Identification and segmentation of myelinated nerve fibers in a cross-sectional optical microscopic image using a deep learning model

Self-supervised segmentation and characterization of fiber bundles in anatomic tracing data

Anatomic tracing is the gold standard tool for delineating brain connections and for validating more recently developed imaging approaches such as diffusion MRI tractography. A key step in the analysis of data from tracer experiments is the careful, manual charting of fiber trajectories on histological sections. This is a very time-consuming process, which limits the amount of annotated tracer data that are available for validation studies. Thus, there is a need to accelerate this process by developing a method for computer-assisted segmentation. Such a method must be robust to the common artifacts in tracer data, including variations in the intensity of stained axons and background, as well as spatial distortions introduced by sectioning and mounting the tissue. The method should also achieve satisfactory performance using limited manually charted data for training. Here we propose the first deeplearning method, with a self-supervised loss function, for segmentation of fiber bundles on histological sections from macaque brains that have received tracer injections. We address the limited availability of manual labels with a semi-supervised training technique that takes advantage of unlabeled data to improve performance. We also introduce anatomic and across-section continuity constraints to improve accuracy. We show that our method can be trained on manually charted sections from a single case and segment unseen sections from different cases, with a true positive rate of ~0.80. We further demonstrate the utility of our method by quantifying the density of fiber bundles as they travel through different white-matter pathways. We show that fiber bundles originating in the same injection site have different levels of density when they travel through different pathways, a finding that can have implications for microstructure-informed tractography methods. The code for our method is available at https://github.com/v-sundaresan/fiberbundle_seg_tracing.

Prediction of lymphoma response to CAR T cells by deep learning-based image analysis

Clinical prognostic scoring systems have limited utility for predicting treatment outcomes in lymphomas. We therefore tested the feasibility of a deep-learning (DL)-based image analysis methodology on pre-treatment diagnostic computed tomography (dCT), low-dose CT (lCT), and 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) images and rule-based reasoning to predict treatment response to chimeric antigen receptor (CAR) T-cell therapy in B-cell lymphomas. Pre-treatment images of 770 lymph node lesions from 39 adult patients with B-cell lymphomas treated with CD19-directed CAR T-cells were analyzed. Transfer learning using a pre-trained neural network model, then retrained for a specific task, was used to predict lesion-level treatment responses from separate dCT, lCT, and FDG-PET images. Patient-level response analysis was performed by applying rule-based reasoning to lesion-level prediction results. Patient-level response prediction was also compared to prediction based on the international prognostic index (IPI) for diffuse large B-cell lymphoma. The average accuracy of lesion-level response prediction based on single whole dCT slice-based input was 0.82+0.05 with sensitivity 0.87+0.07, specificity 0.77+0.12, and AUC 0.91+0.03. Patient-level response prediction from dCT, using the "Majority 60%" rule, had accuracy 0.81, sensitivity 0.75, and specificity 0.88 using 12-month post-treatment patient response as the reference standard and outperformed response prediction based on IPI risk factors (accuracy 0.54, sensitivity 0.38, and specificity 0.61 (p = 0.046)). Prediction of treatment outcome in B-cell lymphomas from pre-treatment medical images using DL-based image analysis and rule-based reasoning is feasible. This approach can potentially provide clinically useful prognostic information for decision-making in advance of initiating CAR T-cell therapy.

Rapid, automated nerve histomorphometry through open-source artificial intelligence

We aimed to develop and validate a deep learning model for automated segmentation and histomorphometry of myelinated peripheral nerve fibers from light microscopic images. A convolutional neural network integrated in the AxonDeepSeg framework was trained for automated axon/myelin segmentation using a dataset of light-microscopic cross-sectional images of osmium tetroxide-stained rat nerves including various axonal regeneration stages. In a second dataset, accuracy of automated segmentation was determined against manual axon/myelin labels. Automated morphometry results, including axon diameter, myelin sheath thickness and g-ratio were compared against manual straight-line measurements and morphometrics extracted from manual labels with AxonDeepSeg as a reference standard. The neural network achieved high pixel-wise accuracy for nerve fiber segmentations with a mean (± standard deviation) ground truth overlap of 0.93 (± 0.03) for axons and 0.99 (± 0.01) for myelin sheaths, respectively. Nerve fibers were identified with a sensitivity of 0.99 and a precision of 0.97. For each nerve fiber, the myelin thickness, axon diameter, g-ratio, solidity, eccentricity, orientation, and individual x -and y-coordinates were determined automatically. Compared to manual morphometry, automated histomorphometry showed superior agreement with the reference standard while reducing the analysis time to below 2.5% of the time needed for manual morphometry. This open-source convolutional neural network provides rapid and accurate morphometry of entire peripheral nerve cross-sections. Given its easy applicability, it could contribute to significant time savings in biomedical research while extracting unprecedented amounts of objective morphologic information from large image datasets.

