Segmentation-Less, Automated, Vascular Vectorization

A Deep Learning Approach for Improving Two-Photon Vascular Imaging Speeds

A potential method for tracking neurovascular disease progression over time in preclinical models is multiphoton fluorescence microscopy (MPM), which can image cerebral vasculature with capillary-level resolution. However, obtaining high-quality, three-dimensional images with traditional point scanning MPM is time-consuming and limits sample sizes for chronic studies. Here, we present a convolutional neural network-based (PSSR Res-U-Net architecture) algorithm for fast upscaling of low-resolution or sparsely sampled images and combine it with a segmentation-less vectorization process for 3D reconstruction and statistical analysis of vascular network structure. In doing so, we also demonstrate that the use of semi-synthetic training data can replace the expensive and arduous process of acquiring low- and high-resolution training pairs without compromising vectorization outcomes, and thus open the possibility of utilizing such approaches for other MPM tasks where collecting training data is challenging. We applied our approach to images with large fields of view from a mouse model and show that our method generalizes across imaging depths, disease states and other differences in neurovasculature. Our pretrained models and lightweight architecture can be used to reduce MPM imaging time by up to fourfold without any changes in underlying hardware, thereby enabling deployability across a range of settings.

Pulse train gating to improve signal generation for in vivo two-photon fluorescence microscopy

Significance

Two-photon microscopy is used routinely for in vivo imaging of neural and vascular structures and functions in rodents with a high resolution. Image quality, however, often degrades in deeper portions of the cerebral cortex. Strategies to improve deep imaging are therefore needed. We introduce such a strategy using the gating of high repetition rate ultrafast pulse trains to increase the signal level.

Aim

We investigate how the signal generation, signal-to-noise ratio (SNR), and signal-to-background ratio (SBR) improve with pulse gating while imaging in vivo mouse cerebral vasculature.

Approach

An electro-optic modulator with a high-power (6 W) 80 MHz repetition rate ytterbium fiber amplifier is used to create gates of pulses at a 1 MHz repetition rate. We first measure signal generation from a Texas Red solution in a cuvette to characterize the system with no gating and at a 50%, 25%, and 12.5% duty cycle. We then compare the signal generation, SNR, and SBR when imaging Texas Red-labeled vasculature using these conditions.

Results

We find up to a 6.73-fold increase in fluorescent signal from a cuvette when using a 12.5% duty cycle pulse gating excitation pattern as opposed to a constant 80 MHz pulse train at the same average power. We verify similar increases for in vivo imaging to that observed in cuvette testing. For deep imaging, we find that pulse gating results in a 2.95-fold increase in the SNR and a 1.37-fold increase in the SBR on average when imaging mouse cortical vasculature at depths ranging from 950 to 1050    μ m .

Conclusions

We demonstrate that a pulse gating strategy can either be used to limit heating when imaging superficial brain regions or used to increase signal generation in deep regions. These findings should encourage others to adopt similar pulse gating excitation schemes for imaging neural structures through two-photon microscopy.

Pulse train gating to improve signal generation for in vivo two-photon fluorescence microscopy

Significance

Two-photon microscopy is used routinely for in vivo imaging of neural and vascular structure and function in rodents with a high resolution. Image quality, however, often degrades in deeper portions of the cerebral cortex. Strategies to improve deep imaging are therefore needed. We introduce such a strategy using gates of high repetition rate ultrafast pulse trains to increase signal level.

Aim

We investigate how signal generation, signal-to-noise ratio (SNR), and signal-to-background ratio (SBR) improve with pulse gating while imaging in vivo mouse cerebral vasculature.

Approach

An electro-optic modulator is used with a high-power (6 W) 80 MHz repetition rate ytterbium fiber amplifier to create gates of pulses at a 1 MHz repetition rate. We first measure signal generation from a Texas Red solution in a cuvette to characterize the system with no gating and at a 50%, 25%, and 12.5% duty cycle. We then compare signal generation, SNR, and SBR when imaging Texas Red-labeled vasculature using these conditions.

Results

We find up to a 6.73-fold increase in fluorescent signal from a cuvette when using a 12.5% duty cycle pulse gating excitation pattern as opposed to a constant 80 MHz pulse train. We verify similar increases for in vivo imaging to that observed in cuvette testing. For deep imaging we find pulse gating to result in a 2.95-fold increase in SNR and a 1.37-fold increase in SBR on average when imaging mouse cortical vasculature at depths ranging from 950 μm to 1050 μm.

Conclusions

We demonstrate that a pulse gating strategy can either be used to limit heating when imaging superficial brain regions or used to increase signal generation in deep regions. These findings should encourage others to adopt similar pulse gating excitation schemes for imaging neural structure through two-photon microscopy.

