Development of the Surgical Implantation and Fixation of an Implanted Middle Ear Microphone, the “UmboMic,” in Cadaveric Sheep

Abstract

Purpose

The UmboMic is a recently developed middle ear microphone that functions by detecting the sound-induced motion of the umbo [1]. To prepare for a live animal study of the in vivo performance of the UmboMic, we developed the surgical implantation and fixation system in cadaveric sheep ears.

Methods

Temporal bones from seven female sheep were prepared for implantation of the UmboMic system. Fixation hardware for sheep was designed to secure the UmboMic sensor in position so that the sensing tip contacted the umbo in the middle-ear cavity and the non-sensing tail end was secured to the surrounding mastoid bone. First, pre-surgical micro-CT scans of the temporal bones and open-source software were used to simulate surgical drilling, and to virtually plan and fit the UmboMic fixation system. Then, in physical sheep temporal bones, sound-induced umbo motion was measured with a laser Doppler vibrometer as a preliminary step. After surgical implantation of the UmboMic system, the position of the UmboMic sensor was evaluated with microscopic visualization and post-surgical micro-CT. Implanted UmboMic sensor function in response to acoustic frequency sweeps to the external ear canal was measured in two specimens.

Results

Measurements of sheep temporal bone anatomical dimensions with micro-CT showed that the facial recess featured a mean maximum height of 2.36 mm, length of 12.2 mm, and depth from outer opening surface of facial recess to the umbo of 11.1 mm. The amplitude of umbo motion prior to UmboMic implantation was consistent with that previously reported by [2], with a displacement of ~ 14 nm/Pa up to a frequency of 5.5 kHz, followed by a gradual drop-off. When bench tested and implanted in sheep temporal bones, the UmboMic sensitivity was as predicted from [1], ~ 1–2 fC/nm.

Conclusion

When implanted in sheep cadaveric ears, UmboMic performance was similar to previous measurements in fresh human cadaveric temporal bones. The method of simulating surgical drilling using micro-CT and simple open-source software is economical and broadly applicable in understanding the 3D biological anatomy and implantable device design and customization. The UmboMic implantation in cadaveric sheep ears was possible, and this study represents a step towards planned live animal studies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s10162-026-01056-w.

Introduction

Cochlear implants can restore hearing to partially and fully deaf people. Over one million cochlear implants have been implanted worldwide [1]. In addition to the implanted electrode and receiver, a conventional cochlear implant requires external components, including a microphone, speech processor, and transmitter. There are several drawbacks of the conventional external microphone, which is anchored behind the pinna and senses sound near the tragus. These include sensing noise when touching the microphone area, noise from wind, interference with protective and other head coverings, inability to be used while sleeping or during some physical activities, discomfort, possibility of becoming lost or broken, and cosmetic concerns [2]. The implanted magnet used to attach the external component can interfere with MRI imaging and cause discomfort or pressure injury [3]. A totally implantable cochlear implant (TICI), with all cochlear implant components internalized, would provide benefit to the cochlear implant user as long as there is not a reduction in sound quality, functionality or durability. Several strategies have been attempted for the implantable microphone. The Acclaim TICI by Envoy Medical (similar to the Esteem device for use with an active middle ear implant) employs an implantable microphone that senses incus body or malleus head motion and is undergoing clinical trials [3]. The Mi2000 TICI by Med-El uses an implantable subcutaneous microphone and is also currently under investigation in a clinical trial [4]. Similarly, Cochlear Ltd. has developed a subcutaneous microphone placed behind the pinna; the TICI G2 [5] is undergoing clinical trials and the Carina system [6] interfaces with active middle ear implants. Implantable microphones with performance competitive with conventional hearing-aid external microphones are still in active development [3, 7, 8].

The UmboMic is based around a bimorphic cantilever sensor made of polyvinylidene fluoride (PVDF), a piezoelectric polymer [9, 10]. The UmboMic senses displacement of the tip of the manubrium of the malleus, at the umbo of the pars tensa, and translates this motion into electrical charge, which is converted to voltage by a custom low-noise amplifier [9]. The placement of the microphone within the middle ear makes use of the natural pressure amplification of the external ear, and the umbo displays the largest unidirectional displacement in the ossicular chain [11, 12].

The sheep has been used as a large animal model whose ear approximates the human middle and inner ear structure [13–17]. Our planned future live sheep study with implantation of the UmboMic and sound stimulation will be important in evaluating noise issues in the middle-ear cavity and determining how stably an implanted UmboMic operates in live sheep. In the present study we prepared for live animal studies by testing the function of the UmboMic in cadaveric sheep temporal bones. To design appropriately sized fixation hardware that secures the UmboMic sensor and to predict the fit in each sheep temporal bone sample, we initially used a simulation approach. For this, pre-surgical and virtually drilled micro-CT scans of the temporal bones were viewed with an open-source 3-dimensional (3D) image processing software (3D Slicer (https://slicer.org). This virtually drilled model was then exported as a stereolithography (STL) file to a computer-aided design (CAD) software (SolidWorks, Waltham MA). SolidWorks was used to both design the fixation hardware and to fit the hardware into the virtually drilled model. Then, we 3D-printed the fixation hardware and implanted the UmboMic system into physical cadaveric sheep specimens. Correct placement of the UmboMic system including the contact between the sensor tip and umbo was checked both by microscopic visualization and post-operative micro-CT. Function of the implanted sensor was verified by measuring sensitivity and comparing against expected values. Further planning for a live animal study is underway and described in the discussion. The UmboMic used here is a functional prototype, still under technical development [18, 19].