A simple and robust method for automating analysis of naïve and regenerating peripheral nerves

BACKGROUND

Manual axon histomorphometry (AH) is time- and resource-intensive, which has inspired many attempts at automation. However, there has been little investigation on implementation of automated programs for widespread use. Ideally such a program should be able to perform AH across imaging modalities and nerve states. AxonDeepSeg (ADS) is an open source deep learning program that has previously been validated in electron microscopy. We evaluated the robustness of ADS for peripheral nerve axonal histomorphometry in light micrographs prepared using two different methods.

METHODS

Axon histomorphometry using ADS and manual analysis (gold-standard) was performed on light micrographs of naïve or regenerating rat median nerve cross-sections prepared with either toluidine-resin or osmium-paraffin embedding protocols. The parameters of interest included axon count, axon diameter, myelin thickness, and g-ratio.

RESULTS

Manual and automatic ADS axon counts demonstrated good agreement in naïve nerves and moderate agreement on regenerating nerves. There were small but consistent differences in measured axon diameter, myelin thickness and g-ratio; however, absolute differences were small. Both methods appropriately identified differences between naïve and regenerating nerves. ADS was faster than manual axon analysis.

CONCLUSIONS

Without any algorithm retraining, ADS was able to appropriately identify critical differences between naïve and regenerating nerves and work with different sample preparation methods of peripheral nerve light micrographs. While there were differences between absolute values between manual and ADS, ADS performed consistently and required much less time. ADS is an accessible and robust tool for AH that can provide consistent analysis across protocols and nerve states.

Myelin detection in fluorescence microscopy images using machine learning

BACKGROUND

The myelin sheath produced by glial cells insulates the axons, and supports the function of the nervous system. Myelin sheath degeneration causes neurodegenerative disorders, such as multiple sclerosis (MS). There are no therapies for MS that promote remyelination. Drug discovery frequently involves screening thousands of compounds. However, this is not feasible for remyelination drugs, since myelin quantification is a manual labor-intensive endeavor. Therefore, the development of assistive software for expedited myelin detection is instrumental for MS drug discovery by enabling high-content image-based drug screens.

NEW METHOD

In this study, we developed a machine learning based expedited myelin detection approach in fluorescence microscopy images. Multi-channel three-dimensional microscopy images of a mouse stem cell-based myelination assay were labeled by experts. A spectro-spatial feature extraction method was introduced to represent local dependencies of voxels both in spatial and spectral domains. Feature extraction yielded two data set of over forty-seven thousand annotated images in total.

RESULTS

Myelin detection performances of 23 different supervised machine learning techniques including a customized-convolutional neural network (CNN), were assessed using various train/test split ratios of the data sets. The highest accuracy values of 98.84±0.09% and 98.46±0.11% were achieved by Boosted Trees and customized-CNN, respectively.

COMPARISON WITH EXISTING METHODS

Our approach can detect myelin in a common experimental setup. Myelin extending in any orientation in 3 dimensions is segmented from 3 channel z-stack fluorescence images.

CONCLUSIONS

Our results suggest that the proposed expedited myelin detection approach is a feasible and robust method for remyelination drug screening.

Predicting the impact of single nucleotide variants on splicing via sequence-based deep neural networks and genomic features

Single nucleotide mutations in exonic regions can significantly affect gene function through a disruption of splicing, and various computational methods have been developed to predict the splicing-related effects of a single nucleotide mutation. We implemented a new method using ensemble learning that combines two types of predictive models: (a) base sequence-based deep neural networks (DNNs) and (b) machine learning models based on genomic attributes. This method was applied to the Massively Parallel Splicing Assay challenge of the Fifth Critical Assessment of Genome Interpretation, in which challenge participants predicted various experimentally-defined exonic splicing mutations, and achieved a promising result. We successfully revealed that combining different predictive models based upon the stacked generalization method led to significant improvement in prediction performance. In addition, whereas most of the genomic features adopted in constructing machine learning models were previously reported, feature values generated with DSSP, a DNN-based splice site prediction tool, were novel and helpful for the prediction. Learning the sequence patterns associated with normal splicing and the change in splicing site probabilities caused by a mutation was presumed to be helpful in predicting splicing disruption.

