Revealing intracochlear microtrauma during cochlear implantation using optical coherence tomography and electrochemical sensing

Abstract

Cochlear implants (CIs) are a highly effective treatment for severe to profound hearing loss, and earlier implantation is associated with improved auditory outcomes. As implantation is now extended to patients with residual hearing, preserving cochlear structures during surgery is a key priority. However, implant insertion trauma remains a risk and can compromise residual hearing. Currently, intraoperative methods capable of detecting such trauma in situ are lacking. We introduce a novel approach for in situ detection of intracochlear trauma using a gerbil model. Specifically, we use optical coherence tomography (OCT) in vivo to visualize structural damage, including basilar membrane ruptures and osseous spiral lamina fractures, and employ modified CI electrode arrays to sense hydrogen peroxide concentration, a marker of oxidative stress. Intracochlear trauma is validated and quantified using contrast-enhanced microcomputed tomography, enabling a novel trauma scale. Our results show that hydrogen peroxide levels are significantly correlated with the trauma scale and trauma volume. A receiver operating characteristic curve was established for the detection of intracochlear trauma, highlighting the sensor’s diagnostic potential. These results show that combining in vivo OCT imaging with electrochemical sensing effectively detects intracochlear trauma, laying the foundation for next-generation CIs with real-time monitoring during implantation.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-26535-7.

Introduction

The risk of developing hearing loss over a lifetime is higher than ever, primarily due to aging, but also influenced by increasing noise pollution, personal listening devices, and high-intensity sound levels allowed at concerts¹. Hearing loss is the most common sensory deficit with serious consequences, including social isolation, socio-professional difficulties, and even the risk of developing dementia^2,3. Today, an estimated 5% of the world’s population suffers from disabling hearing loss, and this number is expected to double by 2050.

Cochlear implants (CIs) form an effective treatment for severe-to-profound hearing loss, bypassing damaged hair cells to directly stimulate the auditory nerve. Over the past decade, patients with residual hearing have received CIs, benefiting from the combined effects of electric and acoustic stimulation^4,5. Therefore, preserving residual hearing is essential to maximizing CI outcomes. However, atraumatic cochlear implantations remain challenging, and insertion trauma forms a major cause of postoperative residual hearing loss, occurring in up to 20% of cases⁶. Currently, a trauma scale developed by Eshraghi is widely used to classify intracochlear trauma based on histopathological findings⁷. The Eshraghi scale classifies intracochlear trauma into four grades, ranging from Grade 0 (no trauma) to Grade 3 (severe trauma involving osseous spiral lamina fractures or modiolar damage) following cochlear implantation. According to this scale, the most common trauma is a dislocation of the electrode array from scala tympani (ST) to scala vestibuli (SV). Other types of trauma include basilar membrane (BM) elevation and disruption, as well as fractures of the osseous spiral lamina (OSL) and modiolus⁸. This scale does not incorporate non-destructive imaging or correlations with electrochemical measurements, but is rather based on histology, which necessitates destructive preprocessing of the samples. We developed and applied a novel trauma scale based on CECT imaging and intracochlear hydrogen peroxide measurements, which differs in both scaling and methodology from the Eshraghi scale.

Insertion trauma may not only impact the efficiency of the electrical stimulation of auditory nerve fibers but can also lead to further structural damage by triggering chronic biological responses like inflammation, fibrosis, and ossification. These processes progressively degrade the cochlear function, putting​ residual hearing preservation at risk⁶. To mitigate these risks, surgeons use trauma-minimizing techniques such as slow insertions to reduce fluid displacements and the administering of corticosteroids to prevent or attenuate potential inflammatory responses. While these measures improve outcomes, they cannot eliminate the occurrence of insertion trauma in the first place, particularly given the variability in cochlear anatomy. This highlights the clinical need for real-time intraoperative methods to detect and prevent cochlear insertion trauma.

Currently, surgeons lack direct, real-time feedback on intracochlear trauma during CI insertion. Intraoperative methods to detect trauma include tactile feedback, impedance measurements, X-ray fluoroscopy, and electrocochleography (ECochG), though each has limitations. Tactile feedback is subjective and requires extensive experience. Impedance techniques, such as electric field telemetry, trans-impedance matrix, and impedance field telemetry, primarily confirm electrode position (e.g., detecting tip fold-over or extracochlear placement), while four-point impedance (4PI) can indicate intracochlear bleeding, often associated with basilar membrane (BM) damage, and correlated with residual hearing loss^9–11. X-ray fluoroscopy can detect electrode misplacement, such as scalar translocation, but does not directly detect soft tissue damage¹². ECochG monitors cochlear function during insertion by capturing hair cell and auditory nerve responses to sound, with drops in amplitude potentially indicating trauma and predicting postoperative hearing loss¹³. Postoperative CT scans detect complications such as tip fold-over or electrode dislocation, but their resolution is insufficient to visualize the delicate intracochlear structures. Contrast-enhanced CT (CECT) produces higher resolution (~ 6 µm) images of soft and bony tissue, but it is limited to ex vivo imaging¹⁴. Optical Coherence Tomography (OCT) provides real-time, high-resolution imaging of intracochlear anatomy and can visualize structures like the BM, OSL, and spiral ligament¹⁵. Additionally, OCT has been increasingly explored for intra-operative diagnostics and research¹⁶. OCT-guided insertion probes are being developed to help surgeons avoid trauma, offering a promising intraoperative tool¹⁷. However, structural imaging alone may not capture the entire picture of a cochlear injury, which involves complex biological responses.