Methods and Materials

Temporal Bone Preparation

The sheep ears were obtained from previously deceased animals from May Family Enterprises, Inc., Manns Choice, Pennsylvania. No animals were killed for the purposes of this research and therefore institutional animal use approval was not applicable. The sheep had been pre-screened for absence of Coxiella burnetii, the cause of Q fever. All sheep were female between 4 and 7 years old. Six were of the Hampshire breed and one was Suffolk. The specimens were previously frozen without use of preservatives. Sheep and human temporal bones have similar anatomy [17]. In a previous study comparing Hampshire and Suffolk-Dorset sheep, Hampshire specimens had better round window membrane visibility, and Hampshire may be a preferred breed for studies related to cochlear implantation [14]. Temporal bones were extracted from the sheep heads with an oscillating saw [13]. The pinna and cartilaginous external ear canal (EC) were transected, and surrounding soft tissue was removed from the temporal bone specimen. Mastoidectomy was performed; then the middle ear cavity was accessed through the facial recess. The facial recess was extended by sacrificing the chorda tympani nerve to increase visibility. Within the middle-ear cavity, excess mucosal tissue was cleared to expose the round window and malleus head, manubrium, and umbo. Imaging of these temporal bones was used for the virtual studies and the physical temporal bones were implanted with the UmboMic system.

When preparing for measurements of umbo motion and sensor operation, a 5-mm hole was drilled into the bony canal to allow for presentation of the speaker and microphone, because the laser beam for measuring umbo motion was occupying the natural EC. In future live animal studies, the sound will be delivered and measured open-field, and a surgical opening to the ear canal will not be made. If there were difficulties in visualizing the umbo due to low light or physical constraints, the bony EC could be drilled down by 2–3 mm.

Sheep numbers (and R,L for right, left temporal bones) are in part used for lab record keeping; for example, sheep 1–8 were used in previous studies [13, 14]. The specific temporal bone numbers used in the different studies are noted in the figures and the table.

Virtual Study with CT Scan and Surgical Simulation

To obtain 512-slice CT scans of the temporal bones (Fig. 1a), the Perkin-Elmer Quantum FX micro-CT scanner at 90-kV voltage and 200-µA current with a 73-mm field of view and 148-μm isotropic resolution was used. Pre-drilled images (5 sheep, 10 temporal bones), post-drilled images (9 temporal bones), and post-implantation images (6 temporal bones) were taken.

Fig. 1.

Schematic diagram of surgical simulation. a Micro-CT scan of extracted temporal bone 15L. b An example of a down-sampled slice (note reduced resolution compared to a), which was then segmented to identify critical anatomical landmarks (yellow: middle ear ossicles; blue: facial nerve, green: tympanic membrane; purple: osseous labyrinth; beige: bone). c 3D rendering of the segmentation (left) and virtually drilled bone (right) using the eraser tool in 3D Slicer. Supplementary Video 1 demonstrates an example of virtual drilling

Bone and other critical anatomical landmarks for implantation including the osseous labyrinth, middle ear ossicles, tympanic membrane, chorda tympani, and facial nerve were segmented from pre-surgery micro-CT scans using 3D Slicer (Fig. 1b). To ensure smooth simulation of virtual drilling, the pre-surgery micro-CT scans were down-sampled to an isotropic resolution of 296 μm.

Virtual surgical drilling of the temporal bone was performed using the eraser tool on the 3D-rendered bone within the Segment Editor Module of 3D Slicer. A demonstration of the virtual drilling process is provided in Supplementary Video 1. The virtually drilled bone (Fig. 1c) was then exported as an STL file to SolidWorks, to be used in designing and fitting the fixation hardware.

Design of the UmboMic Fixation System with CAD

To secure the UmboMic sensor in place, fixation hardware with a ball-and-socket mechanism was designed in SolidWorks. The fixation hardware was then 3D printed in stainless steel by an external vendor (Sculpteo Inc., Cambridge, MA). Stainless steel is a common material for medical implants [20]. For the 3D-printed pieces, stainless steel provided a smoother surface than titanium.

A labeled CAD drawing is in Fig. 2a, b. The sensor’s triangular sensing region extending beyond the “sensor plate,” is in contact with the umbo (without adhesive) and is able to bend. Sensor fabrication and characterization were detailed in previous papers [9, 18, 19]. The sensor tail is glued to the sensor plate of the fixation hardware (red) with J-B Weld “Steel” epoxy adhesive. The “sensor plate” is connected to an arm terminating in a ball “ball-arm”. The ball fits into a socket connected via the “socket arm” to a “screw plate” that is screwed to the bone of the mastoid cortex. The ball-and-socket design allows for some freedom of motion for positioning the sensor. In our original design of the fixation hardware, the screw plate was fixed to the external ear canal wall (Supplemental Figure 1). However, the mastoid cortex was ultimately selected as the fixation site due to its thick, dense cortical structure, which provides reliable screw-anchorage and mechanical stability. It is accessible via a standard mastoidectomy and offers a relatively flat surface that facilitates device positioning. After surgically preparing the cadaveric specimen for facial recess access to the middle ear cavity, specimens were implanted with the UmboMic sensor and fixation system as described below. The sensor plate was 0.5 mm thick and 3.9 mm wide, and the sensor was 0.3 mm thick and 4 mm wide. The socket arm was made in three different lengths (6, 8, and 10 mm) to accommodate individual variation in the distance between the umbo and the mastoid cortex. As shown in Supplemental Video 2, before implantation, socket arms of various lengths were virtually tested in SolidWorks to choose the best-fitting length for each temporal bone. The development and implementation of the fixation system described here prepare for the sheep study; lessons learned through the process will inform the human fixation system under development in our lab.

