Three-dimensional models of cochlear implants: A review of their development and how they could support management and maintenance of cochlear implant performance

Parameterisation and Prediction of Intra-canal Cochlear Structures

Accurate 3D models of the cochlea are useful tools for research in the relationship between the electrode array and nerve fibres. The internal geometry of the cochlear canal plays an important role in understanding and quantifying that relationship. Predicting the location and shapes of the geometry is done by measuring histologic sections and fitting equations that can be used to predict parameters that fully define the geometry. A parameter sensitivity analysis is employed to prove that the size and location of the spiral lamina are the characteristics that most influence current distribution along target nerve fibres. The proposed landmark prediction method more accurately predicts the location of the points defining the spiral lamina in the apical region of the cochlea than methods used in previous modelling attempts. Thus, this technique can be used to generate 2D geometries that can be expanded to 3D models when high-resolution imaging is not available.

An analysis of the spread of electric field within the cochlea for different devices including custom-made electrodes for subtotal cochleoectomy

OBJECTIVE

Cochlear implants (CIs) can restore hearing not only in patients with profound hearing loss and deafness, but also in patients following tumour removal of intra-cochlear schwannomas. In such cases, design and placement differ from conventional electrode insertion, in which the cochlea remains filled with fluid. Despite these technical and surgical differences, previous studies have tended to show positive results in speech perception in tumour patients. The purpose of this study is to retrospectively evaluate the ability to predict speech recognition outcomes using individual electric field spreads and to investigate worldwide unique tumour cases.

STUDY DESIGN

In a retrospective analysis in a university tertiary center electric field spreads were compared between two groups of electrode designs implanted between 2009 and 2020 i.e., between lateral wall electrodes and custom-made perimodiolar electrode carriers from the same company. The voltage gradients were analysed and grouped with speech recognition results.

RESULTS

Differences in electrical field spreads were found between lateral wall electrodes and the custom-made perimodiolar electrodes, whereas a significant influence of electric fields on scores in speech recognition cannot be demonstrated.

CONCLUSION

Prediction of speech recognition outcome based on electric field propagation results seems not feasible. Significant differences in field spread between electrode arrays can be clearly demonstrated. This observation and its relevance to hearing treatment and speech recognition should therefore be further investigated in upcoming studies.

Models of Cochlea Used in Cochlear Implant Research: A Review

As the first clinically translated machine-neural interface, cochlear implants (CI) have demonstrated much success in providing hearing to those with severe to profound hearing loss. Despite their clinical effectiveness, key drawbacks such as hearing damage, partly from insertion forces that arise during implantation, and current spread, which limits focussing ability, prevent wider CI eligibility. In this review, we provide an overview of the anatomical and physical properties of the cochlea as a resource to aid the development of accurate models to improve future CI treatments. We highlight the advancements in the development of various physical, animal, tissue engineering, and computational models of the cochlea and the need for such models, challenges in their use, and a perspective on their future directions.

Optimizing Stimulus Energy for Cochlear Implants with a Machine Learning Model of the Auditory Nerve

Performing simulations with a realistic biophysical auditory nerve fiber model can be very time-consuming, due to the complex nature of the calculations involved. Here, a surrogate (approximate) model of such an auditory nerve fiber model was developed using machine learning methods, to perform simulations more efficiently. Several machine learning models were compared, of which a Convolutional Neural Network showed the best performance. In fact, the Convolutional Neural Network was able to emulate the behavior of the auditory nerve fiber model with extremely high similarity (R²>0.99), tested under a wide range of experimental conditions, whilst reducing the simulation time by five orders of magnitude. In addition, a method for randomly generating charge-balanced waveforms using hyperplane projection is introduced. In the second part of this paper, the Convolutional Neural Network surrogate model was used by an Evolutionary Algorithm to optimize the shape of the stimulus waveform in terms of energy efficiency. The resulting waveforms resemble a positive Gaussian-like peak, preceded by an elongated negative phase. When comparing the energy of the waveforms generated by the Evolutionary Algorithm with the commonly used square wave, energy decreases of 8%-45% were observed for different pulse durations. These results were validated with the original auditory nerve fiber model, which demonstrates that the proposed surrogate model can be used as its accurate and efficient replacement.