Biosensors have the potential to functionally detect insertion trauma by measuring biochemical markers of inflammation. Trauma triggers local increases in concentration of inflammation markers such as hydrogen peroxide (H₂O₂), histamine, and cytokines^18–22. In healthy tissues, H₂O₂ concentrations below 15 μM have been reported, but levels can exceed 100 μM in inflamed or damaged tissue²³. Electrochemical biosensors detect specific analytes with high selectivity through immobilized biorecognition elements (e.g., enzymes, antibodies) that convert molecular interactions into measurable electrical signals. Permselective membranes enhance their response specificity by allowing only the target molecules to reach the electrode²⁴. Electrochemical sensors are widely used in glucose monitors and have been demonstrated for multiple other analytes. They typically make use of three electrodes: a functionalized working electrode, where the species of interest is detected, a stable reference electrode, and a counter electrode. In amperometric sensors, a potential difference applied between the working and reference electrodes generates an electrical current proportional to the analyte concentration.²⁵. While CI electrodes are designed for neural stimulation, their use in electrochemical sensing is emerging. CIs have been modified with molecularly imprinted polymers to detect the inflammatory marker histamine in human cadaveric cochleae²⁶. It was also shown that non-modified CI electrodes can be used for the detection of oxidizable species in vitro and oxygen levels in vivo in rat cochleae²⁷. Recently, we developed a hydrogen peroxide sensor by modifying CI electrodes with an electropolymerized o-phenylenediamine layer and an antifouling dopamine-sulfobetaine coating, validated in vivo in gerbils²⁸.

In this study we investigate a bimodal strategy combining OCT imaging with hydrogen peroxide-sensitive CI electrodes to characterize the relation between trauma and intracochlear hydrogen peroxide levels. We find that OCT can be used to visualize trauma to intracochlear structures and electrode position, while modified CI electrodes can measure peroxide concentrations in real-time. It is shown that hydrogen peroxide concentration changes are indicative of the severity of insertion trauma and can be related to well-known radiological trauma scales, possibly enabling rapid intraoperative intervention.

Results

Ex vivo trauma assessment with contrast-enhanced micro-computed tomography (CECT)

We assessed trauma, induced using a small needle (see methods – surgical procedure), by CECT imaging, which allows for visualizing both soft and bony intracochlear tissue. In 9 cochleae, we identified different forms of trauma. Ruptures of the BM are observed in 2 cochleae, resulting in a fistula between the scala media (SM) and scala tympani (ST). OSL damage was observed in most cases, ranging from minor dents or fractures to extensive damage. In some cochleae, OSL trauma presented as cracks in the OSL, while soft tissue remained intact. In other cochleae, the damage extended to both the bony and soft tissues, including severe fractures. In the most severe cases, this resulted in the dislocation of surrounding structures, such as the spiral limbus, as well as the formation of large fistulae between the ST and scala vestibuli (SV). In a few cases, the modiolus was also minimally affected. Representative trauma types are illustrated in Fig. 1.

Fig. 1.

Different types of intracochlear trauma visible on CECT images. Trauma positions are indicated by the red arrows: examples of (a) a non-traumatized cochlea, (b) small trauma at the modiolar side of the OSL, (c) Cracked OSL with potentially intact soft tissue, (d) rupture of the BM, forming an opening between the ST and SM, (e) OSL trauma with a small fistula formation between ST and SV, (f) severe trauma to OSL at the modiolar side with a large fistula formation. Abbreviations: scala vestibuli (SV), scala media (SM), scala tympani (ST), Reissner’s membrane (RM), spiral limbus (SL), spiral ligament (SLig), osseous spiral lamina (OSL), basilar membrane (BM), organ of Corti (OoC).

Intracochlear trauma was categorized using the following radiological scale: Score 0 indicates the absence of trauma, Score 1 corresponds to structural damage without fistula formation, such as a small fracture of the OSL, modiolus, or elevation of the BM. Fistulas due to severe cracks in the OSL and rupture of the BM were given a Score of 2.

Trauma, induced by a small needle as described in the methods (surgical procedure), was localized between 9 and 35% along the cochlear spiral, corresponding to the basal portion of the first cochlear turn. The extent of trauma, quantified as a proportion of the total spiral length, ranged from 13.7% to 24.8%, while trauma volume ranged from 0.0026 to 0.017 mm³. A higher proportion of trauma did not always correspond to a larger volume. The cochlear spiral length, measured along the lateral edge of the OSL, averaged 9.98 mm. The electrode insertion depth was measured in one representative sample without induced trauma, from the round window membrane (RWM) to the electrode tip as approximately 3.13 mm, which corresponds to about 32% of the total spiral length. Thus, the electrode placement consistently overlapped with the region where the needle-induced trauma was observed, supporting a direct spatial association between the electrode location and the cochlear damage. The trauma proportion and volume for each case are shown in the table in Fig. 2.

Fig. 2.

Representations of electrode and trauma inside two different cochleae, segmented from ex vivo CECT images. (a) the electrode is visible inside the cochlea and the insertion depth is 32% in this case. (b) Cochlear spiral with trauma illustrated in red. Data in the table represent the length of the cochlear spiral for each traumatized cochlea (n = 9) and the proportion of trauma along the cochlear spiral, along with the measured volume of each traumatic insertion. Abbreviations: Osseous spiral lamina (OSL), cochlear implant (CI), round window membrane (RWM), electrode (E), cochlear spiral (CS), insertion depth (ID).