Fig. 2.

a Virtual implementation of the sensor attached to the plate of the fixation hardware in SolidWorks. b Virtual implementation demonstrating the socket attached to the ball, and the screw plate secured into the mastoid cortex

Surgical Implantation of the UmboMic System in Cadaveric Sheep Temporal Bones

The virtual study with 3D Slicer and SolidWorks software was used to predict a successful drilling path and to decide on the length of the socket arm for the physical study. Following these virtual steps, actual sheep temporal bones were drilled and implanted with the UmboMic sensor and fixation hardware. The virtual study was helpful in predicting the drilling path, and the virtually predicted socket arm length proved correct in the actual temporal bones. Two titanium screws were used to secure the screw plate of the fixation hardware into the mastoid cortex posteriorly. Following adjustment so that the sensor interfaced the umbo, the ball and socket were cemented together using UV-curing dental cement (NX3 Light-Cure from Kerr). The cure time was ~ 20 s. After implanting the sensor with fixation hardware, sensor placement was confirmed visually as well as with repeat micro-CT images (Fig. 5c). In preparing for the live animal study, accelerated aging of a cemented ball and socket was performed by soaking in saline at 67 °C. The aging factor is described by a 10-degree rule as $f=2\left[\frac{T-T_{\mathit{ref}}}{{10}^{\circ }C}\right]$ [21, 22]. For example, with T_ref = 37 °C, 60 days of soaking corresponds to approximately 480 days of aging.

Fig. 5.

a Implementation of the sensor with the fixation hardware showing tip contacting the umbo as seen with microscopic visualization. b Labeled photograph of sensor with fixation system, relating to a. c Sensor tip contacting the umbo as seen in a slice of micro-CT as indicated by the blue arrow, and inset provides further identification. Temporal bone #15L is imaged

Stimulus Generation and Data Acquisition

To measure umbo and sensor displacement, a laser Doppler vibrometer (LDV, OFV-534 and OFV-5000 VD06; Polytec, Waldbronn, Germany) beam was directed through the external EC. To present sound and measure ear-canal sound without interfering with the laser beam, an artificial opening was hand-drilled into the bony external EC to fit a tube that connected to a RadioShack speaker and a Sokolich microphone, as described in temporal bone preparation above. The stimulus was typically 40 single frequency tones of 1 s duration, presented at the EC from 400 Hz to 20 kHz at levels from 60 to 90 dB SPL (sound pressure level, dB re 20 μPa). Signal generation and acquisition were performed with a Tucker-Davis Technologies (TDT) System 3 operating at a 130 kHz sampling rate. Matlab (Mathworks, Natick MA) programming was used to interface with the TDT and for analysis. UmboMic sensor output was measured with the same stimulus protocol as for motion measurements. The UmboMic sensors and a low-noise amplifier were fabricated at the Massachusetts Institute of Technology and characterized at Massachusetts Eye and Ear [9, 18, 19].

LDV and sensor response data were Fourier transformed to the frequency domain. The response at each frequency was deemed significant (out of the noise) if its amplitude was three standard deviations above the mean of the noise floor measured in typically ten neighboring frequency bins. With 1-s signal duration, the noise found from the Fourier transform is noise per root-Hz. To convert this to noise in a 1/3 octave band (to present the noise as is standard for hearing aid microphones) the directly measured noise was multiplied by the square root of 0.23*frequency.

Results

Use of Specimens

The sheep heads were purchased and processed sequentially, and some were purchased and used after or before certain aspects of the study. Nine specimens were used for measuring relevant dimensions and for developing the placement of the sensor and fixation hardware (Table 1). (Specimen 15R was abnormal and excluded from the study. In this sample the manubrium was short and malformed, consistent with congenital abnormality or bony erosion from cholesteatoma, and the ossicles were surrounded by mucosal and fibrous tissue.) Umbo motion was measured in nonimplanted specimens (specimen numbers noted in Fig. 4) to determine whether the middle ear was functioning normally in the temporal bone specimens. We tested the functional performance of implanted UmboMic sensors fully in four specimens, 19L,R and 20L,R. However, the umbo displacement in sheep 20 varied from the mean, and this seemingly translated to abnormal implanted responses. Thus we focus on fully implanted responses from 19L,R. Supplemental Figures 2 and 3 show data from 20L,R. The UmboMic sensors have been characterized in bench studies and human temporal bones [9, 18, 19].

Table 1.

