Semi-automated atlas-based analysis of brain histological sections

DLATA: Deep Learning-Assisted transformation alignment of 2D brain slice histology

Accurate alignment of brain slices is crucial for the classification of neuron populations by brain region, and for quantitative analysis in in vitro brain studies. Current semi-automated alignment workflows require labor intensive labeling of feature points on each slice image, which is time-consuming. To speed up the process in large-scale studies, we propose a method called Deep Learning-Assisted Transformation Alignment (DLATA), which uses deep learning to automatically identify feature points in images after training on a few labeled samples. DLATA only requires approximately 10% of the sample size of other semi-automated alignment workflows. Following feature point recognition, local weighted mean method is used as a geometrical transformation to align slice images for registration, achieving better results with about 4 fewer pixels of error than other semi-automated alignment workflows. DLATA can be retrained and successfully applied to the alignment of other biological tissue slices with different stains, including the typically challenging fluorescent stains. Reference codes and trained models for Nissl-stained coronal brain slices of mice can be found at https://github.com/ALIGNMENT2023/DLATA.

Scaling up cell-counting efforts in neuroscience through semi-automated methods

Quantifying how the cellular composition of brain regions vary across development, aging, sex, and disease, is crucial in experimental neuroscience, and the accuracy of different counting methods is continuously debated. Due to the tedious nature of most counting procedures, studies are often restricted to one or a few brain regions. Recently, there have been considerable methodological advances in combining semi-automated feature extraction with brain atlases for cell quantification. Such methods hold great promise for scaling up cell-counting efforts. However, little focus has been paid to how these methods should be implemented and reported to support reproducibility. Here, we provide an overview of practices for conducting and reporting cell counting in mouse and rat brains, showing that critical details for interpretation are typically lacking. We go on to discuss how novel methods may increase efficiency and reproducibility of cell counting studies. Lastly, we provide practical recommendations for researchers planning cell counting.

Machine learning based analysis of stroke lesions on mouse tissue sections

An unbiased, automated and reliable method for analysis of brain lesions in tissue after ischemic stroke is missing. Manual infarct volumetry or by threshold-based semi-automated approaches is laborious, and biased to human error or biased by many false -positive and -negative data, respectively. Thereby, we developed a novel machine learning, atlas-based method for fully automated stroke analysis in mouse brain slices stained with 2% Triphenyltetrazolium-chloride (2% TTC), named "StrokeAnalyst", which runs on a user-friendly graphical interface. StrokeAnalyst registers subject images on a common spatial domain (a novel mouse TTC- brain atlas of 80 average mathematical images), calculates pixel-based, tissue-intensity statistics (z-scores), applies outlier-detection and machine learning (Random-Forest) models to increase accuracy of lesion detection, and produces volumetry data and detailed neuroanatomical information per lesion. We validated StrokeAnalyst in two separate experimental sets using the filament stroke model. StrokeAnalyst detects stroke lesions in a rater-independent and reproducible way, correctly detects hemispheric volumes even in presence of post-stroke edema and significantly minimizes false-positive errors compared to threshold-based approaches (false-positive rate 1.2-2.3%, p < 0.05). It can process scanner-acquired, and even smartphone-captured or pdf-retrieved images. Overall, StrokeAnalyst surpasses all previous TTC-volumetry approaches and increases quality, reproducibility and reliability of stroke detection in relevant preclinical models.