In vivo trauma assessment in the hook region using optical coherence tomography

To assess intracochlear trauma in vivo, we performed transmembranous OCT imaging. Before opening the RWM, OCT images confirmed the intact cochlear anatomy in all 14 cochleae, with well-defined structures including the RWM, osseous spiral lamina (OSL), basilar membrane (BM), organ of Corti (OoC), and spiral ligament (SLig) (see Figures. 3a, g). These pre-opening scans served as the anatomical baseline for trauma evaluation. Following RWM opening, the OCT images revealed the presence of floating debris in the scala tympani in some cases, even in conditions where no trauma was visible (Fig. 3b), suggesting that minor damage can occur due to surgical manipulation alone. When trauma was deliberately induced, OCT imaging consistently revealed structural disruptions indicative of insertion-related damage (Figs. 3c, e, h). Trauma to the BM was identified as a visible separation at its attachment point to the OSL. In OCT images, this presents as a hyporeflective (dark) region in an area that normally appears continuous, indicating a detachment or tear. Damage to the OSL was characterized by increased surface roughness on its scala tympani-facing border. In pre-trauma images, this border appeared smooth and well-defined, whereas post-trauma OCT scans revealed irregularities and disruptions in the surface contour. To quantify these changes, we performed a roughness analysis that showed a marked increase in total variation following trauma (Supplementary information (SI) Fig. 1), supporting its value as an OCT-based indicator of OSL damage. The contrast in surface morphology between trauma and non-trauma conditions was further illustrated using normalized x- and y-axis coordinates (Fig. 3 k).

Fig. 3.

In vivo assessment of cochlear trauma using OCT and validation with ex vivo CECT imaging. (a) and (g) Pre-opening OCT images showing the intact hook region, including key intracochlear structures such as the RWM, OSL), BM, OoC, and SLig; (b) OCT image following RWM opening, showing intact structures; floating debris are visible in the scala tympani due to surgical RWM opening; (c), (e), (h) Post-trauma OCT images. A red asterisk indicates BM disruption; a green asterisk indicates OSL trauma, identified as separation or loss of surface roughness; (d), (f), (i) Corresponding ex vivo CECT images validating structural disruptions observed with OCT. a–c are taken from the same cochlea as shown in (j), where 3D OCT imaging visualizes electrode position in relation to the BM and OSL; (k) Roughness comparison between trauma OSL from (h) and non-trauma OSL from (g), normalized location. Abbreviations: (RWM) round window membrane, (OSL) osseous spiral lamina, (BM) basilar membrane, (OoC) organ of Corti, (SLig) spiral ligament, (CECT) contrast-enhanced microcomputed tomography.

These in vivo OCT findings were verified by ex vivo CECT imaging (Fig. 3d, f, i), which confirmed the location and extent of BM and OSL trauma seen on OCT. The consistency between OCT and CECT validates the use of OCT for real-time, in vivo trauma assessment. Additionally, 3D OCT allowed for in vivo visualization of the electrode within the cochlea, showing its proximity to the OSL and BM. Even in the presence of the electrode, trauma remained clearly detectable, specifically as disruptions in the normal continuity between the BM and its attachment to the OSL.

Sensing of hydrogen peroxide intracochlear concentration

Firstly, we investigated the relation between the measured intracochlear H₂O₂ concentration, the trauma score (as discussed in the previous paragraph), and the trauma volume. Hydrogen peroxide concentrations in the cochlea were measured by implanting modified cochlear implants and using them as peroxide sensors. This methodology is based on previously published work²⁸. CI sensor modifications are described in the methods, in the section titled “hydrogen peroxide biosensor fabrication and calibration in vitro,” and included a PoPD permselective membrane, only allowing perfusion of peroxide, and an antifouling SBMA-based coating. For sensor readout, a 4-step voltage protocol was used. An example of this voltage protocol, raw data, and a calibration curve for a modified CI sensor can be found in the SI (SI Fig. 3). All sensors are calibrated separately in vitro before in vivo experiments. See the methods section for a more detailed explanation of the sensing readout. Peroxide measurements were conducted on cochleae with varying gradations of trauma induced (Score 0: n = 5, score 1: n = 4, score 2: n = 5) over a 2-h period post-implantation, see previous section for explanation of trauma scores identified by CECT. The evolution of H₂O₂ concentration over time, depending on the trauma score, can be seen in Fig. 4a, showing that the concentration of hydrogen peroxide measured was higher directly following the implantation, and then decreased to reach a stable value over the next 2 h. We note that the time interval between the implantation/trauma application and the measurement start, or the total experimental time window may be different for different experiments (i.e., not each experiment was started exactly 15 min after the implantation/trauma application, and in some cases, gerbils died before the 2-h mark was reached). Thus, not all data points displayed in this graph have the same number of repeats, n, (i.e., number of cochleae). Nevertheless, except for score 2 at 15 min (n = 1) and score 1 at 120 min (n = 2), the number of repeats was always larger than 3. Figures 4b and c show that the H₂O₂ concentration correlates with both the trauma score with Spearman correlation coefficient (SCC) and p-value of 0.57 and 0.035, and the volume of trauma with Pearson correlation coefficient PCC = 0.7 and p-value = 0.006. However, Fig. 4d shows that the measured volume of trauma is also correlated with trauma score (SCC = 0.86 and p-value < 0.001). Standardized multiple linear regression revealed that almost the entire effect could be explained by the volume of trauma (Coefficient = 2.1 ± 1.1 ∙ 10³ μM per mm³ of trauma, p-value = 0.086) rather than the trauma classification (0.4 ± 8.2 μM per trauma score step, p-value = 0.96).

