Global Neuron Shape Reasoning with Point Affinity Transformers DOI Creative Commons

Jakob Troidl,

Johannes Knittel, Wanhua Li

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 24, 2024

Connectomics is a subfield of neuroscience that aims to map the brain's intricate wiring diagram. Accurate neuron segmentation from microscopy volumes essential for automating connectome reconstruction. However, current state-of-the-art algorithms use image-based convolutional neural networks are limited local shape context. Thus, we introduce new framework reasons over global with novel point affinity transformer. Our embeds (multi-)neuron cloud into fixed-length feature set which can decode any pair affinities, enabling clustering clouds automatic proofreading. We also show learned easily be mapped contrastive embedding space enables type classification using simple KNN classifier. approach excels in two demanding connectomics tasks: proofreading errors and classifying types. Evaluated on three benchmark datasets derived connectomes, our method outperforms transformers, graph networks, unsupervised baselines.

Language: Английский

Non-destructive X-ray tomography of brain tissue ultrastructure DOI Creative Commons
Carles Bosch, Tomas Aidukas, Mirko Holler

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 17, 2023

Abstract Maps of dense subcellular features in biological tissue are the key to understanding structural basis organ function. Electron microscopy provides necessary resolution, yet - as electrons penetrate samples for only a few 100s nm requires physical sectioning or ablation, which strongly challenges anatomical investigations entire organs such mammalian brains. As demonstrated engineering and sciences, X-ray nanotomography represents promising alternative ultrastructural 3d imaging without 1–15 . Leveraging high brilliance 4th generation synchrotron sources, it has potential non-destructively image mm³-sized at resolution within days 16 A fundamental barrier application life sciences is that, when irradiated with high-intensity X-rays, deform ultimately disintegrate, prohibiting reaching sufficient resolution. Here, we introduce combination solutions defeat this ptychography 17 , coherent diffractive technique. The include cryogenic sample stage stability, high-precision interferometric positioners tailored non-rigid tomographic reconstruction algorithms 18 Furthermore, adapting an epoxy resin developed nuclear aerospace industry, demonstrate radiation resistance doses exceeding 10 Gy. resulting sub-40 isotropic makes possible densely resolve axon bundles, boutons, dendrites reliably identify synapses sectioning. Moreover, validated technique using current gold standard, namely focused ion beam scanning electron (FIB-SEM) 19,20 intact ultrastructure volumes first imaged by X-rays. This unlocks tomography high-resolution imaging, coinciding transformative advancements next-generation synchrotrons worldwide 21

Language: Английский

Citations

5
A Novel Semi-automated Proofreading and Mesh Error Detection Pipeline for Neuron Extension DOI Creative Commons
Justin Joyce,

Rupasri Chalavadi,

Joey Chan

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Oct. 23, 2023

The immense scale and complexity of neuronal electron microscopy (EM) datasets pose significant challenges in data processing, validation, interpretation, necessitating the development efficient, automated, scalable error-detection methodologies. This paper proposes a novel approach that employs mesh processing techniques to identify potential error locations near tips. Error detection at tips is particularly important challenge since these errors usually indicate many synapses are falsely split from their parent neuron, injuring integrity connectomic reconstruction. Additionally, we draw implications results an implementation this semi-automated proofreading pipeline. Manual laborious, costly, currently necessary method for identifying machine learning based segmentation neural tissue. streamlines process by systematically highlighting areas likely contain inaccuracies guiding proofreaders towards continuations, accelerating rate which corrected.

Language: Английский

Citations

3
Selective Labeling Meets Semi-Supervised Neuron Segmentation DOI Creative Commons
Yanchao Zhang, Hao Zhai, Jinyue Guo

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: May 31, 2024

ABSTRACT Semi-supervised learning offers a cost-effective approach for neuron segmentation in electron microscopy (EM) volumes. This technique leverages extensive unlabeled data to regularize supervised training more robust predictions of affinities. However, the distribution mismatch between labeled and datasets, arising from limited annotations diversity neuronal patterns, impedes generalization semi-supervised models. In this paper, we develop dual-level pipeline address inherent issue enhance segmentation. At level, propose an unsupervised heuristic select valuable sub-volumes as based on similarity pretrained feature space, ensuring representative coverage structures. model introduce axial-through mixing strategy into anisotropic integrate it framework. Building this, establish cross-view consistency constraints through intra- inter-mixing which facilitates shared semantics across distributions while avoiding ambiguity Extensive comparative experiments ablation studies publicly available datasets demonstrate effectiveness proposed method different EM modalities spatial resolutions.