A Semi-Automated Workflow for Brain Slice Histology Alignment, Registration, and Cell Quantification (SHARCQ)

Tools for refined cell-specific targeting have significantly contributed to understanding the characteristics and dynamics of distinct cellular populations by brain region. While advanced cell-labeling methods have accelerated the field of neuroscience, specifically in brain mapping, there remains a need to quantify and analyze the data. Here, by modifying a toolkit that localizes electrodes to brain regions (SHARP-Track; Slice Histology Alignment, Registration, and Probe-Track analysis), we introduce a post-imaging analysis tool to map histological images to established mouse brain atlases called SHARCQ (Slice Histology Alignment, Registration, and Cell Quantification). The program requires MATLAB, histological images, and either a manual or automatic cell count of the unprocessed images. SHARCQ simplifies the post-imaging analysis pipeline with a step-by-step GUI. We demonstrate that SHARCQ can be applied for a variety of mouse brain images, regardless of histology technique. In addition, SHARCQ rectifies discrepancies in mouse brain region borders between atlases by allowing the user to select between the Allen Brain Atlas or the digitized and modified Franklin-Paxinos Atlas for quantifying cell counts by region. SHARCQ produces quantitative and qualitative data, including counts of brain-wide region populations and a 3D model of registered cells within the atlas space. In summary, SHARCQ was designed as a neuroscience post-imaging analysis tool for cell-to-brain registration and quantification with a simple, accessible interface. All code is open-source and available for download (https://github.com/wildrootlab/SHARCQ).

Localizing Microemboli within the Rodent Brain through Block-Face Imaging and Atlas Registration

Brain microinfarcts are prevalent in humans, however because of the inherent difficulty of identifying and localizing individual microinfarcts, brain-wide quantification is impractical. In mice, microinfarcts have been created by surgically introducing microemboli into the brain, but a major limitation of this model is the absence of automated methods to identify and localize individual occlusions. We present a novel and semi-automated workflow to identify the anatomic location of fluorescent emboli (microspheres) within the mouse brain through histologic processing and atlas registration. By incorporating vibratome block-face imaging with the QuickNII brain registration tool, we show that the anatomic location of microspheres can be accurately registered to brain structures within the Allen mouse brain (AMB) atlas (e.g., somatomotor areas, hippocampal region, visual areas, etc.). Compared with registering images of slide mounted sections to the AMB atlas, microsphere location was more accurately determined when block-face images were used. As a proof of principle, using this workflow we compared the distribution of microspheres within the brains of mice that were either perfused or immersion fixed. No significant effect of perfusion on total microsphere number or location was detected. In general, microspheres were distributed brain-wide, with the largest density found in the thalamus. In sum, our block-face imaging workflow enables efficient characterization of the widespread distribution of fluorescent microemboli, facilitating future investigation into the impact of microinfarct load and location on brain health.

QUINT: Workflow for Quantification and Spatial Analysis of Features in Histological Images From Rodent Brain

Transgenic animal models are invaluable research tools for elucidating the pathways and mechanisms involved in the development of neurodegenerative diseases. Mechanistic clues can be revealed by applying labelling techniques such as immunohistochemistry or in situ hybridisation to brain tissue sections. Precision in both assigning anatomical location to the sections and quantifying labelled features is crucial for output validity, with a stereological approach or image-based feature extraction typically used. However, both approaches are restricted by the need to manually delineate anatomical regions. To circumvent this limitation, we present the QUINT workflow for quantification and spatial analysis of labelling in series of rodent brain section images based on available 3D reference atlases. The workflow is semi-automated, combining three open source software that can be operated without scripting knowledge, making it accessible to most researchers. As an example, a brain region-specific quantification of amyloid plaques across whole transgenic Tg2576 mouse brain series, immunohistochemically labelled for three amyloid-related antigens is demonstrated. First, the whole brain image series were registered to the Allen Mouse Brain Atlas to produce customised atlas maps adapted to match the cutting plan and proportions of the sections (QuickNII software). Second, the labelling was segmented from the original images by the Random Forest Algorithm for supervised classification (ilastik software). Finally, the segmented images and atlas maps were used to generate plaque quantifications for each region in the reference atlas (Nutil software). The method yielded comparable results to manual delineations and to the output of a stereological method. While the use case demonstrates the QUINT workflow for quantification of amyloid plaques only, the workflow is suited to all mouse or rat brain series with labelling that is visually distinct from the background, for example for the quantification of cells or labelled proteins.

