Long-Term Electromyographic Monitoring of the Stapedius Reflex via Implanted Electrodes in Sheep: Toward Objective Autonomous Cochlear Implant Fitting

1. Introduction

Cochlear implants (CIs) have evolved remarkably since the first single-channel device was implanted in 1957 by Djourno and Eyriès, progressing to modern systems with up to 26 electrodes and smartphone-based control interfaces. These advances have enabled many CI users to achieve high levels of speech recognition. After the development of the first multi-channel electrode arrays the possibility of decoupling CI fitting from subjective patient feedback was discussed and the search for objective fitting strategies began [1,2].

Despite extensive research into objective fitting methods—including electrically evoked auditory brainstem responses (EABR), electrically evoked compound action potentials (ECAP) / neural response telemetry (NRT), and electrically-evoked stapedius reflex thresholds (eSRT)—routine CI fitting still relies heavily on psychoacoustic methods [3,4,5,6,7,8,9,10,11]. These methods are inherently limited in populations that are unable to provide reliable feedback, such as infants or individuals with cognitive impairment, and in patients with no prior auditory experience where adequate perception of new sensory impressions is difficult. Moreover, subjective assessments are time-consuming and prone to misjudgement, often resulting in suboptimal stimulation levels and reduced speech recognition outcomes [11,12,13].

Objective fitting methods are essential for optimizing speech recognition outcomes in cochlear implant users [7,14,15]. Among these approaches, the eSRT has consistently demonstrated high reliability for estimating the maximum comfortable loudness level (MCL) [1,5,6,11,13,14,16,17,18,19,20,21,22,23,24]. The stapedius reflex—a protective muscular response to intense acoustic stimulation—can be assessed either indirectly via middle-ear admittance measurements or directly using electromyography (EMG) [9,25,26,27,28,29]. However, clinical eSRT assessment based on middle-ear admittance is frequently compromised by factors such as middle-ear effusion, tympanic membrane condition, probe placement variability, and limited patient cooperation. These factors reduce measurement accuracy and reproducibility [30,31] but are intrinsic to admittance-based techniques and do not affect EMG-based monitoring of stapedius muscle activity.

The long-term stability of SRT after CI implantation remains a subject of debate. While also very recent studies report adaptive changes in SRT and MCL over time [7,12,13,23,32,33,34,35,36,37,38,39,40,41], others, such as Pitt et al. (2021), found no significant longitudinal variation in SRT across a large cohort [42]. Conversely, Brotherton et al. (2017) demonstrated a significantly lower SRT following a 4-week sealing of one ear and its subsequent normalization [38]. Raghunandhan et al. (2014) reported a postoperative increase in eSRT and MCL, attributed to tissue remodelling around the CI electrode up to 1 year post implantation [13]. In addition, Muigg et al. (2025) demonstrated SRT changes within the first 4 years after CI implantation [41]. These findings suggest that SRT may be influenced by biological and procedural factors, including neuronal adaption, healing dynamics and electrode positioning.

Given the importance of accurate and repeatable fitting, frequent post-implantation assessments are essential. Even minor deviations in mapping can significantly impact speech recognition [7,43,44]. Integrating EMG electrodes into CI systems could enable routine, objective SRT measurements by audiologists and speech therapists. Recent technological advances even suggest the feasibility of autonomous CI fitting based on real-time EMG feedback.

In this study, we investigate the feasibility of long-term monitoring of stapedius muscle activity using chronically implanted custom-made EMG electrodes. Leveraging the anatomical similarity between the ovine and human temporal bone [45], we conducted a six-month longitudinal assessment in adult sheep. This work serves as a foundational step toward developing self-fitting CI systems based on EMG-derived SRT measurements.

2. Materials and Methods

Ethical Approval and Animal Welfare

All procedures were conducted in compliance with European and German animal welfare regulations. The study protocol was approved by the Committee for Animal Research of the State of Thuringia, Germany (approval code: UKJ-19-006). Animal husbandry and all procedures were carried out at the Animal Facility of the Jena University Hospital.

