Bone-conduction hyperacusis induced by superior canal dehiscence in human: the underlying mechanism

Our ability to hear through bone conduction (BC) has long been recognized, but the underlying mechanism is poorly understood. Why certain perturbations affect BC hearing is also unclear. An example is BC hyperacusis (hypersensitive BC hearing)—an unnerving symptom experienced by patients with superior canal dehiscence (SCD). We measured BC-evoked sound pressures in scala vestibuli (PSV) and scala tympani (PST) at the basal cochlea in cadaveric human ears, and estimated hearing by the cochlear input drive (PDIFF = PSV – PST) before and after creating an SCD. Consistent with clinical audiograms, SCD increased BC-driven PDIFF below 1 kHz. However, SCD affected the individual scalae pressures in unexpected ways: SCD increased PSV below 1 kHz, but had little effect on PST. These new findings are inconsistent with the inner-ear compression mechanism that some have used to explain BC hyperacusis. We developed a computational BC model based on the inner-ear fluid-inertia mechanism, and the simulated effects of SCD were similar to the experimental findings. This experimental-modeling study suggests that (1) inner-ear fluid inertia is an important mechanism for BC hearing, and (2) SCD facilitates the flow of sound volume velocity through the cochlear partition at low frequencies, resulting in BC hyperacusis.


Intracochlear pressures during AC stimulation in temporal bone experiments.
For the temporal bone experiments studying the effects of SCD on BC (results shown in Fig. 2 of main manuscript), AC pressure responses were also measured for normal condition and after SCD. The effects of SCD on intracochlear pressures during AC were consistent to earlier reports 1-3 : SCD resulted in P SV , P ST and P DIFF decreasing at low frequencies (from thin solid to thin dashed lines in Fig. S1). Figure S1. Magnitude and phase of (A) P SV and (B) P ST at the cochlear base normalized to the volume velocity of stapes footplate (U STAP ) during AC stimulation. The normalized differential pressure or the cochlear drive (P DIFF /U STAP ) is shown in (C). Pressures measured from individual ears are designated by color, with thin solid lines representing normal (before SCD) and thin dashed lines representing after SCD. The thick black lines represent the medians of the normal inner ear; the dashed black lines represent the medians after SCD. In (B) and (C), the thicker orange lines represent fitted impedances of RW (Z RW ) and cochlear partition (Z DIFF ) of the median curves of normal ears by using acoustic components and the parameters in the column 'Median Z' in Table S1. The thicker blue and red lines represent the fitted low and high boundaries, respectively, of the variations in Z RW and Z DIFF . The parameters used to fit the low and high boundaries of the impedance variations are listed in the columns 'Minimum Z' and 'Maximum Z' in Table S1. In this manner, the median impedance parameters for use in the computational model and the ranges of those parameters were experimentally determined.
Determining impedances of the inner ear from AC intracochlear pressure measurements. During AC stimulation, the effect of SCD on inner-ear sound transmission can be simulated by a lumped-element model with experimentally determined impedances from temporal bone experiments. Figure S2 shows the AC inner-ear model developed for AC sound transmission in the inner ear for normal and SCD conditions 1,4-7 . For a normal ear (switch open in Fig. S2), the volume velocity generated at the stapes footplate (U STAP ) flows through the cochlear partition (with impedance of Z DIFF which includes the helicotrema) and exits the inner ear at the round window (with impedance of Z RW ). The sound pressures measured near the oval and round windows are P SV and P ST , respectively. For an ear with SCD, the superior canal forms a 'thirdwindow' pathway (switch closed in Fig. S2). With SCD, part of U STAP is shunted towards the superior semicircular canal (with impedance of Z SC ) and exits at the dehiscence, while the rest of U STAP goes towards Z DIFF . Figure S2. AC inner-ear model: Schematic of the lumped-element model for sound transmission in the inner ear during AC stimulation. Closing the switch represents an SCD.
In the normal inner ear of temporal bone experimental measurements: Thus, the solid lines in Fig. S1B are Z RW and the solid lines in Fig. S1C are Z DIFF of our temporal bones.
In the SCD condition, impedance of the superior canal Z SC is determined by Figure S3 below plots Z SC of individual ears -obtained from the measurements -with thin colored lines and their medians with thick black lines. Figure S3. Magnitude and phase of the acoustic impedance of the superior semicircular canal with SCD (Z SC ). Thin colored lines represent the impedances obtained from the measurements of individual ears and the thick black lines represent the median. The thick orange line represents the fitted median of Z SC using the parameter in the column "Median Z" in Table S1 . The thick blue and red lines represent the fitted low and high boundaries of the variations in Z SC , respectively . The parameters used to fit the low and high boundaries are listed in the columns 'Minimum Z' and 'Maximum Z' in Table S1, respectively.
To determine representative impedance parameters Z RW , Z DIFF , and Z SC , we fitted the median of Z RW , Z DIFF , and Z SC of the individual ears from the present study using acoustic elements (acoustic compliance, mass, and resistance), as in our previous studies 1,4,6 and others 8,9 . The values of those components are provided in the column 'Median Z' in Table S1 below. In addition, representative parameters of the reverse middle-ear impedance at the oval window, Z ME , and the impedance of the vestibular aqueduct, Z VApreviously derived and published by our group 4,10are also provided in the column 'Median Z' in Table S1. These five experimentally derived impedances are used to test two types of BC models: the BC inner-ear compression model described below in this supplementary information and the BC inner-ear fluid-inertia model described in the main article.
The fitted median impedances are plotted as thick orange lines in Figs. S1B (Z RW ), S1C (Z DIFF ), and S3 (Z SC ). Note that the median of the measured Z RW in Fig. S1B does not correctly reflect the sharp resonant compliance-to-mass transition seen in the individual ears because the frequency of this transition varies across ears. The fitted Z RW has a relatively sharp transition, more representative of the individual measurements.
The inflection at 400 Hz seen in the median of Z SC results from the Z SC variations between the individual ears. Two out of six ears showed this inflection in Z SC but the other four ears did not (Supplemental Fig. S3). Thus the low-frequency inflection is not a common feature of Z SC . The most prominent pattern of Z SC below 4000 Hz is an increase in magnitude with frequency at a rate of 20 dB/decade accompanied by a phase of + 0.25 cycle (i.e. positive 90 degree). This frequency dependence in our experimental results suggests that Z SC resembles an acoustic mass, which is very likely contributed by the mass of the fluid in the superior canal. Thus, we used a single inertance component, L SC , to fit the median of Z SC ( ).
The frequency dependence of Z SC at f > 4000 Hz varies between ears. This variation may reflect complications in our estimates of stapes motion. The volume velocity of the stapes (U STAP_Normal ) used in the calculation of Z SC (Eq. S3) is estimated from a single point 1D LDV measurement, but the motion of the stapes at the high frequencies in AC can be complex and three dimensional.
Thus it is likely that there are errors in the measurement-derived Z SC above 4000 Hz. Given the individual variations and the potential errors, we did not attempt to accurately fit Z SC above 4000 Hz.
We also fitted the low and high boundaries of the inter-individual variations in those impedances, and the corresponding parameters are provided in columns 'Minimum Z' and 'Maximum Z', respectively, in Table S1. Those values were used to study the effects of parameter variation on the BC inertia model, and the relevant results are provided in the section 'Effects of parameter variation on the BC inertia model' below.

