Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 543 - 552
Опубликована: Янв. 1, 2024
Язык: Английский
bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2022, Номер unknown
Опубликована: Ноя. 24, 2022
Abstract Serial section transmission electron microscopy (TEM) has proven to be one of the leading methods for millimeter-scale 3D imaging brain tissues at nanoscale resolution. It is important further improve efficiency acquire larger and more volumes. We report here a three fold increase in speed TEM by using beam deflecting mechanism enable highly efficient acquisition multiple image tiles (nine) each motion mechanical stage. For areas, duty cycle doubles than 30%, yielding net average rate 0.3 gigapixels per second. If fully utilized, an array deflection TEMs should capable dataset cubic millimeter scale several weeks.
Язык: Английский
Процитировано
10
bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown
Опубликована: Окт. 20, 2023
Summary The neocortex is one of the most critical structures that makes us human, and it involved in a variety cognitive functions from perception to sensory integration motor control. Composed repeated modules, or microcircuits, relies on distinct cell types as its fundamental building blocks. Despite significant progress characterizing these 1–5 , an understanding complete synaptic partners associated with individual excitatory remain elusive. Here, we investigate connectivity arguably well recognized studied neuron neocortex: thick tufted layer 5 pyramidal 6–10 also known extra telencephalic (ET) 11 neurons. Although interactions ET neurons have been extensively explored, comprehensive characterization their local remains lacking. To address this knowledge gap, leveraged 1 mm 3 electron microscopic (EM) dataset. We found primarily establish connections inhibitory cells immediate vicinity. However, when they extend axons other cortical regions, tend connect more cells. find targeted by are specific group types, preferentially inhibit Finally, observed common targets IT 6 cells, whereas synapses not common. These findings challenge current views suggest circuit design involves competition among collaboration Our results highlight pattern where subclass forms network offering framework for exploring
Язык: Английский
Процитировано
4
Frontiers in Neuroscience, Год журнала: 2023, Номер 17
Опубликована: Дек. 12, 2023
Serial section multibeam scanning electron microscopy (ssmSEM) is currently among the fastest technologies available for acquiring 3D anatomical data spanning relatively large neural tissue volumes, on order of 1 mm
Язык: Английский
Процитировано
4
bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown
Опубликована: Апрель 5, 2024
Viral projection tracing strategies help establish regional connectomes of mammalian brains. Monosynaptic connectivity with G-deleted rabies virus (RV) establishes retrograde synaptic connectivity, but cannot distinguish networks at cell resolution. We developed ROInet-seq, an accessible spatial method using barcoded ΔG virus, amenable to readout by RNA-sequencing. optimized library complexity and uniformity, such that detection specific barcodes can reliably multiple infected neurons, their individual monosynaptic input networks. In cortical brain areas we found preserved network motives, including co-inputs single neurons from distant local sites. Towards improved resolution simultaneous the cells’ transcriptomes networks, finally sampled hippocampus on a commercial transcriptomics assay reveal details regions’ architecture.
Язык: Английский
Процитировано
1
bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown
Опубликована: Июнь 11, 2024
Abstract Progress in histological methods and microscope technology has enabled dense staining imaging of axons over large brain volumes, but tracing such volumes requires new computational tools for 3D reconstruction data acquired from serial sections. We have developed a pipeline automated volume assembly densely stained imaged sections, which leverages machine learning-based segmentation to enable stitching alignment with the axon traces themselves. validated this segmentation-driven approach individual centimeter-scale sections show application output analysis local orientation proofreading aligned volumes. The is scalable, combined recent advances experimental approaches, should studies mesoscale connectivity function whole human brain.
Язык: Английский
Процитировано
1
Methods in microscopy, Год журнала: 2024, Номер 1(1), С. 49 - 64
Опубликована: Апрель 1, 2024
Abstract Elucidating the 3D nanoscale structure of tissues and cells is essential for understanding complexity biological processes. Electron microscopy (EM) offers resolution needed reliable interpretation, but limited throughput electron microscopes has hindered its ability to effectively image large volumes. We report a workflow volume EM with FAST-EM, novel multibeam scanning transmission microscope that speeds up acquisition by sample in parallel 64 beams. FAST-EM makes use optical detection separate signals individual The reconstruction ultrastructural data from multiple samples demonstrated. results show capable producing reconstructed volumes high contrast address research questions within feasible time frames.