Animals and Study Design

Nine adult female Merino sheep (age: 4–5 years; weight: 70–100 kg) were included in the long-term study, and one adult female Merino sheep cadaver used exclusively for histological analysis. Upon arrival, animals underwent a minimum acclimatization period of five weeks to allow for health screening and intensive handling by trained personnel, which significantly facilitated postoperative care. Animals were included in the study only if the presence of a stapedius reflex was confirmed prior to the implantation procedure.

The left stapedius muscle was accessed via a retrofacial approach [46,47]. Two custom-designed platinum-iridium EMG disk electrodes (MEDEL, Innsbruck, Austria), comprised of a stranded wire (7 single wires, 10IR9.49T) with a tip of approximately 0.65 mm diameter (Figure 1a), were implanted on the surface of the muscle. Recordings were scheduled at four time points: during implantation (baseline), and at 1-, 3-, and 6-months post-implantation.

Following the final measurement, both stapedii (left and right) were surgically excised and subjected to histological examination to assess tissue integrity and potential implant-related changes.

Surgery preparation and pre-screening

The SR monitoring via middle-ear admittance was performed using a commercial tympanometer device (eTymp, Biomed Jena GmbH, Jena, Germany) (Figure 1b, c). A tympanometer custom-made probe with an elongated tip was used to accommodate the anatomical differences between the animal and the ovine external auditory canal. In addition, the outer ear was cleaned and let dry in the stable one day before the surgery.

The sheep were sedated and prepared in the stables for pre-screening with short acting midazolam (0.1 - 0.3 mg/kg, Midazolam-Ratiopharm^® 15 mg/3ml, Ratiopharm, Ulm, Germany) and ketamine i.m. (10 - 20 mg/kg, Ketamin 100 mg/ml, WDT, Garbsen, Germany). The neck and the skin around the ears were shaved with a long hair trimmer (Aesculap^® Vega GT 606, Aesculap, Tuttlingen, Germany) before a Safety IV Catheter with injection port (Vasofix^® Safety, B.Braun, Melsungen, Germany) was placed into the jugular vein. A syringe driver was connected to apply propofol (18 - 40 mg/kg/h, bolus 2 ml/kg BW, Narcofol^®, cp-Pharma, Burgdorf, Germany). The relaxed animals were intubated with a tracheal tube of 8.5 mm or 9 mm (TrachealTube, Smiths Medical Portex^®, Minneapolis, USA) and ventilated with isoflurane (1 – 3.5 Vol%, Isofluran CP^® 1 ml/ml; CP-Pharma). Fentanyl (i.v. 2 µg/kg BW, Fentadon^® 50 µg/ml, Dechra, Aulendorf, Germany) was used for analgesia as bolus. Immediately after induction of anesthesia on the operation table, the presence of the stapedius reflex on each animal was determined. Before the reflex test, the ears were cleaned with cotton buds (Meditrade^®, Kiefersfelden, Germany) to remove any remaining earwax to reduce the risk of probe contamination or clogging.

Electrode Implantation

Immediately after induction of anaesthesia, the presence of an acoustically evoked stapedius reflex was reassessed. From the nine animals included, one did not present any reflexes during this pre-screening and was, therefore, excluded from the study and recovered without implantation. The other eight animals exhibited a reliable reflex and were subsequently implanted. All implantations were performed on the left side to facilitate intraoperative rumen venting via a cannula, thereby preventing abdominal distention during prolonged anaesthesia.

After skin incision caudoventrally of the ear, the tendon of the sternocleidomastoid muscle was partially removed to expose the basis of the mastoid bone. A hole was drilled (UNIDRIVE^® II-plus, Karl Storz, Tuttlingen, Germany) to find the facial nerve which was identified using brief stimulation pulses (3 Hz and 1.0 to 2.5 mA) via a bipolar nerve locator connected to a NeuroSign 100 (MAGSTIM Company Limited, UK). The acoustic control was based on the myoelectrical activity of the facial muscles, which were monitored using 3 subdermal needles (two for bipolar recording + one grounded), placed on the left cheek.

