Classification of amyotrophic lateral sclerosis by brain volume, connectivity, and network dynamics DOI Creative Commons
Janine Thome, Robert Steinbach, Julian Großkreutz

и другие.

Human Brain Mapping, Год журнала: 2021, Номер 43(2), С. 681 - 699

Опубликована: Окт. 16, 2021

Abstract Emerging studies corroborate the importance of neuroimaging biomarkers and machine learning to improve diagnostic classification amyotrophic lateral sclerosis (ALS). While most focus on structural data, recent assessing functional connectivity between brain regions by linear methods highlight role function. These have yet be combined with structure nonlinear features. We investigate features, benefit combining function for ALS classification. patients ( N = 97) healthy controls 59) underwent resting state magnetic resonance imaging. Based key hubs networks, we defined three feature sets comprising volume, (rsFC), as well (nonlinear) dynamics assessed via recurrent neural networks. Unimodal multimodal random forest classifiers were built classify ALS. Out‐of‐sample prediction errors five‐fold cross‐validation. achieved a accuracy 56.35–61.66%. Multimodal outperformed unimodal achieving accuracies 62.85–66.82%. Evaluating ranking individual features' scores across all revealed that rsFC features dominant in univariate analyses reduced patients, more generally indicated deficits information integration networks The present work undermines provides an additional classification, classifiers, while emphasizing capturing both properties identify discriminative

Язык: Английский

Leaving flatland: Advances in 3D behavioral measurement DOI Creative Commons
Jesse D. Marshall,

Tianqing Li,

Joshua H. Wu

и другие.

Current Opinion in Neurobiology, Год журнала: 2022, Номер 73, С. 102522 - 102522

Опубликована: Апрель 1, 2022

Язык: Английский

Процитировано

29
Mouse visual cortex as a limited resource system that self-learns an ecologically-general representation DOI Creative Commons
Aran Nayebi, Nathan C. L. Kong, Chengxu Zhuang

и другие.

PLoS Computational Biology, Год журнала: 2023, Номер 19(10), С. e1011506 - e1011506

Опубликована: Окт. 2, 2023

Studies of the mouse visual system have revealed a variety brain areas that are thought to support multitude behavioral capacities, ranging from stimulus-reward associations, goal-directed navigation, and object-centric discriminations. However, an overall understanding mouse’s cortex, how it supports range behaviors, remains unknown. Here, we take computational approach help address these questions, providing high-fidelity quantitative model cortex identifying key structural functional principles underlying model’s success. Structurally, find comparatively shallow network structure with low-resolution input is optimal for modeling cortex. Our main finding functional—that models trained task-agnostic, self-supervised objective functions based on concept contrastive embeddings much better matches than supervised objectives or alternative methods. This result very unlike in primates where prior work showed two were roughly equivalent, naturally leading us ask question why ones mouse. To this end, show self-supervised, builds general-purpose representation enables achieve transfer out-of-distribution scene reward-based navigation tasks. results suggest low-resolution, makes best use limited resources create light-weight, system—in contrast deep, high-resolution, more categorization-dominated primates.

Язык: Английский

Процитировано

21
Artificial intelligence for life sciences: A comprehensive guide and future trends DOI

Ming Luo,

Wenyu Yang, Long Bai

и другие.

The Innovation Life, Год журнала: 2024, Номер unknown, С. 100105 - 100105

Опубликована: Янв. 1, 2024

<p>Artificial intelligence has had a profound impact on life sciences. This review discusses the application, challenges, and future development directions of artificial in various branches sciences, including zoology, plant science, microbiology, biochemistry, molecular biology, cell developmental genetics, neuroscience, psychology, pharmacology, clinical medicine, biomaterials, ecology, environmental science. It elaborates important roles aspects such as behavior monitoring, population dynamic prediction, microorganism identification, disease detection. At same time, it points out challenges faced by application data quality, black-box problems, ethical concerns. The are prospected from technological innovation interdisciplinary cooperation. integration Bio-Technologies (BT) Information-Technologies (IT) will transform biomedical research into AI for Science paradigm.</p>

Язык: Английский

Процитировано

9
Emergent neural dynamics and geometry for generalization in a transitive inference task DOI Creative Commons
Kenneth Kay, Natalie Biderman, Ramin Khajeh

и другие.