Measurement-oriented deep-learning workflow for improved segmentation of myelin and axons in high-resolution images of human cerebral white matter

BACKGROUND

Standard segmentation of high-contrast electron micrographs (EM) identifies myelin accurately but does not translate easily into measurements of individual axons and their myelin, even in cross-sections of parallel fibers. We describe automated segmentation and measurement of each myelinated axon and its sheath in EMs of arbitrarily oriented human white matter from autopsies.

NEW METHODS

Preliminary segmentation of myelin, axons and background by machine learning, using selected filters, precedes automated correction of systematic errors. Final segmentation is done by a deep neural network (DNN). Automated measurement of each putative fiber rejects measures encountering pre-defined artifacts and excludes fibers failing to satisfy pre-defined conditions.

RESULTS

Improved segmentation of three sets of 30 annotated images each (two sets from human prefrontal white matter and one from human optic nerve) is achieved with a DNN trained only with a subset of the first set from prefrontal white matter. Total number of myelinated axons identified by the DNN differed from expert segmentation by 0.2%, 2.9%, and -5.1%, respectively. G-ratios differed by 2.96%, 0.74% and 2.83%. Intraclass correlation coefficients between DNN and annotated segmentation were mostly >0.9, indicating nearly interchangeable performance.

COMPARISON WITH EXISTING METHOD(S)

Measurement-oriented studies of arbitrarily oriented fibers from central white matter are rare. Published methods are typically applied to cross-sections of fascicles and measure aggregated areas of myelin sheaths and axons, allowing estimation only of average g-ratio.

CONCLUSIONS

Automated segmentation and measurement of axons and myelin is complex. We report a feasible approach that has so far proven comparable to manual segmentation.

Deep learning for high-throughput quantification of oligodendrocyte ensheathment at single-cell resolution

High-throughput quantification of oligodendrocyte myelination is a challenge that, if addressed, would facilitate the development of therapeutics to promote myelin protection and repair. Here, we established a high-throughput method to assess oligodendrocyte ensheathment in-vitro, combining nanofiber culture devices and automated imaging with a heuristic approach that informed the development of a deep learning analytic algorithm. The heuristic approach was developed by modeling general characteristics of oligodendrocyte ensheathments, while the deep learning neural network employed a UNet architecture and a single-cell training method to associate ensheathed segments with individual oligodendrocytes. Reliable extraction of multiple morphological parameters from individual cells, without heuristic approximations, allowed the UNet to match the accuracy of expert-human measurements. The capacity of this technology to perform multi-parametric analyses at the level of individual cells, while reducing manual labor and eliminating human variability, permits the detection of nuanced cellular differences to accelerate the discovery of new insights into oligodendrocyte physiology.

AxonDeepSeg: automatic axon and myelin segmentation from microscopy data using convolutional neural networks

Segmentation of axon and myelin from microscopy images of the nervous system provides useful quantitative information about the tissue microstructure, such as axon density and myelin thickness. This could be used for instance to document cell morphometry across species, or to validate novel non-invasive quantitative magnetic resonance imaging techniques. Most currently-available segmentation algorithms are based on standard image processing and usually require multiple processing steps and/or parameter tuning by the user to adapt to different modalities. Moreover, only a few methods are publicly available. We introduce AxonDeepSeg, an open-source software that performs axon and myelin segmentation of microscopic images using deep learning. AxonDeepSeg features: (i) a convolutional neural network architecture; (ii) an easy training procedure to generate new models based on manually-labelled data and (iii) two ready-to-use models trained from scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Results show high pixel-wise accuracy across various species: 85% on rat SEM, 81% on human SEM, 95% on mice TEM and 84% on macaque TEM. Segmentation of a full rat spinal cord slice is computed and morphological metrics are extracted and compared against the literature. AxonDeepSeg is freely available at https://github.com/neuropoly/axondeepseg .