Parameters of the facial recess in sheep temporal bone as measured by the ruler function in 3D Slicer from pre-surgical micro-CT scans. (A micro-CT scan of 15R was taken but was excluded from the reported values due to its abnormal anatomy as noted in text.) Sheep 13 was a Suffolk breed; the others were Hampshire. The last column lists the socket arm lengths selected from the virtual fitting. The values in the last row are from [23]

Temporal bone sample	Facial recess height max. (mm)	Facial recess length (mm)	Extended facial recess length (mm)	CTy origin to SMF (mm)	Facial nerve to CTy angle (degrees)	Umbo to CTy origin (mm)	Umbo to SMF (mm)	Socket arm length (mm)
13R	2.73	13.3	16.6	4.84	59.6	11.6	13.7	10
13L	2.52	13.3	15.0	6.07	59.3	11.5	12.2	8
14R	2.07	12.8	13.1	5.51	52.5	10.4	10.6	10
14L	2.57	13.8	13.2	6.37	37.5	12.1	9.1	10
15L	3.12	12.3	16.4	3.07	57.7	11.3	13.4	8
17R	1.88	11.0	14.0	4.31	52.7	10.8	12.4	10
17L	1.84	10.2	12.4	5.90	40.8	10.2	11.4	10
18R	2.35	11.1	13.8	5.83	52.1	11.0	12.3	8
18L	2.12	11.7	14.9	4.26	51.3	11.4	13.1	10
Mean±STD (range)	2.36±0.40 (1.84–3.12)	12.2±1.23 (10.2–13.8)	14.4±1.5 (12.4–16.6)	5.13±1.02 (3.07–6.37)	51.5±7.7 (37.5–59.6)	11.1±0.6 (10.2–11.6)	12.0±1.4 (9.1–13.7)
Human Mean±STD (range)	2.65±0.41 (2.24–3.45)	12.41±2.91 (9.4–18.56)	NA	4.08±0.8 (2.06–5.5)	26.91±1.19 (25–28.69)	NA	NA

CTy chorda tympani, SMF stylomastoid foramen

Fig. 4.

Umbo motion in nine Hampshire sheep cadaveric ears. a Mean velocity amplitude normalized to ear-canal sound pressure (blue solid line) with 95% confidence interval (shaded area). b Mean phase relative to ear canal pressure (blue solid line) with 95% confidence interval (shaded area). c Mean displacement normalized to pressure (blue solid line) with 95% confidence interval (shaded area). The black dashed line in A–C is the average result from Péus et al. [25] included for comparison. d Displacement plotted versus SPL from one ear to evaluate linearity of umbo motion. The dashed line in D provides a linear reference. Temporal bones used: 14R,L; 17R,L; 18R,L; 19R,L; 20L

Fig. 3.

3D rendering of anatomical landmarks in 3D Slicer. a Facial nerve and chorda tympani (in blue) and ossicles (in yellow), facial recess maximum height (black line), facial recess length (red line), extended facial recess length (red dashed line), and facial nerve to chorda tympani angle (red arrows). b Tympanic membrane (TM) (in green), osseous labyrinth (in purple), stylomastoid foramen (represented as red dashed circle), umbo to chorda tympani origin, where chorda tympani splits from the facial nerve (black line), umbo to stylomastoid foramen (black dashed line), and chorda tympani origin to stylomastoid foramen (red line). c, d The sensor (tan) in position with the tip contacting the umbo. In c, the transparent ball arm (red) is in the foreground (refer to Fig. 2)

Anatomic Measurements

Anatomic dimensions are needed for the design, fitting and surgical implantation of the UmboMic system. Using the micro-CT scans, linear distances between anatomic landmarks were measured in 3D Slicer using the ruler function. Radiological measurement of facial recess parameters has been found to correlate with surgical measurement [24]. Initially, we measured facial recess parameters with a physical ruler in post-surgical samples. These measurements correlated well with their parallels via micro-CT, so subsequent samples were evaluated radiologically only. Figure 3a, b indicates the seven parameters measured; Fig. 3c, d shows the sensor in position. Similar distance measurements were previously used to describe round window membrane access for cochlear implant surgery in humans [23]. The maximum facial recess height, measured from the facial nerve to the tympanic annulus, was a mean of 2.36 ± 0.40 mm with a range from 1.84 to 3.12 mm. The facial recess length, measured from the chorda tympani origin (where it branches from the facial nerve) to the most posterior point of the short process of the incus, was a mean of 12.2 ± 1.23 mm with a range from 10.2 to 13.8 mm. The extended facial recess length, measured from the stylomastoid foramen to the most posterior point of the short process of the incus, was a mean of 14.4 ± 1.5 mm, with a range from 12.4 to 16.6 mm. The mean distance from the chorda tympani origin to the stylomastoid foramen was 5.13 ± 1.02 mm ranging from 3.07 to 1.02 mm. The mean distance from the umbo to the chorda tympani origin, indicating the depth of the facial recess traversed by the sensor, was 11.1 ± 0.6 mm, ranging from 10.2 to 11.6 mm. The mean distance from the umbo to the stylomastoid foramen was 12.0 ± 1.4 mm, with a range from 9.1 to 13.7 mm. The measured distances for each temporal bone sample are given in Table 1 with human values from [23] included in the final row. Based on the values in columns 2 and 3 of Table 1, the area of the spatial recess is slightly smaller in sheep than in humans.

