the length of the mallear lever arm (m). The two lever arms are on opposite sides of the rotational axis, so when the umbo moves inward, the stapes moves outward (in opposite phase) with about twice the umbo amplitude

Summary and Conclusions

Sound conduction through the middle ear is a multifaceted process involving all the components of the middle ear. Many of these components are well studied, but further investigation is required. Our review indicates:

1. The cadaveric human temporal bone is a useful tool for investigating the mechanical and acoustical function of the human middle [9–11].

2. Modern surgical procedures for the repair of the TM (tympanoplasty) and ossicular chain (ossiculoplasty) do not result in normal hearing, especially at high and extended high frequencies [5, 52, 53, 103].

3. The sound-driven motions of the TM result from the sum of multiple modal patterns [45, 54, 55] that depend on the driven frequency where the continued contribution of the more uniform low-order motion modes at higher frequencies help simplify the motion at the umbo and the input to the rest of the middle ear [45].

4. With impulsive stimulation, the near-simultaneous motion of much of the TM is inconsistent with models that suggest motion waves move from the periphery to the center of the TM. The small delay between the near simultaneous responses of the center of the TM and those parts of the TM adjacent to the umbo and manubrium of the malleus probably reflects the additional load of these structures on the local TM surface [47, 48, 59, 60].

5. The combination of TM motion data and modeling work can define the mechanical properties of the TM and its supporting structure, and efforts to produce life-like artificial TMs for TM repair are promising, but not yet mature enough for clinical adaptation [41, 48–51, 61, 62, 71–74].

6. The sound-driven motions of the ossicles are not explained by simple rigid body rotation of the malleus and incus, and stapes, as the flexible ossicular joints introduce phase delays and losses in motion magnitude between the malleus and stapes [76–78, 101]. While it is indisputable that such losses occur at frequencies above 1 kHz, the studies that define the 3D-motion of the ossicles [77, 78, 101] differ in whether these losses occur at frequencies below 1 kHz. Complications in these comparisons include: the small number of specimens that have been measured in some of the better studies [78, 101], significant individual variation in coupled ossicular motions [77], as well as significant, possibly age-related, variation in the structure of the IMJ [94].

7. The losses of motion across the ossicular joints decouple the motion of the TM and malleus across a broad frequency range from the effects of distal ossicular pathology such as stapes fixation. Clinically, near normal TM motion in the presence of a large conductive hearing loss is a sign of stapes fixation [80–82].

8. Modulation of IMJ stiffness can affect hearing at frequencies above 2 kHz. Loosening the IMJ produces decreases in sound-evoked stapes motion, while stiffening the joint produces increases [83–85]. The observation of stapes motion increasing after stiffening the IMJ [85] adds support to the idea that the normal joint produces a loss in incus and stapes motion.

9. Changes in the modes of ossicular motion introduce frequency-dependent changes in motion and phase along the ossicular chain that contribute to the overall transfer of sound energy from the ear canal to the cochlea, especially at frequencies above a few kHz [89–91]. The significance of these higher-order modes to ossicular reconstruction is not clear, though they suggest that procedures that restore as much of the natural structure as possible, including the joints, may better restore high-frequency ossicular function.

Future Work

Most surgical procedures for the repair of the TM (tympanoplasty) and ossicular chain do not result in normal hearing, especially at high and extended high frequencies [52, 53]. TM replacement procedures could be improved if we better understood the influence of different TM structural features and developed more TM-like replacement grafts. A better understanding of ossicle and joint form and function, as well as a clearer understanding of ossicular modes of motion, could improve ossiculoplasty, and especially help post-surgical high-frequency hearing results. It is also crucial to develop better methods for identifying pathology and surgical techniques that improve high-frequency hearing. An alternative to improvements in surgical reconstruction is the development of effective active middle-ear implantable devices that mimic normal TM and ossicular function over a broad frequency range [92].

The need for improvements in tympanoplasty and ossiculoplasty, especially for improved high-frequency performance, in part, comes from the growing evidence for a conductive component to presbycusis. There is new evidence that the ossicular joints age and introduce high-frequency conductive hearing losses that can be treated by surgery [93, 94]. To better quantify and understand the prevalence of age-related high-frequency conductive hearing loss, research groups at several institutions (including the MEE) are investigating improvements in bone-conduction hearing tests to better define conductive hearing loss at frequencies of 4 to 16 kHz [95].

Another issue that is garnering much interest is to improve our knowledge of the middle ear’s response to high-intensity sound. It is known that the middle ear begins to act nonlinearly with sound stimuli above 125 dB SPL, where evidence of compressive growth of middle-ear output at such high levels has suggested this behavior can protect the inner ear from more intense sounds [96–98]. What is the source of these nonlinearities? and How do they affect middle-ear sound transfer? At the extreme, impulsive sounds of 170 dB SPL or more, such as blasts, can cause catastrophic damage to the middle ear, such as TM rupture and interruption of the ossicular joints [99]. What are the stresses and strains on the TM and ossicles that lead to such damage? Are there “weak spots” in the TM and ossicular chain that are predisposed to damage? Does such damage to the middle ear protect the inner ear? If not, can damage be prevented by pre-treatment?

Funding

The work included in this review that was performed at the MEE and WPI was funded by grants from the NIDCD to Jeffrey Tao Cheng, Cosme Furlong, Saumil Merchant, Heidi Nakajima, William Peake, Aaron Remenschneider, and me. The grant numbers can be found in the original published articles. Artificial Intelligence was not used in preparing this manuscript, including data analysis and figure preparation.

Data Availability

No new data are presented in this review. Requests for data should be made to the authors of the original published articles.

Declarations

Conflict of Interest

While I, the author, have no active conflict of interest in any of this work, I am named on patents for the development of a holography system for the diagnosis of eardrum pathology, and a patent for 3-D print material for the repair of ruptured eardrums.