The facial nerve was then traced in depth to reach the stapedius muscle by the retro-facial approach without opening the middle ear, allowing ipsilateral acoustic stimulation (Figure 2a, b). It is imperative to avoid injury to the facial nerve and the labyrinth structures, which are situated close to the stapedius muscle in sheep (Figure 2c, d). In addition, with permanent irrigation, drilling was carried out at low speeds in the vicinity of the facial nerve to prevent damage due to overheating. The stapedius muscle was carefully exposed caudally to the facial nerve (Figure 2c) and the electrodes were placed at its surface (Figure 2e).

For electrode placement, a pocket between the stapedius muscle and the petrosal bone was prepared by separating the origin of some muscle fibers. The second electrode was placed at the opposite side between the facial nerve and the muscle belly as far away from the other electrode as possible (Figure 2e). The electrodes were fixed in position with fibrin glue (Surgibond, SMI AG, St. Vith, Belgium). The electrode cable was placed within the drilled hole (except the bare ends), which was finally closed using bone cement. After complete fixation, the reflex was monitored again by the middle-ear admittance and by EMG-recording. After several repetitions, the EMG system was disconnected and the wound was rinsed with ampicillin (Ampi-Dry, Veyx, Schwarzenborn, Germany) before the skin incision was closed (Vicryl^TM 3/0 and Prolene^TM 2/0; Ethicon^®, Norderstedt; Germany) and carprofen (1 – 4 mg/kg BW i.v., Carprosol^® 50 mg/ml, cp-Pharma, Burgdorf, Germany) was given. Bandages were applied around the ears and the neck (Inadine^®, Solventum, Kamen, Germany, Artiflex^® Soft, BSN medical, Hamburg, Germany, Elastomull^® haft hospital, Leukoplast^®, BSN medical, Hamburg, Germany, tg^®fix, Lohmann & Rauscher, Rengsdorf, Germany) after connecting an elongation to the Safety IV Catheter. Then the animal wok up. The sheep were monitored over the next few days. The bandages were reapplied daily and the catheter was used to apply analgesia (carprofen 1 - 4 mg/kg BW, i.v. Carprosol^® 50 mg/ml, cp-Pharma, Burgdorf, Germany) and antibiotics (enrofloxacin 5 mg/kg, i.v., 1x daily for 3 days, Baytril^® 50 mg/ml, Bayer, Leverkusen, Germany) for 4 days postoperatively. After 4 days, long-term antibiotics (oxytetracycline 20 mg/kg BW, i.m. Terramycin^® LA 200 mg/ml, Zoetis, Berlin, Germany) were administered. To prevent a painful reaction to the oxytetracycline, lidocaine (1 ml/ per cm injection site, Lidor^® 20 mg/ml, WDT, Garbsen, Germany) was used.

Follow up and final control

First, the bare ends of the electrodes were exposed, the electrodes were connected, and the impedance between the electrodes was measured to verify the integrity of the cables. Then, bipolar EMG measurements were performed during ipsi- and / or contralateral acoustic stimulation. In some cases, alternative monopolar measurements were performed. If impedance exceeded expected values and cable breakage was suspected, access to the electrode lead distal to the breakage site was attempted only during the final (6-month) measurement. If no reflex was recognizable with middle-ear admittance or in the EMG signal, the measurements were repeated after the anaesthesia was lowered as much as possible without compromising the animal’s health.

A flow-chart of the testing, follow-up intervals and histological investigations is depicted in Table 1.

At the end of the final measurements the animals were euthanized in deep anesthesia using pentobarbital-sodium (100 - 150 mg/kg BW, i.v., Euthadorm^®, 500 mg/ml, cp-Pharma, Burgdorf, Germany). Afterwards the electrode positions were verified, and the stapedius muscle of the left and the right side were dissected for further histological analysis (Figure 2d).

Histological treatment

The left and right stapedii of all implanted sheep (n = 8) and one additional control (muscles of a cadaver) were fixated in 4% formalin (Formaldehyd, Otto Fischar, Saarbrücken Germany), decalcified with Osteomoll^® (Sigma Aldrich, Darmstadt, Germany) and subsequently classically histologically treated. As the electrodes were always implanted on the left side the right stapedius muscle served as control (Table 1).

