Using Neurogram Similarity Index Measure (NSIM) to Model Hearing Loss and Cochlear Neural Degeneration
Ahsan, Sunil · 2025
Read Using Neurogram Similarity Index Measure (NSIM) to Model Hearing Loss and Cochlear Neural Degeneration by Ahsan, Sunil (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-12705.txt.
Introduction
In a 2025 arXiv paper, Ahsan J. Cheema and Sunil Puria model hearing loss with nerve response maps. Many people with or without loss still strain to hear in noise. Loss of nerve links in the inner ear may add to that strain and to mixed aid results. The authors score how alike two nerve maps are with a measure called NSIM.
In one test, the score mapped how well people with loss picked out speech sounds. In a second test, the score picked up harm from lost nerve links on its own. That reach marks the score as a possible sign of hidden nerve harm that needs no surgery to check.
Catalog record
| CREATOR | Ahsan, Sunil |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 21 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-12705
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Using Neurogram Similarity Index Measure (NSIM) to Model Hearing Loss and Cochlear Neural Degeneration Ahsan, Sunil Abstract Trouble hearing in noisy situations remains a common complaint for both individuals with hearing loss and individuals with normal hearing. This is hypothesized to arise due to condition called: cochlear neural degeneration (CND) which can also result in significant variabilities in hearing aids outcomes. This paper uses computational models of auditory periphery to simulate various hearing tasks. We present an objective method to quantify hearing loss and CND by comparing auditory nerve fiber responses using a Neurogram Similarity Index Measure (NSIM). Specifically study 1, shows that NSIM can be used to map performance of individuals with hearing loss on phoneme recognition task with reasonable accuracy. In the study 2, we show that NSIM is a sensitive measure that can also be used to capture the deficits resulting from CND and can be a candidate for noninvasive biomarker of auditory synaptopathy. Cheema Puria Harvard UniversityCambridge, USA Massachusetts Eye and Ear Infirmary (MEEI)Boston, USA Harvard Medical SchoolBoston, USA keywords neurogram similarity index measure, cochlear neural degeneration, synaptopathy, hearing aids, image-similarity † † email: [email protected], [email protected] 1 Introduction Difficulty understanding speech in the presence of background noise is one of the most common complaints of patients with sensorineural hearing loss (SNHL) [ 1 ] . Extensive research has shown that the death of cochlear hair cells leads to hearing loss, but it is often preceded by a loss of the synapses linking the hair cells to the auditory-nerve fibers (cochlear neural degeneration). Cochlear neural degeneration (CND) can contribute to difficulty hearing in noise and goes undetected during standard audiometric evaluations [ 2 ] . Hearing aids (HAs) remain the standard treatment for hearing loss, but they primarily address hearing loss due to death of cochlear cells, leaving neural deficits untreated. This limits the effectiveness of hearing aids and can result in variability in patient outcomes. Therefore, development of an objective measure to study the effects of hearing loss and CND remains an essential challenge which can help improve the design and gain compensation strategies for hearing aids. One of the promising approaches to arrive at an objective measures for hearing loss and CND is to simulate the response of auditory nerve fibers (ANF) from a normal cochlea and a cochlea with hearing loss using physiologically inspired auditory nerve model (ANM) of cochlea [ 3 ] and compare differences between the discharge patterns for the two cases. In order to study the differences in discharge patterns of ANF Hines and Harte [ 4 , 5 ] developed a ”Neurogram Similarity Index Measure (NSIM)” which is similar to Structural Similarity Index Measure (SSIM) used widely in the image processing literature [ 6 ] . This measure views the time-frequency discharge patterns from auditory nerve fibers as an image pattern called a ‘neurogram’ and can be used to quantify the differences between ANFs from a normal hearing cochlea with no hearing loss and a cochlea with hearing loss. Previous studies have established NSIM as a quantitative measure to simulate performance of normal hearing individuals on /CVC/ phoneme recognition tasks for a range input stimulus levels [ 5 ] . Mamun et. al. [ 7 ] improved the dynamic range for neurogram based measures