Fig. 4.

Relation between hydrogen peroxide intracochlear concentration and trauma score/trauma volume. (a) H₂O₂ concentration over 2 h post-implantation per trauma score with standard error. Standard errors are calculated from at least 3 repeats (cochleae), except for score 2 at 15 min (n = 1) and score 1 at 120 min (n = 2). (b) Box plots of highest measured H₂O₂ concentration over a 2 h time window after implantation/trauma per trauma score (SCC = 0.57; p-value = 0.035). (c) Dependence of highest measured H₂O₂ concentration over a 2 h time window after implantation/trauma on the trauma volume with a linear fit and a 95% confidence interval band (PCC = 0.7; p-value = 0.006 ). (d) Trauma volume, as determined by segmentation of CECT images, in function of trauma score (SCC = 0.86; p-value < 0.001).

Subsequently, we have evaluated whether the detected peroxide concentration can be used to discriminate between trauma and non-traumatic situations. The comparison of the concentration of hydrogen peroxide in the trauma and control groups over time can be seen in the SI (SI Fig. 4). In the case of trauma, the concentration of hydrogen peroxide is higher at all time points. The maximum peroxide concentration recorded over 2 h in the control (non-trauma) and the trauma groups, respectively, are shown in Fig. 5a. In cochleae where trauma was induced, we measured an average H₂O₂ concentration of 58.3 ± 7.0 μM, compared to 38.7 ± 4.6 μM in control cochleae. The concentrations are significantly higher in the trauma group, as validated by a Welch’s T-test (p = 0.037). Normality was verified for both groups using the Shapiro–Wilk test (p_trauma = 0.11, p_control = 0.75). According to the receiver operating characteristic curve (ROC) displayed in Fig. 5b, the optimal cutoff H₂O₂ concentration value for detecting insertion trauma is 42 μM. The area under the ROC curve (AUC) is 0.84, with a 95% lower confidence limit (CL) of 0.63. In comparison, a test based purely on chance would have an AUC of 0.5, corresponding to the diagonal reference line depicted in Fig. 5b. This indicates that our trauma detection method performs significantly better than random classification. An AUC value higher than 0.8 is generally considered acceptable²⁹​. The optimal cutoff value of 42 μM yields a sensitivity of 78% and specificity of 80%.

Fig. 5.

Intracochlear H₂O₂ concentration as a tool for binary trauma detection (i.e., trauma or no trauma). (a) Box plots of highest measured hydrogen peroxide concentration over a 2 h time window after implantation/trauma in case of trauma and no trauma (Welch’s t-test p-value = 0.037). (b) Receiver operating characteristic for classification of trauma/no trauma based on highest measured hydrogen peroxide concentration over a 2 h time window after implantation/trauma . Trauma is labelled as the positive condition. (AUC = 0.84; 95% CL = [0.63 1.05]).

Discussion

The primary objective of this study was to determine whether intracochlear trauma can be detected in vivo, during surgery, using a bimodal approach combining OCT imaging and electrochemical hydrogen peroxide sensing. We performed a series of cochlear implantations, with and without trauma induction in a gerbil model.

Using OCT we acquired in vivo images revealing intracochlear trauma, including ruptures of the BM, and fractures of the OSL. The presence of trauma was subsequently confirmed using CECT performed post-mortem. OSL trauma was quantified by analyzing surface roughness variations in the OCT images, using the same pre-traumatized OSL as a reference. Since this approach was only tested on two cochleae, the dataset is limited, and this investigation solely serves as a proof of concept. Upon opening the RWM, debris was observed inside the ST even before any mechanical trauma was induced. The 3D reconstructions helped distinguish between true surface changes and shadowing effects caused by the debris. These findings support prior ex vivo studies demonstrating the capability of OCT to image intracochlear structures in the most basal part of the cochlea^15,30. Moreover, our data shows the potential to detect structural trauma of the BM and OSL using OCT. However, extracochlear OCT has a limited imaging depth, and hence, trauma occurring deeper along the cochlear spiral remains undetected. This highlights the need for the development of OCT insertable probes^17,31. To the best of the authors’ knowledge, this is the first in vivo study of intracochlear trauma via the RWM using OCT. One prior study quantified cochlear fibrosis using OCT imaging through the otic capsule, which is possible in rodents but not in human cochleae. This suggests a logical next step for this study, namely, moving from acute studies to chronic models to assess fibrosis formation at implant sites³².

To compare trauma and address the depth limitation of OCT, we used CECT as a post-mortem ground truth. This allowed us to examine the degree of damage to the inner ear and the position of the electrode along the cochlear spiral, and compare trauma visualized on OCT and CECT in a 3D context. Through CECT analysis, we confirmed that trauma was localized within the first 36% of the cochlear spiral, correlating with the fact that most insertion trauma happens within the basal part of the cochlea, where large anatomic variations challenge atraumatic CI insertions³³.