Measurements of Umbo Motion

Figure 4a, b shows the velocity magnitude and phase of the umbo in nine Hampshire sheep temporal bones. The displacements shown in Fig. 4c are calculated from the velocities, thus derived from the same measurements. Velocity is shown to compare to results in the literature from sheep and other mammals, displacement is shown because displacement produces sensor bending. Mean and 95% confidence intervals are plotted in the blue line and blue shaded region, and the black dashed lines are the previous results from Péus et al. [25]. The mean velocity magnitude increased from ~ 0.04 mm/s at 0.5 kHz to a maximum of ~ 0.5 mm/s/Pa, at 5.5 kHz. Above 5.5 kHz, the velocity went through peaks and valleys that mostly averaged out in the mean to a level of ~ 0.3 mm/s/Pa through 20 kHz, the highest frequency measured. For comparison, in human the umbo velocity magnitude increases to a value of ~ 0.2—0.3 mm/s that extends from ~ 0.6 to 8 kHz, and then drops off at higher frequencies [11, 12]. In Fig. 4c, the mean displacement varied nonmonotonically between 12 and 20 nm/Pa to 5.5 kHz, dropping off at higher frequencies. The level and frequency response of the umbo velocity (Fig. 4a) were reasonably consistent with previous measurements from Péus [25]—they overlap at all frequencies available when a ~ 10 dB standard deviation from the Péus study is considered. The phase response (Fig. 4b) was similar in shape and value to that from Péus but shifted to the right in frequency, and a shift to the right was also apparent in the amplitude. This shift suggests that the present bones were stiffer [11]. This could be due to the different breeds of sheep (White-Alpine versus Hampshire), or to the condition of the bones, which were frozen/thawed in the present study and fresh in the Péus study. Umbo motion scaled close to linearly at all frequencies to at least 90 dB SPL (Fig. 4d). Below 1.5 kHz the responses became noisy at 60 dB SPL so are not included. The noise level in the velocity measurement depends on the reflection of light from the bone, which varied in different specimens. At frequencies above 10 kHz notches and peaks in the velocity responses made them sensitive to small changes, and this is likely the reason for the small apparent deviations from linearity; in particular, the deviations are not indicative of motion saturation.

Implantation of UmboMic Sensor with Printed Fixation Hardware in Cadaveric Temporal Bone

The UmboMic system was implanted in 13 temporal bones (13R,L; 14R,L; 15L; 17R,L; 18R,L; 19R,L; 20R,L). The sensor, sensor plate and ball arm were inserted through the surgical opening such that the tip of the sensor was in contact with the umbo (Fig. 5a). The sensor plate thickness was 0.5 mm, thus the combined thickness of the sensor, adhesive, and sensor plate, the total assembly thickness, remained under 1 mm. This ensured adequate clearance within the facial recess, whose maximum height (from chorda tympani to facial nerve) was 1.84 mm or greater in our specimens (Table 1), allowing a degree of rotational freedom of the sensor during implantation.

The curved ball-arm extended from the sensor plate to outside the facial recess (Fig. 2, Fig. 5a). Taking advantage of its curved design, the ball-arm was positioned on one side of the sensor plate, which allowed for improved visualization of the umbo on the opposite side. The socket, prefilled with UV-curable dental cement, was then positioned to capture the ball, completing the ball-and-socket mechanism. The socket was connected to a screw plate via a relatively long socket-arm, which was rotated into position such that the screw plate could be fixed to the mastoid cortex using two titanium bone screws (Figs. 2b and 5b). To account for anatomical variation in the distance between the umbo and the mastoid cortex, the socket arm was fabricated in three different lengths, 6, 8, and 10 mm. (Previously, in our virtual study, various configurations were tested in SolidWorks to determine the optimal fit for each specimen prior to surgery.) In Supplemental Video 2, a virtual study shows that the 6-mm socket arm was too short for the temporal bone in the video and did not reach the edge of the mastoid cortex, while the 10-mm socket arm resulted in a gap between the screw plate and the cortex. The 8-mm version provided the best fit. Table 1 includes the socket arm length chosen for all CT-scanned specimens; notably, the 6-mm socket arm was never virtually predicted or physically used.

In a similar ball-and-socket set, 2 months of accelerated-aging testing in saline at 67 °C confirmed that a cemented ball and socket would be secure for at least 16 months, which is adequate for the 2-month planned in vivo study. After the ball and socket were removed from solution, the two parts could not be manually separated with pliers, and a 4.5-kg weight hung from one side of the piece for more than 2 weeks did not separate the two parts.

In the physical implantation, once the screw plate was secured and the sensor tip was seen to be in contact with the umbo, the dental cement within the socket was cured using UV light to permanently fix the ball in place for the desired configuration. An additional layer of dental cement was then applied around the ball-and-socket joint and cured to increase mechanical stability. The fixation device was straightforward to handle and allowed controlled positioning during implantation. The overall sensor fixation and placement procedure required approximately 30 min. Figure 5a shows an operating microscope view of the implanted fixation hardware with the sensor tip in contact with the umbo. The facial recess was expanded posteriorly (with removal of facial nerve section) in this specimen to achieve an adequate photograph.