The application of the nnU-Net-based automatic segmentation model in assisting carotid artery stenosis and carotid atherosclerotic plaque evaluation

Objective: No new U-net (nnU-Net) is a newly-developed deep learning neural network, whose advantages in medical image segmentation have been noticed recently. This study aimed to investigate the value of the nnU-Net-based model for computed tomography angiography (CTA) imaging in assisting the evaluation of carotid artery stenosis (CAS) and atherosclerotic plaque. Methods: This study retrospectively enrolled 93 CAS-suspected patients who underwent head and neck CTA examination, then randomly divided them into the training set (N = 70) and the validation set (N = 23) in a 3:1 ratio. The radiologist-marked images in the training set were used for the development of the nnU-Net model, which was subsequently tested in the validation set. Results: In the training set, the nnU-Net had already displayed a good performance for CAS diagnosis and atherosclerotic plaque segmentation. Then, its utility was further confirmed in the validation set: the Dice similarity coefficient value of the nnU-Net model in segmenting background, blood vessels, calcification plaques, and dark spots reached 0.975, 0.974 0.795, and 0.498, accordingly. Besides, the nnU-Net model displayed a good consistency with physicians in assessing CAS (Kappa = 0.893), stenosis degree (Kappa = 0.930), the number of calcification plaque (Kappa = 0.922), non-calcification (Kappa = 0.768) and mixed plaque (Kappa = 0.793), as well as the max thickness of calcification plaque (intraclass correlation coefficient = 0.972). Additionally, the evaluation time of the nnU-Net model was shortened compared with the physicians (27.3 ± 4.4 s vs. 296.8 ± 81.1 s, p < 0.001). Conclusion: The automatic segmentation model based on nnU-Net shows good accuracy, reliability, and efficiency in assisting CTA to evaluate CAS and carotid atherosclerotic plaques.

Ultraliser: a framework for creating multiscale, high-fidelity and geometrically realistic 3D models for in silico neuroscience

Ultraliser is a neuroscience-specific software framework capable of creating accurate and biologically realistic 3D models of complex neuroscientific structures at intracellular (e.g. mitochondria and endoplasmic reticula), cellular (e.g. neurons and glia) and even multicellular scales of resolution (e.g. cerebral vasculature and minicolumns). Resulting models are exported as triangulated surface meshes and annotated volumes for multiple applications in in silico neuroscience, allowing scalable supercomputer simulations that can unravel intricate cellular structure-function relationships. Ultraliser implements a high-performance and unconditionally robust voxelization engine adapted to create optimized watertight surface meshes and annotated voxel grids from arbitrary non-watertight triangular soups, digitized morphological skeletons or binary volumetric masks. The framework represents a major leap forward in simulation-based neuroscience, making it possible to employ high-resolution 3D structural models for quantification of surface areas and volumes, which are of the utmost importance for cellular and system simulations. The power of Ultraliser is demonstrated with several use cases in which hundreds of models are created for potential application in diverse types of simulations. Ultraliser is publicly released under the GNU GPL3 license on GitHub (BlueBrain/Ultraliser).

SIGNIFICANCE

There is crystal clear evidence on the impact of cell shape on its signaling mechanisms. Structural models can therefore be insightful to realize the function; the more realistic the structure can be, the further we get insights into the function. Creating realistic structural models from existing ones is challenging, particularly when needed for detailed subcellular simulations. We present Ultraliser, a neuroscience-dedicated framework capable of building these structural models with realistic and detailed cellular geometries that can be used for simulations.

High-resolution three-dimensional blood flow tomography in the subdiffuse regime using laser speckle contrast imaging

SIGNIFICANCE

Visualizing high-resolution hemodynamics in cerebral tissue over a large field of view (FOV), provides important information in studying disease states affecting the brain. Current state-of-the-art optical blood flow imaging techniques either lack spatial resolution or are too slow to provide high temporal resolution reconstruction of flow map over a large FOV.

AIM

We present a high spatial resolution computational optical imaging technique based on principles of laser speckle contrast imaging (LSCI) for reconstructing the blood flow maps in complex tissue over a large FOV provided that the three-dimensional (3D) vascular structure is known or assumed.

APPROACH

Our proposed method uses a perturbation Monte Carlo simulation of the high-resolution 3D geometry for both accurately deriving the speckle contrast forward model and calculating the Jacobian matrix used in our reconstruction algorithm to achieve high resolution. Given the convex nature of our highly nonlinear problem, we implemented a mini-batch gradient descent with an adaptive learning rate optimization method to iteratively reconstruct the blood flow map. Specifically, we implemented advanced optimization techniques combined with efficient parallelization and vectorization of the forward and derivative calculations to make reconstruction of the blood flow map feasible with reconstruction times on the order of tens of minutes.