DeNeRD: high-throughput detection of neurons for brain-wide analysis with deep learning

Mapping the structure of the mammalian brain at cellular resolution is a challenging task and one that requires capturing key anatomical features at the appropriate level of analysis. Although neuroscientific methods have managed to provide significant insights at the micro and macro level, in order to obtain a whole-brain analysis at a cellular resolution requires a meso-scopic approach. A number of methods can be currently used to detect and count cells, with, nevertheless, significant limitations when analyzing data of high complexity. To overcome some of these constraints, we introduce a fully automated Artificial Intelligence (AI)-based method for whole-brain image processing to Detect Neurons in different brain Regions during Development (DeNeRD). We demonstrate a high performance of our deep neural network in detecting neurons labeled with different genetic markers in a range of imaging planes and imaging modalities.

Atlas-based imaging data analysis tool for quantitative mouse brain histology (AIDAhisto)

Cell counting in neuroscience is a routine method of utmost importance to support descriptive in vivo findings with quantitative data on the cellular level. Although known to be error- and bias-prone, manual cell counting of histological stained brain slices remains the gold standard in the field. While the manual approach is limited to small regions-of-interest in the brain, automated tools are needed to up-scale translational approaches and generate whole mouse brain counts in an atlas framework. Our goal was to develop an algorithm which requires no pre-training such as machine learning algorithms, only minimal user input, and adjustable variables to obtain reliable cell counting results for stitched mouse brain slices registered to a common atlas such as the Allen Mouse Brain atlas. We adapted filter banks to extract the maxima from round-shaped cell nuclei and various cell structures. In a qualitative as well as quantitative comparison to other tools and two expert raters, AIDAhisto provides accurate and fast results for cell nuclei as well as immunohistochemical stainings of various types of cells in the mouse brain.

An automated mapping method for Nissl-stained mouse brain histologic sections

BACKGROUND

Histologic evaluation of the central nervous system is often a critical endpoint in in vivo efficacy studies, and is considered the essential component of neurotoxicity assessment in safety studies. Automated image analysis is a powerful tool that can radically reduce the workload associated with evaluating brain histologic sections.

NEW METHOD

We developed an automated brain mapping method that identifies neuroanatomic structures in mouse histologic coronal brain sections. The method utilizes the publicly available Allen Brain Atlas to map brain regions on digitized Nissl-stained sections.

RESULTS

The method's accuracy was first assessed by comparing the mapping results to structure delineations from the Franklin and Paxinos (FP) mouse brain atlas. Brain regions mapped from FP Nissl-stained sections and calculated volumes were similar to structure delineations and volumes derived from corresponding FP illustrations. We subsequently applied our method to mouse brain sections from an in vivo study where the hippocampus was the structure of interest. Nissl-stained sections were mapped and hippocampal boundaries transferred to adjacent immunohistochemically stained sections. Optical density quantification results were comparable to those from time-consuming, manually drawn hippocampal delineations on the IHC-stained sections.

COMPARISON WITH EXISTING METHODS

Compared to other published methods, our method requires less manual input, and has been validated comprehensively using a secondary atlas, as well as manually annotated brain IHC sections from 68 study mice.

CONCLUSIONS

We propose that our automated brain mapping method enables greater efficiency and consistency in mouse neuropathologic assessments.

Cortical and Subcortical Contributions to Short-Term Memory for Orienting Movements

Neural activity in frontal cortical areas has been causally linked to short-term memory (STM), but whether this activity is necessary for forming, maintaining, or reading out STM remains unclear. In rats performing a memory-guided orienting task, the frontal orienting fields in cortex (FOF) are considered critical for STM maintenance, and during each trial display a monotonically increasing neural encoding for STM. Here, we transiently inactivated either the FOF or the superior colliculus and found that the resulting impairments in memory-guided orienting performance followed a monotonically decreasing time course, surprisingly opposite to the neural encoding. A dynamical attractor model in which STM relies equally on cortical and subcortical regions reconciled the encoding and inactivation data. We confirmed key predictions of the model, including a time-dependent relationship between trial difficulty and perturbability, and substantial, supralinear, impairment following simultaneous inactivation of the FOF and superior colliculus during memory maintenance.