The primary challenge in the implantation process was limited visualization of the umbo due to the narrow width of the facial recess. As a result, placement assessment under the surgical microscope was not always reliable. To verify sensor placement, six temporal bones underwent post-implantation micro-CT scanning. An example of a correctly positioned sensor is shown in the post-implantation micro-CT image in Fig. 5c. The inset in panel c identifies structures; Supplemental Video 3 provides additional visualization. The arrow shows the point of connection between the umbo and the sensor tip. The sensor plate appears bright and thicker than its actual thickness due to metal artifacts in the CT scan. The combined thickness of the sensor, epoxy, and plate was measured to be ~ 1.2 mm in the CT scan, whereas the actual physical thickness was approximately 1 mm. Visual inspection alone was insufficient for accurate placement. Therefore, in our final implantations, a functional feedback method was employed as described below, to provide real-time confirmation of sensor tip positioning relative to the umbo during implantation.

Sensor Operation When Fixed in Temporal Bone

Positioning success of the UmboMic sensor to the umbo was improved by recording sensor responses to sound signals during the implantation to provide functional feedback. The performance of implanted sensors was evaluated prior to fixing the ball-joint by delivering sound through a speaker and recording sensor responses into a laptop running the open-source application Audacity (https://www.audacityteam.org). Headphones allowed for live feedback during positioning. Studies in live animals would therefore benefit from intraoperative real-time signal feedback through the implanted microphone.

After positioning and fixing an UmboMic into specimen 19R, Fig. 6a shows the response when sound was delivered to the EC at levels from 60 to 90 dB SPL. In the relatively constant region of largest responses (~ 3 to 11 kHz), the output was ~ 3 mV/0.2 Pa (reading from the 80 dB SPL curve). This corresponds to 8 fC/Pa when the 1.9 mV/fC amplifier gain is accounted for. This is similar to the sensitivity measured in human cadaveric ears [19], although the most sensitive range is shifted up in frequency compared to human. In sheep, the responses below 0.5 kHz were relatively small and often in the noise at 60 dB SPL. Sensor operation in human cadaveric temporal bones has been characterized from 0.1 to 10 kHz [18, 19].

Fig. 6.

Output of UmboMic when implanted in sheep temporal bones 19L and 19R. a Frequency responses of 19R for stimulus levels 60–90 dB SPL. B. Linearity of output, found from data in a. c Frequency responses and noise levels for 19R and 19L (summed in 1/3 octave bandwidth) to determine SNR. Responses before and after applying bone cement are shown. d Sensor sensitivity (UmboMic response/umbo motion) when UmboMic sensors were implanted in 19R and 19L before and after applying bone cement to enhance fixation of sensor to bone

Figure 6b plots Fig. 6a data versus stimulus SPL at several frequencies and illustrates good linearity of the UmboMic operation in the sheep ear for wide dynamic range and wide bandwidth.

The solid lines in Fig. 6c show UmboMic responses when implanted in sheep temporal bones 19L and 19R. The measurements in 19R were made the day after those of Fig. 6a, after disassembling and reassembling the sound system. Different sensors were used in the two specimens, but the same sensor was used in 19R on the two days. The output in specimen 19R was slightly larger on day 2 than on day 1; from 2 to 10 kHz, it is mainly between 0.01 and 0.02 V on day 2, whereas it was close to 0.01 V on day 1. From 2 to 0.5 kHz, the drop off was ~20 dB on day 1 and only 10 dB on day 2. The variation of 19R might have been due to sound stimulus calibration differences in the open-field configuration. Overall, the day-to-day responses were reasonably repeatable. The responses were also reasonably similar in the two specimens, with two sensors. The noise level in Fig. 6c is reported in 1/3 octave bands, which is standard when reporting hearing aid microphone signal-to-noise ratio (SNR) specifications [9]. At stimulus input of 90 dB SPL, the output noise level was ~45 dB lower than the output signal responses in the approximately flat frequency region. Thus, output signals of 45 dB SPL input would be at the noise level. This SNR is about 15 dB worse than what was found in the human temporal bones [19]. The difference is mainly due to an elevated noise level, because the output per Pa delivered at the EC is similar in the roughly flat frequency region of the two species. We are working to understand the cause of the increased noise level and suspect sensor damage due to abrasion through the development of the sheep implantation strategy.

Figure 6c also shows responses in the two specimens after further securing the fixation device to the bone with a liberal application of bone cement (DePuy Synthes high viscosity). The responses are smoother after the application of bone cement; for example, a notch in 19L between 5 and 6 kHz is missing. Following the application of bone cement and the ~ 15-min setting time, the sensors were permanently fixed in the specimens.

Figure 6d shows sensor sensitivity, found by normalizing the sensor output (from Fig. 6c) by the displacement of the umbo. The sensor sensitivities measured when inside the temporal bones were close to bench-tested sensitivities of UmboMic sensors [19] although with more pronounced frequency structure. This was likely due to a degree of looseness in the attachment of the bone screws into the mastoid cortex, which allowed the sensor fixation hardware to rotate (undesired) in addition to the desired bending motion of the sensor tip during the sound stimulation. The dotted lines in Fig. 6d show the sensitivities after the application of bone cement. The frequency responses were smoother; thus the bone cement improved the coupling between the sensor and umbo. Above 10 kHz, pronounced frequency structure emerged even with the bone cement, perhaps due to the umbo motion becoming more complex at high frequencies.