PLoS Computational Biology, Год журнала: 2024, Номер 20(4), С. e1011954 - e1011954

Опубликована: Апрель 25, 2024

Relational cognition—the ability to infer relationships that generalize novel combinations of objects—is fundamental human and animal intelligence. Despite this importance, it remains unclear how relational cognition is implemented in the brain due part a lack hypotheses predictions at levels collective neural activity behavior. Here we discovered, analyzed, experimentally tested networks (NNs) perform transitive inference (TI), classic task (if A > B C, then C). We found NNs (i) generalized perfectly, despite lacking overt structure prior training, (ii) when required working memory (WM), capacity thought be essential brain, (iii) emergently expressed behaviors long observed living subjects, addition order-dependent behavior, (iv) different solutions yielding alternative behavioral predictions. Further, large-scale experiment, subjects performing WM-based TI showed behavior inconsistent with class characteristically an intuitive solution. These findings provide insights into classical ability, wider implications for realizes cognition.

Язык: Английский

Процитировано

7
De novo motor learning creates structure in neural activity that shapes adaptation DOI Creative Commons
Joanna Chang, Matthew G. Perich, Lee E. Miller

и другие.

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

Опубликована: Май 14, 2024

Abstract Animals can quickly adapt learned movements to external perturbations, and their existing motor repertoire likely influences ease of adaptation. Long-term learning causes lasting changes in neural connectivity, which shapes the activity patterns that be produced during Here, we examined how a population’s patterns, acquired through de novo learning, affect subsequent adaptation by modeling cortical population dynamics with recurrent networks. We trained networks on different repertoires comprising varying numbers movements, they following various experiences. Networks multiple had more constrained robust dynamics, were associated defined ‘structure’—organization available patterns. This structure facilitated adaptation, but only when imposed perturbation congruent organization inputs learning. These results highlight trade-offs skill acquisition demonstrate experiences shape geometrical properties

Язык: Английский

Процитировано

7
Decoding the brain: From neural representations to mechanistic models DOI Creative Commons
Mackenzie Weygandt Mathis, Adriana Perez Rotondo, Edward F. Chang

и другие.

Cell, Год журнала: 2024, Номер 187(21), С. 5814 - 5832

Опубликована: Окт. 1, 2024

Язык: Английский

Процитировано

7
Inferring brain-wide interactions using data-constrained recurrent neural network models DOI Creative Commons
Matthew G. Perich, Charlotte Arlt, Sofia Soares

и другие.

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2020, Номер unknown

Опубликована: Дек. 21, 2020

ABSTRACT Behavior arises from the coordinated activity of numerous anatomically and functionally distinct brain regions. Modern experimental tools allow unprecedented access to large neural populations spanning many interacting regions brain-wide. Yet, understanding such large-scale datasets necessitates both scalable computational models extract meaningful features inter-region communication principled theories interpret those features. Here, we introduce Current-Based Decomposition (CURBD), an approach for inferring brain-wide interactions using data-constrained recurrent network that directly reproduce experimentally-obtained data. CURBD leverages functional inferred by reveal directional currents between multiple We first show accurately isolates in simulated networks with known dynamics. then apply multi-region recordings obtained mice during running, macaques Pavlovian conditioning, humans memory retrieval demonstrate widespread applicability untangle underlying behavior a variety datasets.

Язык: Английский

Процитировано

50
A Network Perspective on Sensorimotor Learning DOI
Hansem Sohn, Nicolas Meirhaeghe,

Rishi Rajalingham

и другие.

Trends in Neurosciences, Год журнала: 2020, Номер 44(3), С. 170 - 181

Опубликована: Дек. 18, 2020

Язык: Английский

Процитировано

40
Compartmentalized dynamics within a common multi-area mesoscale manifold represent a repertoire of human hand movements DOI Creative Commons
Nikhilesh Natraj, Daniel B. Silversmith, Edward F. Chang

и другие.

Neuron, Год журнала: 2021, Номер 110(1), С. 154 - 174.e12

Опубликована: Окт. 22, 2021

Язык: Английский

Процитировано

39
Towards the next generation of recurrent network models for cognitive neuroscience DOI Creative Commons
Guangyu Robert Yang, Manuel Molano‐Mazón

Current Opinion in Neurobiology, Год журнала: 2021, Номер 70, С. 182 - 192

Опубликована: Окт. 1, 2021

Язык: Английский

Процитировано

34