Insertion Guidance Based on Impedance Measurements of a Cochlear Electrode Array

The cochlear implantable neuromodulator provides substantial auditory perception to those with severe or profound impaired hearing. Correct electrode array positioning in the cochlea is one of the important factors for quality hearing, and misplacement may lead to additional injury to the cochlea. Visual inspection of the progress of electrode insertion is limited and mainly relies on the surgeon's tactile skills, and there is a need to detect in real-time the electrode array position in the cochlea during insertion. The available clinical measurement presently provides very limited information. Impedance measurement may be used to assist with the insertion of the electrode array. Using computational modeling of the cochlea, and its local tissue layers merging with the associated neuromodulator electrode array parameters, the impedance variations at different insertion depths and the proximities to the cochlea walls have been analyzed. In this study, an anatomical computational model of the temporal region of a patient is used to derive the relationship between impedance variations and the electrode proximity to the cochlea wall and electrode insertion depth. The aim was to examine whether the use of electrode impedance variations can be an effective marker of electrode proximity and electrode insertion depth. The proposed anatomical model simulates the quasi-static electrode impedance variations at different selected points but at considerable computation cost. A much less computationally intensive geometric model (~1/30) provided comparative impedance measurements with differences of <2%. Both use finite element analysis over the entire cross-section area of the scala tympani. It is shown that the magnitude of the impedance varies with both electrode insertion depth and electrode proximity to the adjacent anatomical layers (e.g., cochlea wall). In particular, there is a 1,400% increase when the electrode array is moved very close to the cochlea wall. This may help the surgeon to find the optimal electrode position within the scala tympani by observation of such impedance characteristics. The misplacement of the electrode array within the scala tympani may be eliminated by using the impedance variation metric during electrode array insertion if the results are validated with an experimental study.

A phenomenological computational model of the evoked action potential fitted to human cochlear implant responses

There is a growing interest in biomedical engineering in developing procedures that provide accurate simulations of the neural response to electrical stimulus produced by implants. Moreover, recent research focuses on models that take into account individual patient characteristics.

We present a phenomenological computational model that is customized with the patient’s data provided by the electrically evoked compound action potential (ECAP) for simulating the neural response to electrical stimulus produced by the electrodes of cochlear implants (CIs). The model links the input currents of the electrodes to the simulated ECAP.

Potentials and currents are calculated by solving the quasi-static approximation of the Maxwell equations with the finite element method (FEM). In ECAPs recording, an active electrode generates a current that elicits action potentials in the surrounding auditory nerve fibers (ANFs). The sum of these action potentials is registered by other nearby electrode. Our computational model emulates this phenomenon introducing a set of line current sources replacing the ANFs by a set of virtual neurons (VNs). To fit the ECAP amplitudes we assign a suitable weight to each VN related with the probability of an ANF to be excited. This probability is expressed by a cumulative beta distribution parameterized by two shape parameters that are calculated by means of a differential evolution algorithm (DE). Being the weights function of the current density, any change in the design of the CI affecting the current density produces changes in the weights and, therefore, in the simulated ECAP, which confers to our model a predictive capacity.

The results of the validation with ECAP data from two patients are presented, achieving a satisfactory fit of the experimental data with those provided by the proposed computational model.

The cochlea, found in the inner ear, is the organ where the sound is transformed into an electrical pulse to be transmitted by the neurons to the auditory cortex. Hearing loss can be caused by damage to the hair cells, in which case neuronal excitation is impaired. CIs are devices that replace the normal function of the impaired/damaged Organ of Corti. Computational models allow a better understanding of the mechanisms involved in the electrical stimulation of the auditory nerve. These models can help biomedical engineers to develop new CIs with improved auditory performance. One important aspect of our model is its customization with the patient’s data provided by the recording of the evoked compound action potential (the synchronous firing of a population of electrically stimulated auditory nerve fibers). This phenomenological model allows us to predict the registers of neural stimulation produced when the auditory nerve is stimulated with the CIs. We have validated the proposed model with real data obtained from two patients with CIs.

Apical Reference Stimulation: A Possible Solution to Facial Nerve Stimulation

OBJECTIVES

Postimplantation facial nerve stimulation is a common side-effect of intracochlear electrical stimulation. Facial nerve stimulation occurs when electric current intended to stimulate the auditory nerve, spread beyond the cochlea to excite the nearby facial nerve, causing involuntarily facial muscle contractions. Facial nerve stimulation can often be resolved through adjustments in speech processor fitting but, in some instances, these measures exhibit limited benefit or may have a detrimental effect on speech perception. In this study, apical reference stimulation mode was investigated as a potential intervention to facial nerve stimulation. Apical reference stimulation is a bipolar stimulation strategy in which the most apical electrode is used as the reference electrode for stimulation on all the other intracochlear electrodes.