The prepared muscles were embedded in paraffin (Sigma-Aldrich, Taufkirchen, Germany) before 7 µm thick serial sections were made using a microtome (HM 360 Microm, Marshall Scientific, Hampton, USA). The slices (approx. 300 per muscle) were mounted on microscope slides and stained with Heidenhain's AZAN trichrome stain to discriminate muscle and connective tissue.

All slices were photographed using an Axioscan Z1 (ZEISS, Oberkochen, Germany) and visually analyzed. The evaluation focused on identifying differences between the left and right muscle bellies, specifically the presence of increased muscle fiber atrophy, connective tissue, fibrotic or scar tissue, or conspicuous fat accumulations.

Outcome MeasuresAcoustic Reflex

Stapedius reflex monitoring was performed using a commercial tympanometer (eTymp USB, BioMed Jena, Jena, Germany), equipped with a custom-elongated probe tip to accommodate anatomical differences between the human and ovine external auditory canal (Figure 1).

Acoustic stimulation was delivered using eAudio software (BioMed Jena, Jena, Germany), whereby one to three repetitions of pure-tone stimuli at frequencies of 500 Hz, 1000 Hz, 2000 Hz, and 4000 Hz was applied. Stimulus intensities were variable but always within 70 - 110 dB SPL ((sound pressure level) maximum 100 dB SPL for 4000 Hz), with a fixed duration of 0.5 seconds and a post-stimulus observation window of 1 second. One recording consisted of an intensity and frequency sweep. Intra-operatively, if no response was observed, probe repositioning was attempted or anesthesia was reduced to a minimum, and after a few minutes the test was repeated.

To synchronize EMG recordings with acoustic stimulation, a trigger signal from the tympanometer was integrated into the EMG acquisition system. Reflex responses were classified into two categories based on amplitude, waveform morphology, and reproducibility across repetitions: “Reflex”, and “No Reflex” (Table 1).

Electromyography (EMG)

Myoelectrical activity was detected from the stapedius muscle via a pair of custom-made platinum-iridium disc electrodes (MED-EL, Innsbruck, Austria) (Figure 1a). A subcutaneous needle electrode was inserted as a reference. The bipolar EMG was conditioned with an EMG front-end (Dual Bio-Amp, AD Instruments, Dunedin, New Zealand) and recorded at 100 kHz with a PowerLab (AD Instruments, Dunedin, New Zealand). The front-end was configured to bandpass filter from 1-5000 Hz.

During the recording, electromagnetic noise was kept to the minimum, by disconnecting the operation table, electrosurgical unit, and driller from the power outlets. Cables were shielded either with external shielding tubes or shielded cables were used, if the noise was still high, the 50 Hz notch filter from the EMG front-end was activated.

Signal Processing

The middle ear impedance and EMG signals were integrated in Matlab (The Mathworks Inc., Natick, USA) for post-processing. Both signals were resampled to 10 kHz.

The middle-ear impedance data were used as provided by the eAudio software without further filtering. The data were aligned with the EMG using the trigger signal. To quantify the stapedius reflex response, the absolute value of the signal was computed, and the 90th percentile of the waveform within the first 200 ms following stimulus onset was extracted. This percentile-based metric was selected to approximate the peak response while reducing sensitivity to transient noise artifacts. A reflex was classified as present when the signal exceeded 0.03 ml, corresponding to the standard detection threshold implemented in the eTymp software.

The tympanometry device is design for human anatomy; therefore, probe placement was frequently suboptimal in sheep due to anatomical constraints and time limitations during surgical procedures. This often resulted in noisy recordings, underestimated reflex amplitudes, or a complete absence of detectable responses, even in cases where reflex activity was confirmed by EMG or was expected based on the response pattern at adjacent stimulation intensities. To address these limitations, a second visual screening was performed to exclude false-positive reflex detections caused by noise spikes. Conversely, signals that were not automatically classified as reflexes were re-examined for subthreshold but clear, reproducible, and physiologically consistent responses, as well as for cases in which increasing stimulus intensity should have elicited a larger reflex. This classification strategy ensured that biologically plausible reflex responses were retained, thereby compensating for technical limitations of the tympanometer and improving the overall reliability of reflex detection.