Mapping social behavior-induced brain activation at cellular resolution in the mouse

Understanding how brain activation mediates behaviors is a central goal of systems neuroscience. Here, we apply an automated method for mapping brain activation in the mouse in order to probe how sex-specific social behaviors are represented in the male brain. Our method uses the immediate-early-gene c-fos, a marker of neuronal activation, visualized by serial two-photon tomography: the c-fos-GFP+ neurons are computationally detected, their distribution is registered to a reference brain and a brain atlas, and their numbers are analyzed by statistical tests. Our results reveal distinct and shared female and male interaction-evoked patterns of male brain activation representing sex discrimination and social recognition. We also identify brain regions whose degree of activity correlates to specific features of social behaviors and estimate the total numbers and the densities of activated neurons per brain areas. Our study opens the door to automated screening of behavior-evoked brain activation in the mouse.

Spatial organization of fibroblast nuclear chromocenters: component tree analysis

The nuclei of mouse connective tissue fibroblasts contain chromocenters which are well-defined zones of heterochromatin that can be used as positional landmarks to examine nuclear remodeling in response to a mechanical perturbation. This study used component tree analysis, an image segmentation algorithm that detects high intensity voxels that are topologically connected, to quantify the spatial organization of chromocenters in fibroblasts within whole mouse connective tissue fixed and stained with 4',6-diamidino-2-phenylindole (DAPI). The component tree analysis method was applied to confocal microscopy images of whole mouse areolar connective tissue incubated for 30 min ex vivo with or without static stretch. In stretched tissue, the mean distance between chromocenters within fibroblast nuclei was significantly greater (vs. non-stretched, P < 0.001), corresponding to an average of a 500-nm increase in chromocenter separation (~10% strain). There was no significant difference in chromocenter number or average size between stretch and no stretch. Average chromocenter distance was positively correlated with nuclear cross-sectional area (r = 0.78, P < 0.0001), and nuclear volume (r = 0.42, P < 0.0001), and negatively correlated with nuclear aspect ratio (r = -0.65, P < 0.0001) and nuclear concavity index (r = -0.44, P < 0.0001). These results demonstrate that component trees can be successfully applied to the morphometric analysis of nuclear chromocenters in fibroblasts within whole connective tissue. Static stretching of mouse areolar connective tissue for 30 min resulted in substantially increased separation of nuclear chromocenters in connective tissue fibroblasts. This interior remodeling of the nucleus induced by tissue stretch may impact transcriptionally active euchromatin within the inter-chromocenter space.

In situ characterization of the brain-microdevice interface using device-capture histology

Accurate assessment of brain-implantable microdevice bio-integration remains a formidable challenge. Prevailing histological methods require device extraction prior to tissue processing, often disrupting and removing the tissue of interest which had been surrounding the device. The Device-Capture Histology method, presented here, overcomes many limitations of the conventional Device-Explant Histology method, by collecting the device and surrounding tissue intact for subsequent labeling. With the implant remaining in situ, accurate and precise imaging of the morphologically preserved tissue at the brain/microdevice interface can then be collected and quantified. First, this article presents the Device-Capture Histology method for obtaining and processing the intact, undisturbed microdevice-tissue interface, and imaging using fluorescent labeling and confocal microscopy. Second, this article gives examples of how to quantify features found in the captured peridevice tissue. We also share histological data capturing (1) the impact of microdevice implantation on tissue, (2) the effects of an experimental anti-inflammatory coating, (3) a dense grouping of cell nuclei encapsulating a long-term implant, and (4) atypical oligodendrocyte organization neighboring a long term implant. Data sets collected using the Device-Capture Histology method are presented to demonstrate the significant advantages of processing the intact microdevice-tissue interface, and to underscore the utility of the method in understanding the effects of the brain-implantable microdevices on nearby tissue.