Further testing of the UmboMic sensor system in human temporal bones was recently published and we continue to develop the sensor and age-test it both for sheep studies and eventual human implantation. For example, the recent publication included a quantitative study of the effect of the applied force between the umbo and the sensor, finding that the sensor response was stable over a wide range of force [19].

Discussion

The sheep has been studied as a model for cochlear implantation and for temporal bone dissection in surgical training [15, 26, 27]. For our purposes, the sheep model provided adequate space within the facial recess to test the UmboMic sensor and approximate its function in humans. We confirmed the previously described sheep temporal bone anatomy using both microscopic visualization and micro-CT [13, 28]. In doing this study, the following observations were reaffirmed: The sheep cartilaginous external auditory canal featured a sharp turn. The pars flaccida was larger and more visible than the pars tensa through the external canal [28, 29]. The mastoid was poorly pneumatized, and the mastoid cells were filled with a soft substance. The sheep temporal bone also featured a large hypotympanic cavity. The tegmen was not easily accessible when drilling superiorly; instead, there was a cavity containing a blood vessel that traverses this superior region. The mastoid bone needed to be carefully drilled away in order to enter the middle ear space, without identifiable landmarks like the semicircular canals or tegmen to guide the dissection. Although many specimens contained a large space in the bone that appeared similar to an antral cell, there was no connection between this space and the middle ear space.

We made observations that, to our knowledge, have not been previously reported: Several of the samples contained a proliferation of mucosal and/or fibrous tissue in the middle ear cavity. In some, the mucosal tissue was adherent to the tympanic membrane and/or ossicles. This could indicate a history of middle ear disease. The sheep in this study were at least four years old, and the condition might be age-related. In future live animal studies, pre-screening with otoscopy or prophylactic antibiotics, administered pre-operatively, could be considered [30]. Prior to surgery in live animal studies, evaluation via CT under anesthesia would be beneficial to pre-screen for anatomic abnormalities or middle ear disease such as otitis media with effusion or cholesteatoma.

Using parameters described from human temporal bone CT scans [23], in Table 1, we defined boundaries of the sheep facial recess which limit the size of a middle ear implant in the sheep model. To our knowledge, these parameters have not been previously noted. The values of the facial recess height and length (ranging from 1.8 to 3.1 mm and from 10.2 to 13.3 mm, respectively) limit the dimensions of the implanted device. The distance between the umbo and the chorda tympani origin, where the chorda tympani splits from the facial nerve, ranged from 10.2 to 12.1 mm and corresponds to the depth of the facial recess that contains the UmboMic sensor.

We developed a method of virtual surgical planning and customization of implantable devices using micro-CT images and open-source software (3D Slicer). This method of virtually “drilling” the temporal bone could be used by surgeons or biomedical engineers for several uses: designing an implantable device, verifying fit of a device, adjusting a device to an individual’s specific anatomical measurements, and more. (For designing and fitting the socket arm we used SolidWorks, which is not open-source. Open-source CAD software exists and could likely be substituted.) In summary, the process of virtual simulation based on micro-CT imaging as demonstrated in this study might be used for other implantable devices in animal or human studies.

Lessons learned through the iterative process of virtual and actual drilling, and UmboMic implantation included the following: Obtaining visibility of the umbo required technical skill and was impossible in certain specimens without extension of the facial recess. The chorda tympani would need to be sacrificed in most preparations to obtain adequate visibility. Visualization might be improved with endoscopic imaging. Removal of the incus could be considered, but further studies would be needed to determine if incus removal significantly improves visibility and placement of the UmboMic. Possible effects on the UmboMic function with such a modification would also have to be determined. It would prevent ossicular sound transmission to the cochlea, preventing transmission of sound for ears with residual hearing. However, for future in vivo sheep study with the goal of studying UmboMic performance (without cochlear implant), such modification might be acceptable.

The strategy for the fixation hardware that secured the UmboMic sensor was successful in that sensitivity of the UmboMic was maintained when implanted and fixed in place with the hardware. The addition of bone cement at the location of the bone screws improved the stability, flattening the sensitivity frequency response curve. Finally, we observed variations in sensitivity based on contact of the sensor with the umbo, as was previously noted in studies in human cadaveric ears [19]. To obtain good sensitivity, as in the values noted here, required good visualization and real-time testing via auditory feedback while sound was presented to the EC. Therefore, live studies would benefit from intraoperative real-time sound feedback through the implanted microphone.

The amplitude of umbo velocity in sheep and human (without coupled sensors) increases with frequency at ~ 6 dB/octave to a corner frequency, which is ~ 0.6 kHz in human and ~ 5 kHz in sheep (Fig. 4, [11, 12]). Above the corner frequency, the velocity amplitude is ~ 0.3 mm/s/Pa in both human and sheep. Thus, both species include a frequency region of fairly wide-band sensitivity. However, in sheep the response pattern is shifted to higher frequencies, so that the response is smaller than in human below ~ 4 kHz and greater above ~ 10 kHz. This shift to higher frequency is also apparent in the sheep audiogram [31, 32] and was apparent in the output of the UmboMic sensor upon sound stimulation, reported in Fig. 6. Thus, the sheep is a reasonable, albeit imperfect, large animal model of human hearing, and for studying an implanted microphone. The frequency response of the UmboMic’s sensitivity to sound stimulation in the sheep will guide our work going forward. For example, in test stimuli, frequencies below 4 kHz could be boosted so the microphone output would provide a better model of human responses.