Additionally, CECT provided a detailed investigation of structural disruptions, such as BM ruptures, OSL fractures, fractures on the modiolus, and fistulas between scalae. The pixel resolution of CECT scans ranged between 1.7 µm and 2.5 µm, meaning that any trauma occurring at a finer molecular or cellular level was beyond the detection threshold. This limitation suggests that certain early-stage trauma markers, such as localized fibrosis, cellular apoptosis, or oxidative stress responses, may remain undetected in purely anatomical imaging. Future studies could address this limitation by using fluorescence microscopy or molecular contrast agents to visualize cellular and biochemical alterations associated with trauma in vivo⁶.

To classify the severity of insertion-related injury, we propose a novel trauma scale defined by correlating CECT images with intracochlear peroxide concentration levels. This scale categorizes trauma into three levels: Score 0 = no trauma (intact intracochlear structures, 38.7 ± 4.6 µM H₂O₂), Score 1 = minor trauma (OSL microfractures or BM deformations without full rupture, 50.9 ± 5.1 µM H₂O₂), and Score 2 = Major trauma (complete BM rupture, OSL fistula, 64 ± 11 µM H₂O₂). An important extension of this scale, distinguishing it from already existing scales (e.g. Eshraghi⁷), is the fact that we were able to correlate (SCC = 0.57, p_scale = 0.035) our trauma scale with the corresponding hydrogen peroxide concentrations, representing oxidative stress inside the cochlea.

The volume of the trauma segmented on CECT images is also significantly correlated with the concentration of hydrogen peroxide (PCC = 0.696, p_volume = 0.006) measured in the cochlea. This suggests an upregulation of hydrogen peroxide resulting from cellular damage induced by the trauma. A possible explanation, not yet verified in the cochlea, is activation of the NADPH oxidase DUOX by calcium influx due to disruption of cell membranes³⁴. DUOX has been found in the rat cochlea and is predominantly located in the organ of Corti³⁵. Additionally, mitochondria in stressed neurons also increase H₂O₂ production and this effect also correlated with an increase of intracellular calcium concentration³⁶. H₂O₂ is then involved in paracrine signaling and attracts neutrophils to the site of injury which can in turn cause an increase in peroxide concentration through an oxidative burst. To distinguish whether H₂O₂ concentration is mostly influenced by trauma volume or trauma score, we performed a multiple linear regression analysis. This revealed that H₂O₂ concentration is mostly influenced by the volume of trauma, rather than by the trauma score. Importantly, our results indicate that H₂O₂ concentrations correlate well with trauma severity, supporting the potential of electrochemical biosensors as early, in situ indicators of intracochlear stress.

Intracochlear hydrogen peroxide concentrations went from as low as 25 μM, in control conditions, up to 100 μM in the case of severe trauma (score 2). Hydrogen peroxide levels in mammals tend to vary heavily depending on the location and condition of the tissue measured. In zebrafish tissue, concentration gradients from 50 μM to 0.5 μM have been measured depending on the distance to the location of injury and concentrations seemed to peak roughly 20 min after injury ³⁷. In rat brains, following ischemia and reperfusion, concentrations can vary between 60–160 μM²³. In general, it is assumed that a concentration around 1–15 μM is the upper limit for healthy tissues, but it can reach over 100 μM within inflamed tissues³⁸. These literature values from other tissues correspond well with the concentrations measured during our experiments. Specifically for the cochlea, no absolute hydrogen peroxide concentrations have been measured before, but redox homeostasis dysregulation resulting in an increase in oxidative stress by increased concentrations of reactive oxygen species, such as hydrogen peroxide, is thought to be a significant contributor to noise-induced hearing loss³⁹.

Relatively high baseline hydrogen peroxide concentrations were measured in the control cochleae, which can be explained by the fact that even during the atraumatic implantation process, an opening still must be made into the RWM, which inherently causes trauma to this structure. Although the RWM is located further (~ 0.1 mm) from the implanted electrodes, this might cause an increase in baseline hydrogen peroxide levels. Debris resulting from the RWM opening can also be seen in the OCT images, as described earlier. Additionally, some studies report a controversial effect of ketamine on oxidative stress that might cause an increase in hydrogen peroxide concentration, although this effect is not confirmed for cochlear tissue⁴⁰.

In our experiments, the concentration of hydrogen peroxide seems to peak early after the trauma occurred (~ 30 min), similar to hydrogen peroxide dynamics measured in zebrafish³⁷, and then stabilizes but remains elevated in comparison with cochleae in which no trauma was sustained. Even in control cochleae, i.e., without visible trauma, hydrogen peroxide concentration were highest initially after implantation.

The ROC plot analysis reveals that by using peroxide concentration recordings with the modified CI electrode array, we can effectively discriminate between traumatic and atraumatic implantation. Previous work, by Bester et al., was able to construct a ROC curve correlating elevated 4-point impedance (4PI) levels and residual hearing loss 3 months after implantation with an AUC of 0.72, allowing for the prediction of implantation-related hearing loss based on intraoperative impedance measurements. It was shown that 4PI levels are elevated when cochlear bleeding occurs ⁹. On the other hand, peroxide monitoring can identify even less severe trauma, without bleeding. The time points used for the hydrogen peroxide measurements (30 min – 2 h after implantation) are also within the timescale of a CI implantation surgery and are feasible in a clinical context.

