Large-scale annotated dataset for cochlear hair cell detection and classification.

Our sense of hearing is mediated by cochlear hair cells, of which there are two types organized in one row of inner hair cells and three rows of outer hair cells. Each cochlea contains 5-15 thousand terminally differentiated hair cells, and their survival is essential for hearing as they do not regenerate after insult. It is often desirable in hearing research to quantify the number of hair cells within cochlear samples, in both pathological conditions, and in response to treatment. Machine learning can be used to automate the quantification process but requires a vast and diverse dataset for effective training. In this study, we present a large collection of annotated cochlear hair-cell datasets, labeled with commonly used hair-cell markers and imaged using various fluorescence microscopy techniques. The collection includes samples from mouse, rat, guinea pig, pig, primate, and human cochlear tissue, from normal conditions and following in-vivo and in-vitro ototoxic drug application. The dataset includes over 107,000 hair cells which have been identified and annotated as either inner or outer hair cells. This dataset is the result of a collaborative effort from multiple laboratories and has been carefully curated to represent a variety of imaging techniques. With suggested usage parameters and a well-described annotation procedure, this collection can facilitate the development of generalizable cochlear hair-cell detection models or serve as a starting point for fine-tuning models for other analysis tasks. By providing this dataset, we aim to give other hearing research groups the opportunity to develop their own tools with which to analyze cochlear imaging data more fully, accurately, and with greater ease.

Large-scale annotated dataset for cochlear hair cell detection and classification.

Our sense of hearing is mediated by cochlear hair cells, localized within the sensory epithelium called the organ of Corti. There are two types of hair cells in the cochlea, which are organized in one row of inner hair cells and three rows of outer hair cells. Each cochlea contains a few thousands of hair cells, and their survival is essential for our perception of sound because they are terminally differentiated and do not regenerate after insult. It is often desirable in hearing research to quantify the number of hair cells within cochlear samples, in both pathological conditions, and in response to treatment. However, the sheer number of cells along the cochlea makes manual quantification impractical. Machine learning can be used to overcome this challenge by automating the quantification process but requires a vast and diverse dataset for effective training. In this study, we present a large collection of annotated cochlear hair-cell datasets, labeled with commonly used hair-cell markers and imaged using various fluorescence microscopy techniques. The collection includes samples from mouse, human, pig and guinea pig cochlear tissue, from normal conditions and following in-vivo and in-vitro ototoxic drug application. The dataset includes over 90'000 hair cells, all of which have been manually identified and annotated as one of two cell types: inner hair cells and outer hair cells. This dataset is the result of a collaborative effort from multiple laboratories and has been carefully curated to represent a variety of imaging techniques. With suggested usage parameters and a well-described annotation procedure, this collection can facilitate the development of generalizable cochlear hair cell detection models or serve as a starting point for fine-tuning models for other analysis tasks. By providing this dataset, we aim to supply other groups within the hearing research community with the opportunity to develop their own tools with which to analyze cochlear imaging data more fully, accurately, and with greater ease.

SKOOTS: Skeleton oriented object segmentation for mitochondria.

The segmentation of individual instances of mitochondria from imaging datasets is informative, yet time-consuming to do by hand, sparking interest in developing automated algorithms using deep neural networks. Existing solutions for various segmentation tasks are largely optimized for one of two types of biomedical imaging: high resolution three-dimensional (whole neuron segmentation in volumetric electron microscopy datasets) or two-dimensional low resolution (whole cell segmentation of light microscopy images). The former requires consistently predictable boundaries to segment large structures, while the latter is boundary invariant but struggles with segmentation of large 3D objects without downscaling. Mitochondria in whole cell 3D EM datasets often occupy the challenging middle ground: large with ambiguous borders, limiting accuracy with existing tools. To rectify this, we have developed skeleton oriented object segmentation (SKOOTS); a new segmentation approach which efficiently handles large, densely packed mitochondria. We show that SKOOTS can accurately, and efficiently, segment 3D mitochondria in previously difficult situations. Furthermore, we will release a new, manually annotated, 3D mitochondria segmentation dataset. Finally, we show this approach can be extended to segment objects in 3D light microscopy datasets. These results bridge the gap between existing segmentation approaches and increases the accessibility for three-dimensional biomedical image analysis.

The hair cell analysis toolbox is a precise and fully automated pipeline for whole cochlea hair cell quantification.

Our sense of hearing is mediated by sensory hair cells, precisely arranged and highly specialized cells subdivided into outer hair cells (OHCs) and inner hair cells (IHCs). Light microscopy tools allow for imaging of auditory hair cells along the full length of the cochlea, often yielding more data than feasible to manually analyze. Currently, there are no widely applicable tools for fast, unsupervised, unbiased, and comprehensive image analysis of auditory hair cells that work well either with imaging datasets containing an entire cochlea or smaller sampled regions. Here, we present a highly accurate machine learning-based hair cell analysis toolbox (HCAT) for the comprehensive analysis of whole cochleae (or smaller regions of interest) across light microscopy imaging modalities and species. The HCAT is a software that automates common image analysis tasks such as counting hair cells, classifying them by subtype (IHCs versus OHCs), determining their best frequency based on their location along the cochlea, and generating cochleograms. These automated tools remove a considerable barrier in cochlear image analysis, allowing for faster, unbiased, and more comprehensive data analysis practices. Furthermore, HCAT can serve as a template for deep learning-based detection tasks in other types of biological tissue: With some training data, HCAT's core codebase can be trained to develop a custom deep learning detection model for any object on an image.