Raw EMG signals were filtered using a 32nd-order IIR bandpass filter (80 – 800 Hz) to isolate stapedius muscle activity and suppress low-frequency artifacts and high-frequency noise. Persistent 50 Hz powerline interference and harmonics were removed using a pattern extraction filter based on 40-cycle modelling. Additional frequency-specific noise arose from unshielded cable sections and elevated impedance due to mechanical stress or unstable micrograbber connections. These vulnerabilities led to contamination from the tympanometry device probe tone (226 Hz) and stimulus (500 Hz), and additional cross-over from surgical equipment ranged between 500 – 550 Hz. These artifacts were selectively attenuated using 32nd-order IIR bandstop filters, each targeting ±10 Hz around the identified frequencies, or broader ranges when peaks clustered closely.

To determine the presence of a stapedius reflex in the EMG recordings, two quantitative features were extracted from each stimulus response. The main metric was the maximum amplitude of the EMG envelope within the same post-stimulus window (30 – 130 ms). The envelope was calculated based on the moving root mean square (RMS) with a window length of 10 ms. The second metric was the RMS ratio, calculated by dividing the RMS value of the signal within a 100 ms window following stimulation onset (specifically, from 30 to 130 ms) by the RMS value of the baseline signal recorded during the 200 ms preceding the stimulus. While the first metric looks for a minimum amount of activity expected from a muscle, the second one prevented the false detection of reflexes due to high baseline noise.

A response was classified as a reflex if the RMS ratio exceeded 1.15 and the peak amplitude reached or surpassed a value of 3 µV_RMS. A second visual screening was performed to identify responses in which EMG activity was slightly below threshold but became clearly detectable later within the stimulation burst. Conversely, responses attributable to transient noise spikes were reclassified as no reflex. The selected thresholds were empirically derived from the distribution of confirmed reflex responses and optimized to balance sensitivity and specificity.

Finally, to determine the overall outcome of each measurement, a cross-validation was performed using the reference signal (middle-ear admittance) and electrode integrity data. A measurement was classified as a “success” if an EMG signal indicative of a stapedius reflex was detected at one or more stimulation intensities. Conversely, it was considered a “failure” if no EMG signal was observed, despite reflex confirmation by the middle-ear admittance signal at one or more stimulation intensities, and no evidence of electrode breakage was present. In cases where the absence of EMG could plausibly be attributed to either a lack of reflex activation or compromised electrode function, the measurement outcome was labelled as “undefined”.

3. Results

3.1. Animals

Of the nine sheep initially enrolled, eight underwent the full protocol and one animal (S08) was excluded prior to implantation due to the absence of a stapedius reflex during pre-screening. From the eight implanted animals, S04 was categorized as sham-implanted control, because post-mortem analysis revealed that both EMG electrodes had been incorrectly positioned and were not in contact with the stapedius muscle. Therefore, only the histological and middle-ear admittance data were valid, and EMG recordings were discarded and marked as “not measured” in Table 1. Across the eight treated animals, a total of 27 measurement sessions were conducted, encompassing implantation, follow-up, and termination procedures. The second follow-up session was omitted in all animals (except S03) due to electrode breakage, which rendered EMG recordings unfeasible even when reflexes were observed via middle-ear admittance. During the final measurement sessions, when breakages were identified, electrode access was restored by drilling through the bone cement to expose the distal end of the lead beyond the site of damage. In total, data from seven sheep (S03, S05, S06, S07, S09, S10, and S11) were included in the EMG analyses, while middle-ear admittance reflex assessments were available from eight animals (S03, S04, S05, S06, S07, S09, S10, and S11). Histological data were analyzed from the eight implanted animals (S03, S05, S06, S07, S09, S10, S11), 1 Sham-implanted control and one cadaveric control (see Table 1). The left stapedius muscle of S03 was lost during dissection, but all remaining samples were successfully preserved for histological evaluation.