Simulate effects of SCD on AC-evoked intracochlear pressures
We used the AC inner-ear model of Fig. S2 and the median impedance parameters Z RW , Z DIFF , and Z SC described in Table S1 to simulate P SV /U STAP , P ST /U STAP , and P DIFF /U STAP with and without SCD in Fig. S4. The simulations for the normal inner ear are plotted as solid black lines; the simulations of an ear with SCD are plotted with dashed black lines. These predictions are similar to our temporal bone experimental results plotted in Fig. S1. It should be noted that the AC model in Fig. S2 does not include any natural inner-ear leakage channels such as the vestibular aqueduct and cochlear aqueduct. We assume that the acoustic impedances of those inner-ear leakage channels are too high to greatly influence the sound transmission in the inner ear during AC over the tested frequency range. This assumption is based on evidence of equal volume velocity for round and oval windows during AC 11 .

Bone Conduction
Temporal bone intracochlear sound pressures during BC. Figure S5. Magnitude frequency response of P SV , P ST , and P DIFF measured in 6 ears in response to BC stimulation generated by a Baha BP100 driven by a sinusoidal voltage having a magnitude of 40 mV RMS . Black lines plot the pressures under the normal condition; red lines plot the pressures after SCD; grey dashed lines plot the noise floors of the pressure sensors.

Outer and middle-ear BC mechanisms
The outer ear can contribute to BC hearing by expansion and compression of the compliant cartilaginous wall of the lateral part of the ear canal wall. The middle ear can contribute to BC hearing by the inertia of the ossicles 12 . The sound produced by these two BC mechanisms enters the inner ear via a pathway similar to that during AC. If SCD affects either of these BC mechanisms, SCD should then have similar effects for AC and BC sound transmission. However, SCD induces almost opposite effects for AC and BC: AC hearing decreases and BC hearing increases at low frequencies. Our intracochlear pressure measurements are consistent with these clinical findings. Therefore, the effect of SCD on BC sound transmission does not act through the outer-or middle-ear BC mechanisms.

Inner-ear bone compression model for BC.
Expansion and compression of the bony wall of the inner ear has been proposed as a mechanism for BC 7,13,14 . This concept is illustrated in Fig. S6A. In the normal ear, the bone vibration is believed to result in compression, squeezing the cochlear fluid towards the oval and the round windows (arrows in Fig. S6A), and then expanding. This would generate sound pressures near the oval window (P SV ) and round window (P ST ). Previously, based on this mechanism, we built a BC bone-compression model that resulted in predictions consistent with an increase in P DIFF associated with SCD 7 . But that model was not tested against the individual effects on P SV and P ST .
To quantitatively determine the SCD effects on P SV and P ST under the bone-compression mechanism, we re-created the BC bone-compression model as shown in Fig. S6B.
The two volume velocity sources, U SV and U ST , represent the volume velocities generated by the change in SV space and the ST space, respectively, during volume compression and expansion.
The impedances in this model are the same as those obtained from AC experimental data described above and in supplementary Table S1. Because U SV and U ST during bone compression are unknown, we try different ratios of U SV /U ST (from 0.1 to 10), to test if the simulated effects of SCD fit our experimental results.  With the bone-compression model, SCD can increase P DIFF at low frequencies (if U SV /U ST ≤ ¼, as shown in Fig. S7A). However, SCD decreases both P SV and P ST at low frequencies. This is inconsistent with temporal bone experimental results where SCD increases P SV at low frequencies and has limited effect on P ST , as shown in Fig. 2 in the main article. Thus, we conclude that the inner-ear wall compression mechanism is inconsistent with the SCD effects on BC sound transmission.