The cochlea remains a “black box” with limited tools available to explore its physiological responses in situ. Electrochemical sensors can enhance our understanding of cochlear mechanics, inflammation, and trauma, improving cochlear implantation techniques and hearing treatments. On the short term, peroxide monitoring could enable the evaluation of acute intra-operative trauma, facilitating comparisons between implantation techniques and pharmacological interventions, such as dexamethasone-eluting implants. With further refinements, real-time intra-operative feedback could help surgeons minimize trauma, preserve residual hearing, and optimize electrode placement. In clinical trials, real-time trauma assessment could allow surgeons to adjust insertion parameters, initiate anti-inflammatory treatment for severe trauma, and assess the need for reimplantation. Post-operatively, trauma detection can help predict hearing outcomes, guide targeted care, and monitor implant-site inflammation, enabling personalized treatment and improved long-term cochlear health. This method is not limited to cochlear implants but can also be applied to all other implants consisting of an electrode array, e.g., deep brain stimulation devices and spinal cord stimulators ⁴¹.

In the future, developing an OCT probe for cochlear implantation could enhance visualization of the implant’s trajectory, extending imaging beyond the cochlear hook region⁴². This would enable real-time OCT-guided insertion tools, allowing surgeons to adapt the implant trajectory before trauma occurs, making the procedure more precise and minimally invasive. Additionally, automated OCT-based trauma detection could revolutionize cochlear implantation by reducing reliance on manual interpretation. By applying machine learning algorithms to roughness analysis, trauma on the OSL and other intracochlear structures could be automatically identified, detecting subtle changes indicative of damage. This approach would improve diagnostic accuracy, standardize trauma assessment, and support safer, more consistent implantation techniques.

Future studies on hydrogen peroxide sensing with CIs should investigate the spatial distribution of increased hydrogen peroxide concentration relative to the trauma site and its potential for pinpointing trauma location along the electrode array. Additionally, assessing the impact of locally administered anti-inflammatory steroids, such as dexamethasone, on hydrogen peroxide levels could provide valuable insights. Before conducting clinical experiments, the toxicity and biocompatibility of the sensor modification layers must be thoroughly evaluated, and the safety of the applied voltage protocol used for sensing should be assessed. However, the currents used in the sensing measurements (maximum 1–2 μA) are significantly lower than the threshold for eliciting an auditory response in human CI patients (> 100 μA) ²⁷.

Conclusion

The aim of our study was to investigate the feasibility of intra-operative trauma detection. To achieve this, we conducted in vivo experiments in gerbils (n = 9 trauma, n = 5 control), where insertion trauma was induced and monitored using OCT and continuous electrochemical sensing of hydrogen peroxide. Results showed that OCT can successfully detect BM ruptures and OSL fractures, confirming trauma in vivo. Hydrogen peroxide levels were significantly upregulated when trauma was present compared to controls, and both correlated with trauma volume and a novel trauma scale, which was developed using post-mortem CECT imaging. The results of a receiver operating characteristic analysis yielded an area under the curve (AUC) of 0.84, with a cutoff value of 42 μM for trauma detection, enabling a binary decision for the detection of trauma versus non-trauma. Based on these results, we can conclude that this combined approach enables real-time identification of cochlear trauma based on both structural and biochemical markers.

The ability to detect and quantify cochlear trauma in vivo represents a crucial step toward improving atraumatic implantation techniques, thereby reducing trauma-induced residual hearing loss. To our knowledge, this is the first intraoperative demonstration of trauma detection using integrated imaging and sensing. This study highlights the potential to develop “smart” CIs that not only restore hearing but also may protect the inner ear during implantation by combining OCT imaging and implantable biosensors.

Methods

Surgical procedure

All animal procedures have been approved by the Animal Ethics Committee of the University of Leuven (P087/2022) and comply with the guidelines of the European Community Council Directive. All animal procedures in this study were performed in accordance with the relevant guidelines and regulations. All experimental procedures and reporting are in accordance with the ARRIVE guidelines (Animal Research: Reporting of In Vivo Experiments). Ten-week-old female Mongolian gerbils (Janvier labs, France) were used in these experiments, weighing between 59–75 g. Gerbils were anesthetized using a mixture of 10% ketamine (Nimatek, Dechra, The Netherlands), 7.5% xylazine (2%, Xyl-M, Livestock Pharma, Belgium), and 82.5% saline (0,9%, Fresenius Kabi, France) administered intraperitoneally. The postauricular region was shaved and sterilized, whereafter a postauricular incision was made to expose the bulla. A small fenestration in the bulla was made to access the RWM, which was carefully opened using a micro pick. In the trauma group, controlled trauma was induced using a soft microneedle (diameter 200 µm). OCT imaging was performed before and after the RWM opening to verify the absence or presence of trauma and to get an initial idea of trauma severity. A modified CI electrode was inserted into the ST and fixed, subsequently secured using a small piece of muscle tissue to prevent perilymph leakage. A silver/silver chloride (Ag/AgCl) reference electrode and a platinum (Pt) wire counter electrode were implanted subcutaneously for biosensor measurements. The time of trauma induction and electrode insertion was always considered while evaluating hydrogen peroxide (H₂O₂) levels during a 2-h time frame. After the experiment, gerbils were overdosed with the anesthetic mixture, and death was confirmed by measuring the heartbeat. The bulla was extracted, and OCT imaging was performed through the otic capsule to visualize deeper embedded trauma. In Fig. 6 an illustration of the experimental flow is added.

Fig. 6.