Further Planning for Live Animal Study

Looking forward, the live animal study is planned as a chronic implantation, with UmboMic signal recordings made for 2 months following surgical implantation. Recordings will be made in awake animals. Sounds and words will be delivered open-field to the sheep, and the output of the implanted microphone and amplifier will be delivered via a percutaneous connector to a radio frequency (RF) transmitter that will be held in a jacket when testing is underway and removed (uncoupled from the connector) otherwise. The connector exit point will be between the shoulder blades or at the back of the neck where the sheep cannot scratch and damage it. The signal will be transmitted to an RF receiver approximately 2 m away. The RF transmitter/receiver are already in-lab and in use in preparation for the live animal study. An UmboMic amplifier of volume ~ 1.5 × 1.5 × 0.3 cm³ will be embedded in a medical-grade epoxy that has been used to seal integrated circuits for implantation in the body [33]. Amplifier power is provided by the battery in the RF transmitter. Accelerated-aging tests of the amplifier embedding and UmboMic sensor are underway; aging tests of the fixation epoxy were described above and are continuing.

Surgery will be done by an otologic surgeon on the team and veterinary staff. Sheep will placed prone with hind legs lateral in a natural resting position, as described in [15]. Based on that prior work, this positioning is expected to give adequateaccess to the surgical site and be tolerated by the sheep. We forsee that surgical positioning will be a challenge and plan to address this with directed practice with whole heads, and a mobile operating table. The surgical approach is based on the transmastoid extended facial recess approach, a standardapproach for cochlear implantation [14, 15, 34]. In sheep, there are some anatomical differences regarding mastoid aeration and dimensions of middle ear spaces, as noted in Table 1, and some modifications are necessary. Otologic surgeons with experience and comfort working through the facial recess will be performing these procedures. Practice with both temporal bones and whole heads is necessary, and we have continued our preparation with whole head sheep dissections. Standard surgical and anesthesia protocols and management of adverse events will be followed. Enhanced exposure to the sheep middle ear through sacrifice of the chorda tympani nerve might be necessary but is not expected to cause significant complications. However, chorda tympani sacrifice was associated with taste alterations and salivary disorders in some patients [35, 36], and the effects of chorda tympani sacrifice in sheep are unknown. Adverse conditions will be managed following veterinary consultation.

The implantation is expected to be simpler than previously described incus-contacting procedures but more complex than subcutaneous microphones. The complexity of the procedure is greater than conventional cochlear implants, but not excessively so. Over the course of this study, once the implantation method was refined the implantation was straightforward; it is expected to add ~ 30 min to a mastoidectomy procedural time. Challenges in visualization of sensor placement were improved by real-time sound feedback. In ongoing work in whole heads, endoscopic visualization is being implemented.

The planned live-animal study will serve to inform (1) the ability of the sensors to remain functional for an extended period of animal movement and mechanical impact (head-butting or rubbing); (2) the degree of signal contamination due to chewing, breathing and other body-produced sounds; and (3) the degree of signal contamination from other components of noise in the natural environment. It is noteworthy that sheep ruminate—they chew their food twice, once after regurgitation. This will likely increase the incidence of chewing noise. We will initially record sound when the sheep is still anesthetized and not ruminating.

Building on the studies described in the present paper, whole head studies are underway so that the surgical team is fully trained in the surgical access and implantation and fixation procedure prior to embarking on the live animal study. (See also [16], which describes cadaveric studies preceding live animal middle ear implantation, in sheep.)

Conclusion

The UmboMic is a bimorph piezoelectric microphone developed for use in a totally implantable cochlear implant. In preparation for live large animal trials in sheep, we studied the sheep temporal bone anatomy and tested the UmboMic in cadaveric sheep ears. Firstly, we defined relevant parameters of the sheep facial recess. Secondly, we developed methods of virtually drilling micro-CT scans to design and customize UmboMic fixation hardware. The physical 3D-printed fixation hardware successfully fit each sheep temporal bone sample during physical implantation. These methods could be used in surgical planning and customization of other implantable devices. In testing the implanted UmboMic with fixation hardware in sheep cadaveric temporal bones, sensor sensitivity was preserved as compared to bench tests of the UmboMic. Thus, the UmboMic implantation in cadaveric sheep ears was possible, and this study represents a step towards planned live animal studies.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution

ICC and CZ performed experiments and wrote MS; DLN assisted in experiments; EFW and JZZ constructed sensors and amplifiers for testing and edited MS; IK advised; DBW and JHL advised and edited manuscript; HHN advised, edited manuscript, and oversaw project; ESO participated in experiments, oversaw project, and wrote MS; JHL, HHN, and ESO secured funding.

Funding

This work was supported by NIH NIDCD grants R01DC021996 and R01DC016874.

Data Availability

Data from this article will be available via https://doi.org/10.6084/m9.figshare.31999599.

Code Availability

Not applicable.

Declarations

Ethics Approval

The sheep ears were obtained from previously deceased animals. No animals were killed for the purposes of this research, and therefore, institutional animal use approval was not applicable.

Conflict of interest

The authors declare no competing interests.

Use of AI

AI was not used in this work or manuscript preparation.