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

VASCilia (Vision Analysis StereoCilia): A Napari Plugin for Deep Learning-Based 3D Analysis of Cochlear Hair Cell Stereocilia Bundles

Cochlear hair cells are essential for hearing, and their stereocilia bundles are critical for mechanotransduction. However, analyzing the 3D morphology of these bundles can be challenging due to their complex organization and the presence of other cellular structures in the tissue. To address this, we developed VASCilia (Vision Analysis StereoCilia), a Napari plugin suite that automates the analysis of 3D confocal microscopy datasets of phalloidin-stained cochlear hair cell bundles. VASCilia includes five deep learning-based models that streamline the analysis process, including: (1) Z-Focus Tracker (ZFT) for selecting relevant slices in a 3D image stack; (2) PCPAlignNet (Planar Cell Polarity Alignment Network) for automated orientation of image stacks; (3) a segmentation model for identifying and delineating stereocilia bundles; (4) a tonotopic Position Prediction tool; and (5) a classification tool for identifying hair cell subtypes. In addition, VASCilia provides automated computational tools and measurement capabilities. Using VASCilia, we found that the total actin content of stereocilia bundles (as measured by phalloidin staining) does not necessarily increase with bundle height, which is likely due to differences in stereocilia thickness and number. This novel biological finding demonstrates the power of VASCilia in facilitating detailed quantitative analysis of stereocilia. VASCilia also provides a user-friendly interface that allows researchers to easily navigate and use the tool, with the added capability to reload all their analyses for review or sharing purposes. We believe that VASCilia will be a valuable resource for researchers studying cochlear hair cell development and function, addressing a longstanding need in the hair cell research community for specialized deep learning-based tools capable of high-throughput image quantitation. We have released our code along with a manually annotated dataset that includes approximately 55 3D stacks featuring instance segmentation (https://github.com/ucsdmanorlab/Napari-VASCilia). This dataset comprises a total of 502 inner and 1,703 outer hair cell bundles annotated in 3D. As the first open-source dataset of its kind, we aim to establish a foundational resource for constructing a comprehensive atlas of cochlea hair cell images. Ultimately, this initiative will support the development of foundational models adaptable to various species, markers, and imaging scales to accelerate advances within the hearing research community.

PCP auto count: a novel Fiji/ImageJ plug-in for automated quantification of planar cell polarity and cell counting

Introdution: During development, planes of cells give rise to complex tissues and organs. The proper functioning of these tissues is critically dependent on proper inter- and intra-cellular spatial orientation, a feature known as planar cell polarity (PCP). To study the genetic and environmental factors affecting planar cell polarity, investigators must often manually measure cell orientations, which is a time-consuming endeavor. To automate cell counting and planar cell polarity data collection we developed a Fiji/ImageJ plug-in called PCP Auto Count (PCPA). Methods: PCPA analyzes binary images and identifies "chunks" of white pixels that contain "caves" of infiltrated black pixels. For validation, inner ear sensory epithelia including cochleae and utricles from mice were immunostained for βII-spectrin and imaged with a confocal microscope. Images were preprocessed using existing Fiji functionality to enhance contrast, make binary, and reduce noise. An investigator rated PCPA cochlear hair cell angle measurements for accuracy using a one to five agreement scale. For utricle samples, PCPA derived measurements were directly compared against manually derived angle measurements and the concordance correlation coefficient (CCC) and Bland-Altman limits of agreement were calculated. PCPA was also tested against previously published images examining PCP in various tissues and across various species suggesting fairly broad utility. Results: PCPA was able to recognize and count 99.81% of cochlear hair cells, and was able to obtain ideally accurate planar cell polarity measurements for at least 96% of hair cells. When allowing for a <10° deviation from "perfect" measurements, PCPA's accuracy increased to 98%-100% for all users and across all samples. When PCPA's measurements were compared with manual angle measurements for E17.5 utricles there was negligible bias (<0.5°), and a CCC of 0.999. Qualitative examination of example images of Drosophila ommatidia, mouse ependymal cells, and mouse radial progenitors revealed a high level of accuracy for PCPA across a variety of stains, tissue types, and species. Discussion: Altogether, the data suggest that the PCPA plug-in suite is a robust and accurate tool for the automated collection of cell counts and PCP angle measurements.

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.

PCP Auto Count: A Novel Fiji/ImageJ plug-in for automated quantification of planar cell polarity and cell counting

Background

During development, planes of cells give rise to complex tissues and organs. The proper functioning of these tissues is critically dependent on proper inter- and intra-cellular spatial orientation, a feature known as planar cell polarity (PCP). To study the genetic and environmental factors affecting planar cell polarity investigators must often manually measure cell orientations, which is a time-consuming endeavor.

Methodology

To automate cell counting and planar cell polarity data collection we developed a Fiji/ImageJ plug-in called PCP Auto Count (PCPA). PCPA analyzes binary images and identifies "chunks" of white pixels that contain "caves" of infiltrated black pixels. Inner ear sensory epithelia including cochleae (P4) and utricles (E17.5) from mice were immunostained for βII-spectrin and imaged on a confocal microscope. Images were preprocessed using existing Fiji functionality to enhance contrast, make binary, and reduce noise. An investigator rated PCPA cochlear angle measurements for accuracy using a 1-5 agreement scale. For utricle samples, we directly compared PCPA derived measurements against manually derived angle measurements using concordance correlation coefficients (CCC) and Bland-Altman limits of agreement. Finally, PCPA was tested against a variety of images copied from publications examining PCP in various tissues and across various species.

Results

PCPA was able to recognize and count 99.81% of cochlear hair cells (n = 1,1541 hair cells) in a sample set, and was able to obtain ideally accurate planar cell polarity measurements for over 96% of hair cells. When allowing for a <10° deviation from "perfect" measurements, PCPA's accuracy increased to >98%. When manual angle measurements for E17.5 utricles were compared, PCPA's measurements fell within -9 to +10 degrees of manually obtained mean angle measures with a CCC of 0.999. Qualitative examination of example images of Drosophila ommatidia, mouse ependymal cells, and mouse radial progenitors revealed a high level of accuracy for PCPA across a variety of stains, tissue types, and species. Altogether, the data suggest that the PCPA plug-in suite is a robust and accurate tool for the automated collection of cell counts and PCP angle measurements.

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.