Flowchart illustrating the experimental workflow: (1) First step in the process where trauma is induced and imaged with OCT, to insert the electrode afterward and visualize the electrode inside the cochlea. (2) Two-hour electrochemical sensing measurements. (3) Extraction of the cochlea from post-mortem gerbil and submersion in contrast agent. (4) CECT scanning after 24 h submersion. (5) Segmentation of trauma and cochlear features. (6) Correlation of CECT, OCT and electrode sensing data.

In vivo trauma assessment using optical coherence tomography

OCT imaging was performed using a spectral-domain OCT system (Telesto TEL220C1; Thorlabs, Germany). Detailed settings are described in a previous study¹⁵. Transmembranous RWM imaging was carried out before opening the RWM, before trauma induction, and after trauma induction.

Trauma visible on OCT was qualitatively described and compared with CECT images. A quantitative assessment of trauma was performed by analyzing the roughness of the OSL surface in pre-trauma and post-trauma-induced OCT images. Regions of interest (ROIs) encompassing the OSL were manually selected using Python (version 3.11, https://www.python.org/), with the Matplotlib library (version 3.10.0) and OpenCV (version 4.11.0)^43,44. The images were binarized via Otsu thresholding and refined using morphological closing. The OSL contour was extracted by identifying the first non-zero pixel in each image row within the ROI (SI Fig. 1). The resulting 2D shape was normalized and used to compute roughness metrics. These included the variance of first differences, total variation, second derivative variance (curvature), and residuals from a Savitzky-Golay filter fit. Comparisons between normal and trauma-exposed samples were visualized and saved, and quantitative metrics were reported in the SI Fig. 2 (created using Matplotlib⁴⁴).

Ex vivo trauma assessment with contrast-enhanced microcomputed tomography

Following euthanasia, the gerbil cochleae were extracted and immersed in a 1:2 Hafnium-substituted Wells–Dawson polyoxometalate (Hf-WD 1:2 POM), mixed with 3 mg/mL LiCl staining solution for a minimum of 24 h at room temperature^45,46. CECT imaging was performed with a Phoenix Nanotom M (GE Measurement and Control Solutions). The pixel size varies between 1.75 – 2.5 µm, the exposure time was 500 ms per frame, and 2400 frames were acquired over a 360° rotation. The scanning analysis and procedure in human temporal bones and gerbil cochleae are described in previous studies^14,28. Trauma volume segmentation and cochlear spiral length measurements were done using Avizo (Thermo Fisher Scientific). Trauma severity was classified based on imaging and electrochemical sensing. The proposed three-level scale seems to be effective for quantifying the trauma severity. Score 0 indicated no trauma, Score 1 represented structural damage without fistula formation, and Score 2 denoted severe trauma with fistula formation between the SV and ST or between ST and scala media (SM). Electrode insertion depth was determined by measuring the distance from the RWM entry point to the electrode tip. This distance was then divided by the total cochlear spiral length, measured along the lateral part of the OSL, and multiplied by 100 to show the proportion of insertion. Likewise, the proportion of trauma was calculated along the cochlear spiral length to compare with the insertion depth of the electrode.

Hydrogen peroxide biosensor fabrication and calibration in vitro

Hydrogen peroxide biosensors were prepared as described in Wellens et al.²⁸. 3 electrode CI arrays were provided by Cochlear Ltd. Electrodes were cleaned by sonication in 1% Hellmanex at 40 degrees for 5 min. Afterward, they were cleaned in a UV-ozone oven (UVO cleaner Model 342–220, Jelight) for 20 min on each side. A final cleaning step consisted of performing cyclic voltammetry in 0.5 M H₂SO₄ until a stable CV signal was obtained (-0.2 to 1.6 V, 50 mV/s). Poly O-phenylenediamine (PoPD) was then electrodeposited by cyclic voltammetry in 5 mM OPD in degassed phosphate-buffered saline (PBS) and afterward rinsed thoroughly with ultra-pure water (UPW). The CV parameters were:—0.3 to 0.9 V, 60 cycles, scan rate of 50 mV/s. Ag/AgCl was used as a reference electrode in all electrochemical measurements. We then deposited an antifouling layer on top based on polydopamine (PDA) and sulfobetaine as previously described ⁴⁷. Firstly, PDA was deposited by incubating the electrodes in a solution of 1 mg/ml dopamine hydrochloride and 1.7 mg of ammonium persulfate (APS) in PBS for 3.5 h. After rinsing with UPW, they were incubated in 1 mg/ml polyethyleneimine (PEI) in PBS overnight and then rinsed with UPW. Finally, a copolymer of SBMA and MA was conjugated to the amine groups of the PEI via EDC-NHS chemistry. This was done by incubating the electrodes in 1 mg/ml Poly(SBMA-co-MA) and 1 mM EDC-NHS in 50 mM MES buffer (pH = 6.5) for 4 h and rinsed with UPW afterward. A detailed description of Poly(SBMA-co-MA) synthesis can be found in Wellens et al.⁴⁷ Electrodes were stored at room temperature.