During the implantation session, pre-operative stapedius reflexes were detected via middle-ear admittance in all eight animals, although intra-operative reflexes were not detectable in three cases (S05, S10, S11). EMG responses were observed in at least one measurement in four animals out of seven animals with valid EMG datasets (S03, S06, S07, S09).

During the follow-up sessions, intra-operative reflexes were detected via middle-ear admittance in all but 1 animal (S10). However, EMG recordings were compromised in all cases due to lead breakages affecting one or both electrodes. In one animal (S03), reflex-related EMG activity was successfully recorded using a monopolar configuration.

Termination measurements showed consistency with the implantation results. The same three animals that lacked intra-operative reflexes during implantation (S05, S10, S11) also showed no reflexes at termination, and correspondingly, no EMG signals were detected. Middle-ear admittance data show reflexes in the remaining five animals, including S04. Among these, all four animals with correctly placed electrodes exhibited EMG activity during acoustic stimulation.

3.2. Reflexes Detection

Overall, SR detection via middle-ear admittance corresponded with the presence of the EMG and both signals showed a good temporal correlation if both were measurable (Figure 3).

Figure 3 depicts an example of the recorded signals from a single measurement (Sheep S03), showing the responses to acoustic stimulation compared to the filtered EMG. The 3rd row shows the EMG envelope. As expected, the middle-ear admittance signal and EMG increased clearly as the sound level was raised. In a few cases, the EMG signal already showed a reflex that had not yet been detected by the tympanometer device.

Four different patterns could be distinguished in the reflex-related EMG activity: a short burst at the beginning of the stimulus; a short burst followed by smaller tonic activity; then a series of burst spikes; and finally, a tonic activity fusing those spikes. Generally speaking, these four behaviors corresponded to reflexes going from small to very large reflexes.

From the sessions where reflexes were observed, 1423 stimuli along all combinations (intensity and frequency) were analyzed. From those 886 were classified as reflexes with activity peaks between 3 and 68 µV RMS and RMS ratios between 1.0 and 12.2. The classification parameters worked relatively well, with only 16 records visually re-classified from no reflex to reflex, and 18 records re-classified from reflex to no reflex.

3.3. Histology

During implantation, it was noticed that the facial nerve and the stapedius muscle were not enclosed separately by bone as in humans, but were located together in the sinus tympani. Only a thick layer of connective tissue isolated the two structures from each other. In addition, the stapedius muscle was bulkier than in humans, in which it is relatively slender and long (Figure 2d). The stapedius could be reached in every animal via the retrofacial approach, as it always became visible during drilling at the very lateral aspect of the geniculum of the facial canal. The labyrinth was in some cases very close to the muscle (0.5 mm). It was accidently injured in one animal (Figure 2c). The animal recovered within a few days and the experiment was continued as planned. Access via the middle ear, on the other hand, would be much more complicated in sheep, as the tendon of the muscle disappeared under the facial nerve and no pyramidal eminence is developed (Figure 2c).

No differences were found between the left and right stapedius muscle of the two controls (S04 and cadaveric control) or in the implanted animals. No signs of atrophy or fibrosis were evident in any of the left stapedius muscles. In addition, no scar tissue or noticeable accumulations of fat were detected despite the implantation of EMG electrodes (Figure 4).

Since the electrodes were placed on the surface of the muscle rather than inside, the implantation did not appear to cause any damage to the muscle tissue or noticeable scarring. Thus, the electrodes could be removed without any problems after 6 months. Even the detachment of the muscle fibers from the wall of the cavity to prepare the pocket did not lead to any visible muscular changes.

However, all examined stapedius muscles could be divided into three morphologically distinct regions: a distal, a middle, and a proximal region (Figure 5). Each region was characterized by a specific composition.