Language: Английский

Citations

0
NeuroVerse: Immersive exploration of 3D ultrastructural brain reconstructions for education and collaborative analysis DOI
Corrado Calì, Marco Agus

Published: Aug. 28, 2024

Language: Английский

Citations

0
Beyond neurons: computer vision methods for analysis of morphologically complex astrocytes DOI Creative Commons
Tabish A. Syed, Mohammed Youssef, Alexandra L. Schober

et al.

Frontiers in Computer Science, Journal Year: 2024, Volume and Issue: 6

Published: Sept. 25, 2024

The study of the geometric organization biological tissues has a rich history in literature. However, geometry and architecture individual cells within traditionally relied upon manual or indirect measures shape. Such rudimentary are largely result challenges associated with acquiring high resolution images cellular components, as well lack computational approaches to analyze large volumes high-resolution data. This is especially true brain tissue, which composed complex array cells. Here we review tools that have been applied unravel nanoarchitecture astrocytes, type cell increasingly being shown be essential for function. Astrocytes among most structurally functionally diverse mammalian body partner neurons. Light microscopy does not allow adequate astrocyte morphology, however, large-scale serial electron data, provides nanometer 3D models, enabling visualization fine, convoluted structure astrocytes. Application computer vision methods resulting nanoscale models helping reveal organizing principles but complete understanding its functional implications will require further adaptation existing tools, development new approaches.

Language: Английский

Citations

0
SegNeuron: 3D Neuron Instance Segmentation in Any EM Volume with a Generalist Model DOI
Yanchao Zhang, Jinyue Guo, Hao Zhai

et al.

Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 589 - 600

Published: Jan. 1, 2024

Language: Английский

Citations

0
Cell nuclei segmentation in mm-scale x-ray holographic nanotomography images of mouse brain tissue DOI Creative Commons
Andrea Nathansen,

M. Clausen,

Manuel Berning

et al.

Published: Oct. 23, 2024

Biological soft tissues are functional agglomerates of cells. They constitute the microenvironment where intercellular communication occurs. In turn, their woven structure underlies mechanical properties that contribute to roles in context organs and organisms contain them. Therefore, determining density spatial distribution cells within tissue offers key information for understanding its physiological state. X-ray holographic nanotomography is a non-destructive imaging technique capable resolving subcellular details biological has shown promising advantages study neuronal circuits. However, dimensions datasets required – covering volume landscapes ~mm3 make manual annotation individual nuclei an unrealistic task. We developed trained automated image segmentation classifier accurately detects segments cell mouse brain imaged with x-ray nanotomography, generalises similar obtained from replicates minimal additional ground truth. It provides locations morphologies ~80k per dataset high recall. harnesses strengths high-performance computing cluster embeds curated results two main simplified outcomes: data table explorable segmentations meshes associated original dataset, browser-compatible format simplifies proofreading by multiple users. The we present here can be readily integrated into analytical pipeline histological synchrotron systems neuroscience as well broader life science studies.

Language: Английский

Citations

0
Global Neuron Shape Reasoning with Point Affinity Transformers DOI Creative Commons

Jakob Troidl,

Johannes Knittel, Wanhua Li

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 24, 2024

Connectomics is a subfield of neuroscience that aims to map the brain's intricate wiring diagram. Accurate neuron segmentation from microscopy volumes essential for automating connectome reconstruction. However, current state-of-the-art algorithms use image-based convolutional neural networks are limited local shape context. Thus, we introduce new framework reasons over global with novel point affinity transformer. Our embeds (multi-)neuron cloud into fixed-length feature set which can decode any pair affinities, enabling clustering clouds automatic proofreading. We also show learned easily be mapped contrastive embedding space enables type classification using simple KNN classifier. approach excels in two demanding connectomics tasks: proofreading errors and classifying types. Evaluated on three benchmark datasets derived connectomes, our method outperforms transformers, graph networks, unsupervised baselines.

Language: Английский

Citations

0