The calibration of electrodes was performed at room temperature in artificial perilymph. Artificial perilymph is composed of PBS supplemented with 2 mg/ml BSA, 396 μM Glycine, 175 μM phenylalanine and 40 μM proline⁴⁸. Hydrogen peroxide was measured via a voltage protocol modified from Weltin et al. ²⁷. Firstly, a potential of 0.7 V vs Ag/AgCl is applied to oxidize the Pt/Ir electrode surface for 1 s. Secondly, a potential of 0.55 V is applied for 3 s to catalyze the oxidation of hydrogen peroxide at the electrode surface. The current value at the end of the 3 s period is linearly correlated with the concentration of hydrogen peroxide and is used for the calibration of the amperometric sensor. Thirdly, a potential of -0.4 V is applied for 2 s to reduce the oxidized Pt surface back to its original state. Then the electrode is left on open circuit potential for 24 s, at this potential no net current flows across the electrode, allowing H₂O₂ to diffuse back to the surface and be replenished. In vivo hydrogen peroxide concentrations were calculated based on this calibration. Typically, for every calibration data point, the measurement was repeated 6–10 times, and the first measurement of this cycle is always discarded. As CI electrodes are handmade and costly, we opted to reuse CI electrodes for different animal experiments. Electrodes that were reused were cleaned, sensing layers were redeposited, and then electrodes were recalibrated as described above.

In vivo hydrogen peroxide sensor measurements

All measurements were performed with a Palmsens4 potentiostat in bipot mode (measuring 2 electrodes at the same time) except for the case when only 1 electrode was still functional. A Ag/AgCl reference electrode (MI-402, Microelectrodes Inc.) and Pt counter electrode (BASi MW-1032, Palmsens) were implanted subcutaneously. For determination of the hydrogen peroxide concentration in the cochlea at any time point, we took the average of the value measured with the 2 sensing electrodes, if both were still functional. Measurements were taken over a 2-h period after administration of trauma or after implantation if no trauma was induced. For statistical analysis, T-tests, correlation and ROC curve, the highest average H₂O₂ concentration measured on the 2 electrodes over this 2-h period was used.

Statistical analysis

The sample size (i.e. number of tested cochleae) was n = 9 in the case of trauma and n = 5 cochleae for control conditions, respectively. Specifically, this study included 5 cochleae with trauma score 0, 4 cochleae with score 1 and 5 cochleae with score 2. While more data was collected during this study, some data was excluded from the final analysis. This data primarily served to test sensor functionality, but due to missing complementary CECT data, they could not be incorporated into the trauma classification framework. Nevertheless, the increased sample size improved the statistical robustness of the analysis. Hydrogen peroxide levels were measured over a 2-h timeframe and the highest measured value over this time was used for statistical analysis. Measurements started as soon as possible after implantation (15–30 min). In case gerbils died before the 2-h timeframe, the highest measured hydrogen peroxide value before death was used. Differences between control and trauma conditions were assessed using Welch’s T-test. The normality of data sets was verified with Shapiro–Wilk normality test. Correlation of hydrogen peroxide concentration with volume of observed trauma was verified using Pearson’s correlation coefficient (PCC). Correlation of hydrogen peroxide concentration with trauma score was measured with Spearman’s correlation coefficient (SCC). To elucidate the most important contributor to the measured hydrogen peroxide level (volume of trauma or trauma scale), we performed a multiple linear regression analysis in Python using the statsmodels package. Trauma score was treated as a pseudo-continuous ordinal value. Encoding trauma classification as a dummy gave similar results and conclusions. The Predictive power of sensor readout was evaluated using a receiver operating characteristic curve (ROC). ROC analysis progressively shifts the decision threshold, in our case the hydrogen peroxide concentration, and at each level calculates the true positive rate (TPR), the proportion of trauma-induced cochleae whose hydrogen peroxide concentration exceeds the threshold, and the false positive rate (FPR), the proportion of control cochleae whose hydrogen peroxide concentration exceeds the same threshold. Plotting these paired rates forms the ROC curve. The area under this curve (AUC) provides a threshold-independent measure of discriminative performance with values close to one indicating almost perfect separation of traumatized and control cochleae, whereas values near 0.5 indicate performance no better than chance. The optimal cutoff value is determined by minimizing the value |1-TPR-FPR|. This is equivalent to finding the point that is closest to the “ideal” situation where the TPR is 1 and the FPR is 0. For more details see publication by Nahm et al.²⁹. All statistical analyses were performed using Origin 2023 software unless otherwise specified.

Author contributions

Lore Kerkhofs: Methodology, Validation, Formal analysis, Investigation, Data Curation, Writing – Original Draft, Visualisation. Visualisation. Jolan Wellens: Methodology, Validation, Formal analysis, Investigation, Data Curation, Writing – Original Draft, Visualisation. Olivier Deschaume: Writing – Review & Editing. Myles McLaughlin: Supervision, Writing – Review & Editing. Tristan Putzeys: Writing – Review & Editing, Supervision, Methodology. Carmen Bartic: Conceptualization, Methodology, Writing – Review & Editing, Supervision, Resource and Project coordination, Funding acquisition. Nicolas Verhaert: Conceptualization, Methodology, Writing – Review & Editing, Funding acquisition, Supervision, Resource and Project coordination. Data availability. Full datasets can be provided by Nicolas Verhaert (nicolas.verhaert@kuleuven.be) upon reasonable request.

Funding

This work was supported by the Research Foundation – Flanders (FWO) under grant numbers 11D5723N, G088619N, 1S64622N, 1804816N, and G0A9Q25N, as well as by the KU Leuven Research Council (Onderzoeksraad, KU Leuven) through grants IDN/21/021 and IDN/23/011.

Data availability

Full datasets can be provided by Nicolas Verhaert (nicolas.verhaert@kuleuven.be) upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.