Components of impedance in a cochlear implant animal model with TGFß1-accelerated fibrosis.

Outcomes of cochlear implantation are likely influenced by the biological state of the cochlea. Fibrosis is a pathological change frequently seen in implanted ears. The goal of this work was to investigate the relationship between fibrosis and impedance. To that end, we employed an animal model of extensive fibrosis and tested whether aspects of impedance differed from controls. Specifically, an adenovirus with a TGF-ß1 gene insert (Ad.TGF-ß1) was injected into guinea pig scala tympani to elicit rapid onset fibrosis and investigate the relation between fibrosis and impedance. We found a significant correlation between treatment and rate of impedance increase. A physical circuit model of impedance was used to separate the effect of fibrosis from other confounding factors. Supported by preliminary, yet nonconclusive, electron microscopy data, this modeling suggested that deposits on the electrode surface are an important contributor to impedance change over time.

How electrically evoked compound action potentials in chronically implanted guinea pigs relate to auditory nerve health and electrode impedance.

This study examined how multiple measures based on the electrically evoked compound action potential (ECAP) amplitude-growth functions (AGFs) were related to estimates of neural [spiral ganglion neuron (SGN) density and cell size] and electrode impedance measures in 34 specific pathogen free pigmented guinea pigs that were chronically implanted (4.9-15.4 months) with a cochlear implant electrode array. Two interphase gaps (IPGs) were used for the biphasic pulses and the effect of the IPG on each ECAP measure was measured ("IPG effect"). When using a stimulus with a constant IPG, SGN density was related to the across-subject variance in ECAP AGF linear slope, peak amplitude, and N1 latency. The SGN density values also help to explain a significant proportion of variance in the IPG effect for AGF linear slope and peak amplitude measures. Regression modeling revealed that SGN density was the primary dependent variable contributing to across-subject variance for ECAP measures; SGN cell size did not significantly improve the fitting of the model. Results showed that simple impedance measures were weakly related to most ECAP measures but did not typically improve the fit of the regression model.

Effects of Electrode Location on Estimates of Neural Health in Humans with Cochlear Implants.

There are a number of psychophysical and electrophysiological measures that are correlated with SGN density in animal models, and these same measures can be performed in humans with cochlear implants (CIs). Thus, these measures are potentially applicable in humans for estimating the condition of the neural population (so called "neural health" or "cochlear health") at individual sites along the electrode array and possibly adjusting the stimulation strategy in the CI sound processor accordingly. Some measures used to estimate neural health in animals have included the electrically evoked compound potential (ECAP), psychophysical detection thresholds, and multipulse integration (MPI). With regard to ECAP measures, it has been shown that the change in the ECAP response as a function of increasing the stimulus interphase gap ("IPG Effect") also reflects neural density in implanted animals. These animal studies have typically been conducted using preparations in which the electrode was in a fixed position with respect to the neural population, whereas in human cochlear implant users, the position of individual electrodes varies widely within an electrode array and also across subjects. The current study evaluated the effects of electrode location in the implanted cochlea (specifically medial-lateral location) on various electrophysiological and psychophysical measures in eleven human subjects. The results demonstrated that some measures of interest, specifically ECAP thresholds, psychophysical detection thresholds, and ECAP amplitude-growth function (AGF) linear slope, were significantly related to the distances between the electrode and mid-modiolar axis (MMA). These same measures were less strongly related or not significantly related to the electrode to medial wall (MW) distance. In contrast, neither the IPG Effect for the ECAP AGF slope or threshold, nor the MPI slopes were significantly related to MMA or MW distance from the electrodes. These results suggest that "within-channel" estimates of neural health such as the IPG Effect and MPI slope might be more suitable for estimating nerve condition in humans for clinical application since they appear to be relatively independent of electrode position.

Changes over time in the electrically evoked compound action potential (ECAP) interphase gap (IPG) effect following cochlear implantation in Guinea pigs.

The electrically-evoked compound action potential (ECAP) is correlated with spiral ganglion neuron (SGN) density in cochlear implanted animals. In a previous study, we showed that ECAP amplitude growth function (AGF) linear slopes for stimuli with a constant interphase gap (IPG) changed significantly over time following implantation. Related studies have also shown that 1) IPG sensitivity for ECAP measures ("IPG Effect") is related to SGN density in animals and 2) the ECAP IPG Effect is related to speech recognition performance in humans with cochlear implants. The current study examined how the ECAP IPG Effect changed following cochlear implantation in four non-deafened guinea pigs with residual inner hair cells (IHCs) and 5 deafened, neurotrophin-treated guinea pigs. Simple impedances were measured on the same days as the ECAP measures. Generally, non-deafened implanted animals with higher SGN survival demonstrated higher ECAP AGF linear slope and peak amplitude values than the deafened, implanted guinea pigs. The ECAP IPG Effect for the AGF slopes and peak amplitudes was also larger in the hearing animals. The N1 latencies for a constant IPG were not different between groups, but the N1 latency IPG Effect was smaller in the non-deafened, implanted animals. Similar to previously reported results, ECAP measures using a fixed or changing IPG required as many as three months after implantation before a stable point could be calculated, but this was dependent on the animal and condition. For all ECAP measures most animals showed greater variance in the first 30 days post-implantation. Post-implantation changes in ECAPs and impedances were not correlated with one another. Results from this study are helpful for estimating the mechanisms underlying ECAP characteristics and have implications for clinical application of the ECAP measures in long-term human cochlear implant recipients. Specifically, these measures could help to monitor neural health over a period of time, or during a time of stability these measures could be used to help select electrode sites for activation in clinical programming.