using Neurogram orthogonal polynomial to map performance of normal hearing individuals, but this study only included 1 hearing loss individual and thus was not generalizable to study the effect of hearing loss. No previous study has comprehensively (on a large enough datasets of individuals with hearing loss) explored the utility of neurogram based objective measures to study the effects of hearing loss on ANF responses and link that to the performance on speech recognition tasks. Similarly, the effects of CND on neurogram based metrics has also not been explored in previous literature. In this work we present two studies that evaluate NSIM as a metric. In Study 1, NSIM is used to evaluate performance of individuals with varying degrees of hearing loss. In Study 2, we use NSIM to model varying degrees of CND. 2 Methods 2.1 Phenomenological Model of Cochlea and Auditory Nerve Fiber Response This study used ANM [ 3 ] to simulate the responses of ANF and generate neurogram for a given sound (.wav file) and for a specific hearing loss profile. The model allows the simulation of hearing loss by decreasing OHC gain and making the frequency response of the cochlear filters broader [ 3 ] . The output of the ANM is in the form of post stimulus histogram (PSTH) which represents the spiking activity of the auditory nerve fibers in time bins. Each time bin in the PSTH can be converted to frequency response and generate a neurogram, which maps the spiking activity on a time-frequency map. The ANM allowed simulation of the three different types of auditory nerve fibers: 1) Low Spontaneous (LS) rate fibers with high thresholds, 2) Medium Spontaneous (MS) rate Fibers with medium thresholds, and 3) High Spontaneous (HS) rate fibers with low thresholds. The summed spiking activity of all fibers can be used to get one neurogram that represents activity for all three ANF types in response to an acoustic stimulus, or three separate neurograms representing each ANF fiber types. Additionally, to capture long term and short-term properties of auditory nerve, the model allows control of the time bins used for the PSTH which in turn allows generation of fine timing (FT) neurograms and mean rate (MR) neurograms. 2.2 Neurogram Similarity Index Measure: To quantify the similarity between the normal hearing neurograms and degraded hearing neurograms (neurograms from a hearing loss cochlea), methods previously developed [ 6 , 4 ] were used to calculated NSIM. The image processing based method Structural Similarity Index Measure (SSIM) compares luminance(l), contrast(c) and structure(s) between a reference image and degraded image using a suitable window size [ 6 ] using Eq. 1 . The constants \alpha , \beta , and \gamma are the weighting coefficients that are applied to weigh the different components of the similarity equation. The coefficients C_{1},C_{2},C_{3} have negligible effect on the results and are added to prevent unstable results at boundaries. Hines et. al [ 5 ] applied the same method and derived an equivalent similarity measure for comparing neurograms. They showed that for the neurograms, in the Eq. 1 contrast and component weightings have negligible effect. This reduces Eq. 1 to Eq. 2 which is then calculated over a 3X3 gaussian window of radius 0.5 [ 5 , 8 ] and averaged over all the points in the neurogram to calculate the NSIM (eq. 3 ). The general design of experiments is given in Fig. 1 . SSI(r,d)=\left[l(r,d)\right]^{\alpha}\cdot\left[c(r,d)\right]^{\beta}\cdot\left[s(r,d)\right]^{\gamma} (1) where: l(r,d)=\frac{2\mu_{r}\mu_{d}+C_{1}}{\mu_{r}^{2}+\mu_{d}^{2}+C_{1}},\\ c(r,d)=\frac{2\sigma_{rd}+C_{2}}{\sigma_{r}^{2}+\sigma_{d}^{2}+C_{2}},\\ s(r,d)=\frac{\sigma_{rd}+C_{3}}{\sigma_{r}\sigma_{d}+C_{3}},\\ \mu_{r}=\text{mean of reference image intensity over window}, \mu_{d}=\text{mean of degraded image intensity over window}, \sigma_{r}=\text{standard deviation of reference image over window}, \sigma_{r}^{2}=\text{variance of reference image over window}, \sigma_{d}=\text{standard deviation of degraded image over window}, \sigma_{d}^{2}=\text{variance of degraded image over window}, \sigma_{rd}=\text{covariance matrix over the window}, C_{1}=0.01L,\ \ \ C_{2}=(0.03L)^{2},\ \ \ C_{3}=\dfrac{C_{2}}{2},\ \ \ L=intensity\ range NSI(r,d)=\frac{2\mu_{r}\mu_{d}+C_{1}}{\mu_{r}^{2}+\mu_{d}^{2}+C_{1}}\cdot\frac{\sigma_{rd}+C_{3}}{\sigma_{r}\sigma_{d}+C_{3}} (2) \text{NSIM}(r,d)=\frac{1}{N\cdot M}\sum_{f=1}^{N}\sum_{t=1}^{M}\text{NSI}(r,d) (3) N=total\ frequency\ bins,\\ \ M=total\ time\ bins Figure 1: Design of experiment to simulate Neurogram Similarity Index Measure 2.3 Study