DESIGN

A person-specific model of the human cochlea, facial nerve and electrode array, coupled with a neural model, was used to predict excitation of auditory and facial nerve fibers. These predictions were used to evaluate the effectiveness in reducing facial nerve stimulation using apical reference stimulation. Predictions were confirmed in psychoacoustic tests by determining auditory comfort and threshold levels for the apical reference stimulation mode while capturing electromyography data in two participants.

RESULTS

Models predicted a favorable outcome for apical reference stimulation, as facial nerve fiber thresholds were higher and auditory thresholds were lower, in direct comparison to conventional monopolar stimulation. Psychophysical tests also illustrated decreased auditory thresholds and increased dynamic range during apical reference stimulation. Furthermore, apical reference stimulation resulted in lower electromyography energy levels, compared to conventional monopolar stimulation, which suggests a reduction in facial nerve stimulation. Subjective feedback corroborated that apical reference stimulation alleviated facial nerve stimulation.

CONCLUSION

Apical reference stimulation may be a viable strategy to alleviate facial nerve stimulation considering the improvements in dynamic range and auditory thresholds, complemented with a reduction in facial nerve stimulation symptoms.

Facial nerve stimulation in a post-meningitic cochlear implant user: using computational modelling as a tool to probe mechanisms and progression of complications on a case-by-case basis

Facial nerve stimulation (FNS) is a side-effect of cochlear implantation that can result in severe discomfort for the user and essentially limits the optimal use of the implant. Three-dimensional cochlear implant modelling research has led to the progression from generic models to user-specific models with one of the intentions to develop model-based diagnostic tools. The objective of this study is to investigate the mechanisms that underlie the manifestation of FNS in the post-meningitic cochleae of a specific CI user through computational modelling. Bilateral models were created using a method previously developed for the construction of a three-dimensional user-specific volume conduction model of the cochlea and was expanded to include the facial nerve geometry. Reduced temporal bone density based on bone densitometry, cochlear duct ossification and degenerate auditory neural fibres were incorporated into a comprehensive FNS model. Auditory and facial nerve thresholds were predicted with the models showing good correspondence to perceptual thresholds and the user's FNS experience. Ossified cochlear ducts appear to aggravate the increase in thresholds caused by the otic capsule's decreased resistivity. This translational case study demonstrates the application of computational modelling as a clinical instrument in the assessment and management of complications with cochlear implantation.

Light sheet microscopy of the gerbil cochlea

Light sheet fluorescence microscopy (LSFM) provides a rapid and complete three-dimensional image of the cochlea. The method retains anatomical relationships-on a micrometer scale-between internal structures such as hair cells, basilar membrane (BM), and modiolus with external surface structures such as the round and oval windows. Immunolabeled hair cells were used to visualize the spiraling BM in the intact cochlea without time intensive dissections or additional histological processing; yet material prepared for LSFM could be rehydrated, the BM dissected out and reimaged at higher resolution with the confocal microscope. In immersion-fixed material, details of the cochlear vasculature were seen throughout the cochlea. Hair cell counts (both inner and outer) as well as frequency maps of the BM were comparable to those obtained by other methods, but with the added dimension of depth. The material provided measures of angular, linear, and vector distance between characteristic frequency regions along the BM. Thus, LSFM provides a unique ability to rapidly image the entire cochlea in a manner applicable to model and interpret physiological results. Furthermore, the three-dimensional organization of the cochlea can be studied at the organ and cellular level with LSFM, and this same material can be taken to the confocal microscope for detailed analysis at the subcellular level.

