MTNeuro: A Benchmark for Evaluating Representations of Brain Structure Across Multiple Levels of Abstraction DOI Creative Commons
Jorge R. Quesada, Lakshmi Sathidevi, Ran Liu

и другие.

arXiv (Cornell University), Год журнала: 2023, Номер unknown

Опубликована: Янв. 1, 2023

There are multiple scales of abstraction from which we can describe the same image, depending on whether focusing fine-grained details or a more global attribute image. In brain mapping, learning to automatically parse images build representations both small-scale features (e.g., presence cells blood vessels) and properties an image region comes from) is crucial open challenge. However, most existing datasets benchmarks for neuroanatomy consider only single downstream task at time. To bridge this gap, introduce new dataset, annotations, tasks that provide diverse ways readout information about structure architecture Our multi-task neuroimaging benchmark (MTNeuro) built volumetric, micrometer-resolution X-ray microtomography spanning large thalamocortical section mouse brain, encompassing cortical subcortical regions. We generated number different prediction challenges evaluated several supervised self-supervised models brain-region pixel-level semantic segmentation microstructures. experiments not highlight rich heterogeneity but also insights into how approaches be used learn capture attributes perform well variety tasks. Datasets, code, pre-trained baseline provided at: https://mtneuro.github.io/ .

Язык: Английский

Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish DOI Creative Commons
Lena Smirnova, Brian Caffo, David H. Gracias

и другие.

Frontiers in Science, Год журнала: 2023, Номер 1

Опубликована: Фев. 28, 2023

Recent advances in human stem cell-derived brain organoids promise to replicate critical molecular and cellular aspects of learning memory possibly cognition vitro . Coining the term “organoid intelligence” (OI) encompass these developments, we present a collaborative program implement vision multidisciplinary field OI. This aims establish OI as form genuine biological computing that harnesses using scientific bioengineering an ethically responsible manner. Standardized, 3D, myelinated can now be produced with high cell density enriched levels glial cells gene expression for learning. Integrated microfluidic perfusion systems support scalable durable culturing, spatiotemporal chemical signaling. Novel 3D microelectrode arrays permit high-resolution electrophysiological signaling recording explore capacity recapitulate mechanisms formation and, ultimately, their computational potential. Technologies could enable novel biocomputing models via stimulus-response training organoid-computer interfaces are development. We envisage complex, networked whereby connected real-world sensors output devices, ultimately each other sensory organ (e.g. retinal organoids), trained biofeedback, big-data warehousing, machine methods. In parallel, emphasize embedded ethics approach analyze ethical raised by research iterative, manner involving all relevant stakeholders. The many possible applications this urge strategic development discipline. anticipate OI-based allow faster decision-making, continuous during tasks, greater energy data efficiency. Furthermore, “intelligence-in-a-dish” help elucidate pathophysiology devastating developmental degenerative diseases (such dementia), potentially aiding identification therapeutic approaches address major global unmet needs.

Язык: Английский

Процитировано

156
brainlife.io: a decentralized and open-source cloud platform to support neuroscience research DOI Creative Commons
Soichi Hayashi, Bradley Caron, Anibal Sólon Heinsfeld

и другие.

Nature Methods, Год журнала: 2024, Номер 21(5), С. 809 - 813

Опубликована: Апрель 11, 2024

Neuroscience is advancing standardization and tool development to support rigor transparency. Consequently, data pipeline complexity has increased, hindering FAIR (findable, accessible, interoperable reusable) access. brainlife.io was developed democratize neuroimaging research. The platform provides standardization, management, visualization processing automatically tracks the provenance history of thousands objects. Here, described evaluated for validity, reliability, reproducibility, replicability scientific utility using four modalities 3,200 participants.

Язык: Английский

Процитировано

33
Petascale pipeline for precise alignment of images from serial section electron microscopy DOI Creative Commons
Sergiy Popovych, Thomas Macrina, Nico Kemnitz

и другие.