Effects of parameter variation on the BC inertia model
The Supplemental Figs. S8-S12 below show the simulated changes in P SV , P ST , and P DIFF due to SCD in the BC inertia model (Fig. 3C) as the impedances in the model are varied, one at a time, from the minimum to the maximum (see Supplemental Table S1) while the other impedances were held at their median or average values. Figure S8. Simulated changes in P SV , P ST , and P DIFF due to SCD in the BC inertia model when Z SC is set at the minimum, median, and maximum values listed in Supplemental Table S1. Figure S9. Simulated changes in P SV , P ST , and P DIFF due to SCD in the BC inertia model when Z VA is set at the minimum, median, and maximum values listed in Supplemental Table S1. Figure S10. Simulated changes in P SV , P ST , and P DIFF due to SCD in the BC inertia model when Z DIFF is set at the minimum, median, and maximum values listed in Supplemental Table S1. Figure S11. Simulated changes in P SV , P ST , and P DIFF due to SCD in the BC inertia model when Z ME is set at the minimum, median, and maximum values listed in Supplemental Table S1. Figure S12. Simulated changes in P SV , P ST , and P DIFF due to SCD in the BC inertia model when Z RW is set at the minimum, median, and maximum values listed in Supplemental Table S1. Figure S13. Structure based acoustic circuit representing a superior semicircular canal with an SCD. Z T represents the termination impedance interfacing the SCD. Figure S13 shows an acoustic circuit of the superior canal with an SCD. The two arms of the superior canal and the SCD are treated as tubes of small diameters. The resistors (R SC ) and inductors (L SC ) represent the acoustic resistance and mass associated with the water-like lymph in the tube, with values that depend on the length and diameter of the tube (Eq. 5.48 and 5.49 in Beranek). Z T is the termination impedance at the SCD interface. If the SCD interfaces the brain, Z T represents the impedance of the fluid-and tissue-filled cranial vault; if the SCD interfaces the air-filled middle ear cavity, Z T represents the impedance of the air in the cavity. Table S2 summarizes the parameters used to calculate the theoretical impedance of the superior canal (Z SC_Theo ), the theoretical impedance of a circular SCD of various diameters from 0.1 mm to 1 mm (Z SCD_Theo ), and the termination impedance (Z T ) for the two conditions -the middle ear cavity and the cranial vault.

Prediction of the effect of SCD size and location on BC hearing
The total impedance of the superior canal branch, Z SC_Total , is determined by Table S2. Dimensions and equations used to calculate the impedances in Figure S13.  Figure S14 plots Z SC_Total with different sizes of SCD and different terminating conditions. When the dehiscence is extremely small (diameter of 0.1 mm), Z SC_Total is relatively high and is dominated by Z SCD_Theo . As the size of the dehiscence increases, Z SCD_Theo decreases and so does Z SC_Total . When the diameter of SCD ≥ 0.4 mm, Z SC_Total is dominated by Z SC_Theo , and changes little with further increases in the size of the SCD.
When the diameter of the dehiscence ≤ 0.2 mm, the Z SC_Total is insensitive to Z T because Z SC_Total is dominated by Z SCD_Theo . For SCDs of diameters ≥ 0.4 mm when the dehiscence is terminated by the cranial vault, Z SC_Total are smaller at f < 150 Hz, due to a resonance between Z SC_Theo and Z T . Figure S15. Simulated effects of SCD size on the changes of P SV , P ST , and P DIFF in the BC inertia model when the SCD faces the cranial vault.
To study the effects of SCD size and the different termination conditions on BC-elicited intracochlear pressures, we replaced Z SC with Z SC_Total in the BC inertia model in Fig. 3C. Figure  S15 shows the predicted changes of P SV , P ST , and P DIFF with varying diameter of the SCD when the dehiscence faces the cranial vault. As the size of SCD increases, the low-frequency increases in P SV , P ST , and P DIFF become more prominent. However, when the SCD diameter is 0.4 mm or greater, the changes in P SV , P ST , and P DIFF are insensitive to the size of the dehiscence.
When we change the termination impedance from the cranial space to the middle ear cavity, the changes in P SV , P ST , and P DIFF for a given size of the SCD remain the same (therefore simulation results are not shown). The model suggests that BC hearing would increase by the same amount due to the SCD regardless of which space the SCD interfaces. Figure S16. Flow chart of the experimental protocol.