Язык: Английский
Процитировано
1
Опубликована: Июль 1, 2024
Язык: Английский
Процитировано
1
Sensors, Год журнала: 2024, Номер 24(17), С. 5759 - 5759
Опубликована: Сен. 4, 2024
Image stitching aims to construct a wide field of view with high spatial resolution, which cannot be achieved in single exposure. Typically, conventional image techniques, other than deep learning, require complex computation and are thus computationally expensive, especially for large raw images. In this study, inspired by the multiscale feature fluid turbulence, we developed fast point detection algorithm named local-peak scale-invariant transform (LP-SIFT), based on local peaks method. By combining LP-SIFT RANSAC stitching, speed can improved orders compared original SIFT Benefiting from adjustable size interrogation window, demonstrates comparable or even less time commonly used algorithms, while achieving better results. Nine images (over 2600 × 1600 pixels), arranged randomly without prior knowledge, stitched within 158.94 s. The is highly practical applications requiring diverse application scenes, e.g., terrain mapping, biological analysis, criminal investigation.
Язык: Английский
Процитировано
1
BMC Biology, Год журнала: 2023, Номер 21(1)
Опубликована: Сен. 25, 2023
As an extension of electron tomography (ET), serial section (serial ET) aims to align the tomographic images multiple thick tissue sections together, break through volume limitation single and preserve sub-nanoscale voxel size. It could be applied reconstruct intact synapse, which expands about one micrometer contains nanoscale vesicles. However, there are several drawbacks existing ET methods. First, locating imaging regions interest (ROIs) in during shooting process is time-consuming. Second, alignment volumes difficult due missing information caused by cutting imaging. Here we report a workflow simplify acquisition ROIs sections, automatically ET, semi-automatically target synaptic structure.We propose intelligent synapse with sub-nanometer Our includes rapid localization automatic alignment, restoration, assembly volumes, semi-automatic structure segmentation. For use affine transformations calculate their approximate position based on relative location orderly placed sections. consecutive significantly distinct appearances, multi-scale image feature matching elastic belief propagation (BP-Elastic) algorithm them from coarse fine. restoration first estimate number lost pixel changes adjacent after alignment. Then, present generation network that appropriate for small-sample using pre-training interpolation distillation learning. And it generate achieve whole reconstruction. reconstruction ultrastructures, 3D neural obtain quickly. In summary, our can quickly locate acquire align, restore, assemble complete segmentation result minimal manual manipulation. Multiple synapses wild-type rat were reconstructed at size 0.664 nm/voxel demonstrate effectiveness workflow.Our contributes obtaining structures scale ROI locating, reconstruction, We have open-sourced relevant code workflow, so easy apply other labs ultrastructures similar
Язык: Английский
Процитировано
2
bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown
Опубликована: Авг. 13, 2023
ABSTRACT Complex cognitive functions in a mammalian brain are distributed across many anatomically and functionally distinct areas rely on highly dynamic routing of neural activity the network. While modern electrophysiology methods enable recording spiking from increasingly large neuronal populations at cellular level, development probabilistic to extract these inter-area interactions is lagging. Here, we introduce an unsupervised machine learning model that infers connectivity recorded population synchrony their activity. As opposed traditional decoding models reveal dynamics whole population, produces cellular-level cell-type specific functional otherwise omitted analysis. The evaluated ground truth synthetic data compared alternative ensure quality quantification predictions. Our strategy incorporates two sequential stages – extraction static structure network followed by inference temporal changes connection strength. This two-stage architecture enables detailed statistical criteria be developed evaluate confidence predictions comparison with descriptive methods. We applied analyze large-scale in-vivo recordings visual cortices. discovery patterns local long-range circuits cortex temporally varying strength feedforward feedback drives during sensory stimulation. approach provides conceptual link between slow brain-wide studied neuroimaging fast enabled may help uncover often overlooked dimensions code.
Язык: Английский
Процитировано
1