Nature Communications, Год журнала: 2024, Номер 15(1)

Опубликована: Янв. 4, 2024

Abstract The reconstruction of neural circuits from serial section electron microscopy (ssEM) images is being accelerated by automatic image segmentation methods. Segmentation accuracy often limited the preceding step aligning 2D to create a 3D stack. Precise and robust alignment in presence artifacts challenging, especially as datasets are attaining petascale. We present computational pipeline for ssEM with several key elements. Self-supervised convolutional nets trained via metric learning encode align pairs, they used initialize iterative fine-tuning alignment. A procedure called vector voting increases robustness or missing data. For speedup series divided into blocks that distributed workers aligned each other composing transformations decay, which achieves global without resorting time-consuming optimization. apply our whole fly brain dataset, show improved relative prior state art. also demonstrate scales cubic millimeter mouse visual cortex. Our publicly available through two open source Python packages.

Язык: Английский

Процитировано

16
EM-Compressor: Electron Microscopy Image Compression in Connectomics with Variational Autoencoders DOI
Yicong Li, Core Francisco Park, Daniel Xenes

и другие.

Lecture notes in computer science, Год журнала: 2025, Номер unknown, С. 160 - 169

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

1
How innovations in methodology offer new prospects for volume electron microscopy DOI Creative Commons
Arent J. Kievits, Ryan Lane, Elizabeth C. Carroll

и другие.

Journal of Microscopy, Год журнала: 2022, Номер 287(3), С. 114 - 137

Опубликована: Июль 10, 2022

Detailed knowledge of biological structure has been key in understanding biology at several levels organisation, from organs to cells and proteins. Volume electron microscopy (volume EM) provides high resolution 3D structural information about tissues on the nanometre scale. However, throughput rate conventional microscopes limited volume size number samples that can be imaged. Recent improvements methodology are currently driving a revolution EM, making possible imaging whole small organisms. In turn, these recent developments image acquisition have created or stressed bottlenecks other parts pipeline, like sample preparation, analysis data management. While progress is stunning due advent automatic segmentation server-based annotation tools, challenges remain. Here we discuss trends emerging methods for increasing implications

Язык: Английский

Процитировано

27
DVID: Distributed Versioned Image-Oriented Dataservice DOI Creative Commons
William Katz, Stephen M. Plaza

Frontiers in Neural Circuits, Год журнала: 2019, Номер 13

Опубликована: Фев. 5, 2019

Open-source software development has skyrocketed in part due to community tools like github.com, which allows publication of code as well the ability create branches and push accepted modifications back original repository. As number size EM-based datasets increases, connectomics faces similar issues when we publish snapshot data corresponding a publication. Ideally, there would be mechanism where remote collaborators could modify then flexibly reintegrate results via moderated acceptance changes. The DVID system provides web-based API first steps toward such distributed versioning approach datasets. Through its use central resource for Janelia's FlyEM team, have integrated concepts into reconstruction workflows, allowing support proofreader training segmentation experiments through branched, versioned data. also supports persistence variety storage systems from high-speed local SSDs cloud-based object stores, deployment on laptops large servers. tailoring backend each type leads efficient fast queries. is freely available open-source with an increasing supported options.

Язык: Английский

Процитировано

31
A Survey of Visualization and Analysis in High‐Resolution Connectomics DOI Creative Commons
Johanna Beyer,

Jakob Troidl,

Saeed Boorboor

и другие.

Computer Graphics Forum, Год журнала: 2022, Номер 41(3), С. 573 - 607

Опубликована: Июнь 1, 2022

Abstract The field of connectomics aims to reconstruct the wiring diagram Neurons and synapses enable new insights into workings brain. Reconstructing analyzing Neuronal connectivity, however, relies on many individual steps, starting from high‐resolution data acquisition automated segmentation, proofreading, interactive exploration, circuit analysis. All these steps have handle large complex datasets rely or benefit integrated visualization methods. In this state‐of‐the‐art report, we describe methods that can be applied throughout pipeline, We first define different pipeline focus how is currently steps. also survey open science initiatives in connectomics, including usable open‐source tools publicly available datasets. Finally, discuss challenges possible future directions exciting research field.