Distally, i.e., close to the stapes, the intramuscular tendon was very prominent in all animals and occupied a large area of a cross-section. The few muscle fibers were longitudinally aligned in the cross-sections and spanned between the wall of the cavity and the tendon (Figure 5a). In the middle region, the tendon lost its compactness and spread across the entire cross-section (Figure 5b). It remained prominent, but was now almost completely surrounded by muscle fibers. On the side facing the facial nerve, the muscle fibers ran transversely, and on the opposite side continued to run longitudinally. The tendon was also surrounded by some fat cells. Superficially, near the bone wall, thicker nerve branches followed the course of the stapedius muscle. Proximally, the muscle fibers clearly predominated (Figure 5c). The tendon receded into the background but was still surrounded by fat cells. The majority of the muscle fibers, approximately 2/3, were oriented transversely. The rest attached to the tendon at a 90° angle. The longitudinally oriented fibers were found throughout the stapedius muscle on the side facing away from the facial nerve. In the proximal region, close to the wall of the cavity of the petrous bone the innervating nerve branches were located. It seemed, that they formed a kind of network on the surface of the stapedius muscle (Figure 5d).

3.4. Adverse Events

In 2 sheep (S03, S10), the vestibular system was damaged during the implantation surgery. The severity of the damage varied from slight to problematic for 3 days post-implantation. In all cases the sheep recovered.

In another case, there was an infection on the implantation area (S07) and another sheep had skin healing problems around the connector (S04). Both issues had no serious consequences and recovered within one to three days.

4. Discussion

The results of this study demonstrate that chronically implanted EMG electrodes can reliably detect stapedius reflex activity in sheep after acoustic stimulation over a six-month period, provided that electrode integrity and correct placement are maintained. This confirms the feasibility of long-term electromyographic monitoring of the stapedius muscle and supports the potential for integrating such electrodes into CI systems to enable objective fitting. The temporal correlation between EMG and middle-ear admittance signals, particularly the graded increase in response amplitude with increasing stimulus intensity, reinforces the validity of EMG-based detection. Moreover, the classification metrics used—peak amplitude and RMS ratio—were robust, with only minor manual corrections required during visual screening.

When comparing all objective methods for determining the MCL, the best correlation was found for postoperative eSRT [7,11,18,22,48], which is therefore ideal for CI fitting. However, clinical application remains limited (<39%) [31,40], because additional equipment (impedance meter) is required, which must be operated by an experienced audiologist and the structures between the eardrum and the stapedius muscle must be intact. In addition, the results can be affected due to fluid in the middle ear, the condition of the eardrum, inflammation or problems within the ossicle chain, and, above all, movement by the patient during the measurements [30,49]. This makes the objective determination of the reflex threshold inaccurate and time-consuming especially with young children or patients with disabilities [41]. Our use of long-term implanted EMG electrodes addresses many of these limitations by bypassing the middle ear and enabling direct measurement of stapedius muscle activity. This could facilitate the work of the audiologists, improve the quality of patient care, and stapedius reflex related research studies.

In contrast to Pitt et al. (2021) [42], several studies have demonstrated changes in both eSRT and MCL, particularly during the first year after CI implantation [7,12,13,23,31,36,37,38,39,50,51]. Observed changes in neural response telemetry (NRI) (or ECAP electrically evoked compound action potentials) in the first postoperative year support these findings [12,33]. Additionally, further studies have shown that even small changes can affect subsequent speech recognition [37,41]. Consequently, we assume that patients would benefit from regular and accurate determination of eSRT [50] and the associated fitting of the implant. The reason Pitt et al. (2021) were unable to detect any changes of eSRT over time [42], was likely due to their recruitment of patients who, on average, already had 56 months of CI experience. However, the strongest increases in eSRT occur within 6 months and further changes are completed within a year [7,41]. This can be attributed to the ingrowth of the CI electrodes and the corresponding scarring [30,49]. However, as CIs are current-controlled devices that deliver a constant current regardless of the surrounding tissue, it is likely that the changes are primarily due to neuroplasticity—i.e., the brain's adaptation to new stimuli.