RESULTS

We tested our reconstruction algorithm through simulation of both a flow phantom model as well as an anatomically correct murine cerebral tissue and vasculature captured via two-photon microscopy. Additionally, we performed a noise study, examining the robustness of our inverse model in presence of 0.1% and 1% additive noise. In all cases, the blood flow reconstruction error was <2  %   for most of the vasculature, except for the peripheral vasculature which suffered from insufficient photon sampling. Descending vasculature and deeper structures showed slightly higher sensitivity to noise compared with vasculature with a horizontal orientation at the more superficial layers. Our results show high-resolution reconstruction of the blood flow map in tissue down to 500  μm and beyond.

CONCLUSIONS

We have demonstrated a high-resolution computational imaging technique for visualizing blood flow map in complex tissue over a large FOV. Once a high-resolution structural image is captured, our reconstruction algorithm only requires a few LSCI images captured through a camera to reconstruct the blood flow map computationally at a high resolution. We note that the combination of high temporal and spatial resolution of our reconstruction algorithm makes the solution well-suited for applications involving fast monitoring of flow dynamics over a large FOV, such as in functional neural imaging.

Improving the efficiency of plant root system phenotyping through digitization and automation

Root system architecture (RSA) determines unevenly distributed water and nutrient availability in soil. Genetic improvement of RSA, therefore, is related to crop production. However, RSA phenotyping has been carried out less frequently than above-ground phenotyping because measuring roots in the soil is difficult and labor intensive. Recent advancements have led to the digitalization of plant measurements; this digital phenotyping has been widely used for measurements of both above-ground and RSA traits. Digital phenotyping for RSA is slower and more difficult than for above-ground traits because the roots are hidden underground. In this review, we summarized recent trends in digital phenotyping for RSA traits. We classified the sample types into three categories: soil block containing roots, section of soil block, and root sample. Examples of the use of digital phenotyping are presented for each category. We also discussed room for improvement in digital phenotyping in each category.

Evaluation of resonant scanning as a high-speed imaging technique for two-photon imaging of cortical vasculature

We demonstrate a simple, low-cost two-photon microscope design with both galvo-galvo and resonant-galvo scanning capabilities. We quantify and compare the signal-to-noise ratios and imaging speeds of the galvo-galvo and resonant-galvo scanning modes when used for murine neurovascular imaging. The two scanning modes perform as expected under shot-noise limited detection and are found to achieve comparable signal-to-noise ratios. Resonant-galvo scanning is capable of reaching desired signal-to-noise ratios using less acquisition time when higher excitation power can be used. Given equal excitation power and total pixel dwell time between the two methods, galvo-galvo scanning outperforms resonant-galvo scanning in image quality when detection deviates from being shot-noise limited.

Deep Learning for Automated Analysis of Cellular and Extracellular Components of the Foreign Body Response in Multiphoton Microscopy Images

The Foreign body response (FBR) is a major unresolved challenge that compromises medical implant integration and function by inflammation and fibrotic encapsulation. Mice implanted with polymeric scaffolds coupled to intravital non-linear multiphoton microscopy acquisition enable multiparametric, longitudinal investigation of the FBR evolution and interference strategies. However, follow-up analyses based on visual localization and manual segmentation are extremely time-consuming, subject to human error, and do not allow for automated parameter extraction. We developed an integrated computational pipeline based on an innovative and versatile variant of the U-Net neural network to segment and quantify cellular and extracellular structures of interest, which is maintained across different objectives without impairing accuracy. This software for automatically detecting the elements of the FBR shows promise to unravel the complexity of this pathophysiological process.

Euclidean Distance Approximations From Replacement Product Graphs

We introduce a new chamfering paradigm, locally connecting pixels to produce path distances that approximate Euclidean space by building a small network (a replacement product) inside each pixel. These " RE -grid graphs" maintain near-Euclidean polygonal distance contours even in noisy data sets, making them useful tools for approximation when exact numerical solutions are unobtainable or impractical. The RE -grid graph creates a modular global architecture with lower pixel-to-pixel valency and simplified topology at the cost of increased computational complexity due to its internal structure. We present an introduction to chamfering replacement products with a number of case study examples to demonstrate the potential of these graphs for path-finding in high frequency and low resolution image spaces which motivate further study. Possible future applications include morphology, watershed segmentation, halftoning, neural network design, anisotropic image processing, image skeletonization, dendritic shaping, and cellular automata.