Язык: Английский

Процитировано

17
FAIR African brain data: challenges and opportunities DOI Creative Commons
Eberechi Wogu, George Ogoh, Patrick Filima

и другие.

Frontiers in Neuroinformatics, Год журнала: 2025, Номер 19

Опубликована: Март 3, 2025

Introduction The effectiveness of research and innovation often relies on the diversity or heterogeneity datasets that are Findable, Accessible, Interoperable Reusable (FAIR). However, global landscape brain data is yet to achieve desired levels can facilitate generalisable outputs. Brain from low-and middle-income countries Africa still missing in open science ecosystem. This mean decades may not be populations Africa. Methods combined experiential learning with a survey questionnaire. involved deriving insights direct, hands-on experiences collecting African view making it FAIR. was critical process action, reflection, doing collection. A questionnaire then used validate findings provide wider contexts for these findings. Results revealed major challenges FAIR categorised as socio-cultural, economic, technical, ethical legal challenges. It also highlighted opportunities growth include capacity development, development technical infrastructure, funding well policy regulatory changes. showed neuroscience community believes ranked order priority follows: Technical, socio-cultural Conclusion We conclude researchers need work together address way maximise efforts build thriving ecosystem socially acceptable, ethically responsible, technically robust legally compliant.

Язык: Английский

Процитировано

0
Automated segmentation of synchrotron-scanned fossils DOI Creative Commons
Melanie A. D. During, Jordan Matelsky, Fredrik K. Gustafsson

и другие.

Fossil record, Год журнала: 2025, Номер 28(1), С. 103 - 114

Опубликована: Март 6, 2025

Computed tomography has revolutionised the study of internal three-dimensional structure fossils. Historically, fossils typically spent years in preparation to be freed from enclosing rock. Now, X-ray and synchrotron reveal structures that are otherwise invisible, data acquisition can fast. However, manual segmentation these 3D volumes still take months years. This is especially challenging for resource-poor teams, as scanning may free, but computing power (AI-assisted) software required handle resulting large sets complex use expensive. Here we present a browser-based tool reduces computational overhead by splitting into small chunks, allowing processing on low-memory, inexpensive hardware. Our also speeds up collaborative ground-truth generation visualisation, all in-browser. We developed evaluated our pipeline various open-data scans differing contrast, resolution, textural complexity, size. successfully isolated Thrinaxodon Broomistega pair an Early Triassic burrow. It cranial bones Cretaceous acipenseriform Parapsephurus willybemisi both 45.53 µm 13.67 resolution (voxel size) data. Middle sauropterygian Nothosaurus scan squamate embryo inside egg dating back Cretaceous. reliably reproduces expert-supervised at fraction time cost, offering greater accessibility than existing tools. Beyond online tool, code open source, enabling contributions palaeontology community further this emerging machine-learning ecosystem.

Язык: Английский

Процитировано

0
Structural Diversity of Mitochondria in the Neuromuscular System across Development Revealed by 3D Electron Microscopy DOI Creative Commons
J. Alexander Bae, Myung-Kyu Choi, Soungyub Ahn

и другие.

Advanced Science, Год журнала: 2025, Номер unknown

Опубликована: Март 6, 2025

As an animal matures, its neural circuit undergoes alterations, yet the developmental changes in intracellular organelles to facilitate these is less understood. Using 3D electron microscopy and deep learning, study develops semi-automated methods for reconstructing mitochondria C. elegans collected reconstructions from normal reproductive stages dauer, enabling comparative on structure within neuromuscular system. It found that various structural properties neurons correlate with synaptic connections are preserved across development different circuits. To test necessity of universal properties, examines behavior drp-1 mutants impaired fission discovers it causes behavioral deficits. Moreover, observed dauer display distinctive mitochondrial features, muscles exhibit unique reticulum-like structure. proposed specialized structures may serve as adaptive mechanism support stage-specific physiological needs.

Язык: Английский

Процитировано

0