To reliably record the myoelectrical signals of the stapedius muscle for eSRT measurements, correct electrode placement is essential. Our histological analysis showed no evidence of atrophy, fibrosis, or scarring in the implanted stapedii. This finding suggests that surface electrode placement via the retrofacial approach [46,47] at the muscle belly does not compromise muscle integrity, even after six months. This is a crucial observation, as the stapedius muscle is a small and delicate structure, containing on average only about 400 muscle fibers [52], and measuring approximately 2 mm in diameter and 5 mm in length. Thus, any intramuscular damage could easily impair its reflex function. Based on the findings of Walluks et al. (2024), who reported the formation of a 500 µm thick connective tissue ring around an intramuscular implanted electrode [53], a complete loss of the stapedius muscle would be expected after this kind of implantation. Nevertheless, Zarowski et al. (2021) demonstrated successful stapedius reflex measurements by a middle-ear admittance following the chronic implantation of an EMG electrode [29]. In this study, EMG measurements were performed exclusively intraoperatively in humans. Postoperatively, the researchers monitored over a six-month period whether the stapedius reflex could still be elicited. Since the monopolar electrode they used was positioned at the pyramidal eminence, i.e., in the distal region of the muscle with fewer muscle fibres and a prominent tendon, this may have prevented damage to the actual proximal muscle belly and led to the positive results.

The reflex absence in some animals despite histological evidence of the integrity of the treated stapedius muscle, intact electrodes and preoperative reflex confirmation highlights the sensitivity of the stapedius reflex to anaesthesia and surgical manipulation [29]. The disappearance of the reflex could be attributed to a stapedius nerve damage, which enters the muscle at the implantation side. However, the nerve regeneration growth rate of 1 mm per day [54,55] and the continuous co-contraction of the tensor tympani muscle [52] argue against such an injury. The tensor tympani muscle contractions should still be measurable by middle-ear admittance, and the stapedius muscle is expected to be reinnervated within a month. We therefore assume that the use of isoflurane, which is known to suppress reflex responses [56], likely contributed to the intraoperative loss of reflexes in several cases.

Technical challenges such as signal variability due to cable shielding and connector stability were consistent with the general knowledge, but are expected to be mitigated in clinical CI systems through integrated, miniaturized hardware. Importantly, these limitations did not preclude reflex detection in animals with intact electrodes and preserved reflex pathways.

Limitations

Although the ovine model offers anatomical similarities to the human temporal bone [45], key differences—such as the absence of the pyramidal eminence and the shared cavity of the stapedius muscle and facial nerve — limit the direct translatability of the surgical approach. These anatomical constraints may not fully reflect the clinical conditions during human cochlear implant (CI) surgeries, particularly regarding electrode placement and access to the stapedius muscle. However, Arnold et al. 2022 and Guntinas et al. 2022 demonstrated that access to the stapedius muscle via the retrofacial approach is feasible also in humans [46,47].

The use of acoustic stimulation in this study, while advantageous for avoiding electrical artifacts, does not replicate the conditions under which eSRT is typically measured in CI users. Future studies will need to address the challenge of recording EMG signals in the presence of electrical stimulation artifacts, which are inherent to CI-based eSRT measurements [25,26,27,28,29]. The integration of artifact suppression techniques will be essential for clinical translation.

5. Conclusions

This study confirms the feasibility of long-term EMG monitoring of the stapedius reflex using chronically implanted surface electrodes in a large animal model. Reflex-related EMG signals were successfully recorded over six months, and histological analysis confirmed muscle integrity, supporting the safety of the implantation technique.

By enabling direct measurement of stapedius muscle activity, this approach addresses limitations of conventional eSRT methods and offers a promising tool for objective CI fitting. The six-month monitoring window aligns with the critical period of eSRT and MCL adaptation, suggesting that EMG-based tracking could enhance fitting precision during early post-implantation stages.

Although technical and anatomical challenges remain, the results support further development of EMG-integrated CI systems and lay the groundwork for autonomous or clinician-assisted mapping strategies. Future work should focus on artifact suppression and human validation to fully realize the clinical potential of this technology.