Computational Evaluation of Cochlear Implant Surgery Outcomes Accounting for Uncertainty and Parameter Variability

Cochlear implantation (CI) is a complex surgical procedure that restores hearing in patients with severe deafness. The successful outcome of the implanted device relies on a group of factors, some of them unpredictable or difficult to control. Uncertainties on the electrode array position and the electrical properties of the bone make it difficult to accurately compute the current propagation delivered by the implant and the resulting neural activation. In this context, we use uncertainty quantification methods to explore how these uncertainties propagate through all the stages of CI computational simulations. To this end, we employ an automatic framework, encompassing from the finite element generation of CI models to the assessment of the neural response induced by the implant stimulation. To estimate the confidence intervals of the simulated neural response, we propose two approaches. First, we encode the variability of the cochlear morphology among the population through a statistical shape model. This allows us to generate a population of virtual patients using Monte Carlo sampling and to assign to each of them a set of parameter values according to a statistical distribution. The framework is implemented and parallelized in a High Throughput Computing environment that enables to maximize the available computing resources. Secondly, we perform a patient-specific study to evaluate the computed neural response to seek the optimal post-implantation stimulus levels. Considering a single cochlear morphology, the uncertainty in tissue electrical resistivity and surgical insertion parameters is propagated using the Probabilistic Collocation method, which reduces the number of samples to evaluate. Results show that bone resistivity has the highest influence on CI outcomes. In conjunction with the variability of the cochlear length, worst outcomes are obtained for small cochleae with high resistivity values. However, the effect of the surgical insertion length on the CI outcomes could not be clearly observed, since its impact may be concealed by the other considered parameters. Whereas the Monte Carlo approach implies a high computational cost, Probabilistic Collocation presents a suitable trade-off between precision and computational time. Results suggest that the proposed framework has a great potential to help in both surgical planning decisions and in the audiological setting process.

Analysis of a purely conductance-based stochastic nerve fibre model as applied to compound models of populations of human auditory nerve fibres used in cochlear implant simulations

The study presents the application of a purely conductance-based stochastic nerve fibre model to human auditory nerve fibres within finite element volume conduction models of a semi-generic head and user-specific cochleae. The stochastic, threshold and temporal characteristics of the human model are compared and successfully validated against physiological feline results with the application of a mono-polar, bi-phasic, cathodic first stimulus. Stochastic characteristics validated include: (i) the log(Relative Spread) versus log(fibre diameter) distribution for the discharge probability versus stimulus intensity plots and (ii) the required exponential membrane noise versus transmembrane voltage distribution. Intra-user, and to a lesser degree inter-user, comparisons are made with respect to threshold and dynamic range at short and long pulse widths for full versus degenerate single fibres as well as for populations of degenerate fibres of a single user having distributed and aligned somas with varying and equal diameters. Temporal characteristics validated through application of different stimulus pulse rates and different stimulus intensities include: (i) discharge rate, latency and latency standard deviation versus stimulus intensity, (ii) period histograms and (iii) interspike interval histograms. Although the stochastic population model does not reduce the modelled single deterministic fibre threshold, the simulated stochastic and temporal characteristics show that it could be used in future studies to model user-specific temporally encoded information, which influences the speech perception of CI users.

Validation of a Cochlear Implant Patient-Specific Model of the Voltage Distribution in a Clinical Setting

Cochlear Implants (CIs) are medical implantable devices that can restore the sense of hearing in people with profound hearing loss. Clinical trials assessing speech intelligibility in CI users have found large intersubject variability. One possibility to explain the variability is the individual differences in the interface created between electrodes of the CI and the auditory nerve. In order to understand the variability, models of the voltage distribution of the electrically stimulated cochlea may be useful. With this purpose in mind, we developed a parametric model that can be adapted to each CI user based on landmarks from individual cone beam computed tomography (CBCT) scans of the cochlea before and after implantation. The conductivity values of each cochlea compartment as well as the weighting factors of different grounding modes have also been parameterized. Simulations were performed modeling the cochlea and electrode positions of 12 CI users. Three models were compared with different levels of detail: a homogeneous model (HM), a non-patient-specific model (NPSM), and a patient-specific model (PSM). The model simulations were compared with voltage distribution measurements obtained from the backward telemetry of the 12 CI users. Results show that the PSM produces the lowest error when predicting individual voltage distributions. Given a patient-specific geometry and electrode positions, we show an example on how to optimize the parameters of the model and how to couple it to an auditory nerve model. The model here presented may help to understand speech performance variability and support the development of new sound coding strategies for CIs.

The history and future of neural modeling for cochlear implants

This special issue of Network: Computation in Neural Systems on the topic of "Computational models of the electrically stimulated auditory system" incorporates review articles spanning a wide range of approaches to modeling cochlear implant stimulation of the auditory system. The purpose of this overview paper is to provide a historical context for the different modeling endeavors and to point toward how computational modeling could play a key role in the understanding, evaluation, and improvement of cochlear implants in the future.