@article{sun_foundation_2024,
	title = {A foundation model for enhancing magnetic resonance images and downstream segmentation, registration and diagnostic tasks},
	copyright = {2024 The Author(s), under exclusive licence to Springer Nature Limited},
	issn = {2157-846X},
	url = {https://www.nature.com/articles/s41551-024-01283-7},
	doi = {10.1038/s41551-024-01283-7},
	abstract = {In structural magnetic resonance (MR) imaging, motion artefacts, low resolution, imaging noise and variability in acquisition protocols frequently degrade image quality and confound downstream analyses. Here we report a foundation model for the motion correction, resolution enhancement, denoising and harmonization of MR images. Specifically, we trained a tissue-classification neural network to predict tissue labels, which are then leveraged by a ‘tissue-aware’ enhancement network to generate high-quality MR images. We validated the model’s effectiveness on a large and diverse dataset comprising 2,448 deliberately corrupted images and 10,963 images spanning a wide age range (from foetuses to elderly individuals) acquired using a variety of clinical scanners across 19 public datasets. The model consistently outperformed state-of-the-art algorithms in improving the quality of MR images, handling pathological brains with multiple sclerosis or gliomas, generating 7-T-like images from 3 T scans and harmonizing images acquired from different scanners. The high-quality, high-resolution and harmonized images generated by the model can be used to enhance the performance of models for tissue segmentation, registration, diagnosis and other downstream tasks.},
	language = {en},
	urldate = {2025-01-06},
	journal = {Nature Biomedical Engineering},
	author = {Sun, Yue and Wang, Limei and Li, Gang and Lin, Weili and Wang, Li},
	month = dec,
	year = {2024},
	note = {Publisher: Nature Publishing Group},
	keywords = {Computational models, Image processing, Machine learning},
	pages = {1--18},
}

In structural magnetic resonance (MR) imaging, motion artefacts, low resolution, imaging noise and variability in acquisition protocols frequently degrade image quality and confound downstream analyses. Here we report a foundation model for the motion correction, resolution enhancement, denoising and harmonization of MR images. Specifically, we trained a tissue-classification neural network to predict tissue labels, which are then leveraged by a ‘tissue-aware’ enhancement network to generate high-quality MR images. We validated the model’s effectiveness on a large and diverse dataset comprising 2,448 deliberately corrupted images and 10,963 images spanning a wide age range (from foetuses to elderly individuals) acquired using a variety of clinical scanners across 19 public datasets. The model consistently outperformed state-of-the-art algorithms in improving the quality of MR images, handling pathological brains with multiple sclerosis or gliomas, generating 7-T-like images from 3 T scans and harmonizing images acquired from different scanners. The high-quality, high-resolution and harmonized images generated by the model can be used to enhance the performance of models for tissue segmentation, registration, diagnosis and other downstream tasks.

@article{wang_large-scale_2024,
	title = {Large-scale foundation models and generative {AI} for {BigData} neuroscience},
	issn = {0168-0102},
	url = {https://www.sciencedirect.com/science/article/pii/S0168010224000750},
	doi = {10.1016/j.neures.2024.06.003},
	abstract = {Recent advances in machine learning have led to revolutionary breakthroughs in computer games, image and natural language understanding, and scientific discovery. Foundation models and large-scale language models (LLMs) have recently achieved human-like intelligence thanks to BigData. With the help of self-supervised learning (SSL) and transfer learning, these models may potentially reshape the landscapes of neuroscience research and make a significant impact on the future. Here we present a mini-review on recent advances in foundation models and generative AI models as well as their applications in neuroscience, including natural language and speech, semantic memory, brain-machine interfaces (BMIs), and data augmentation. We argue that this paradigm-shift framework will open new avenues for many neuroscience research directions and discuss the accompanying challenges and opportunities.},
	urldate = {2025-01-06},
	journal = {Neuroscience Research},
	author = {Wang, Ran and Chen, Zhe Sage},
	month = jun,
	year = {2024},
	keywords = {BigData, Brain-machine interface, Embedding, Foundation model, Generative AI, Representation learning, Self-supervised learning, Transfer learning, Transformer},
}

Recent advances in machine learning have led to revolutionary breakthroughs in computer games, image and natural language understanding, and scientific discovery. Foundation models and large-scale language models (LLMs) have recently achieved human-like intelligence thanks to BigData. With the help of self-supervised learning (SSL) and transfer learning, these models may potentially reshape the landscapes of neuroscience research and make a significant impact on the future. Here we present a mini-review on recent advances in foundation models and generative AI models as well as their applications in neuroscience, including natural language and speech, semantic memory, brain-machine interfaces (BMIs), and data augmentation. We argue that this paradigm-shift framework will open new avenues for many neuroscience research directions and discuss the accompanying challenges and opportunities.

@article{helmer_stability_2024,
	title = {On the stability of canonical correlation analysis and partial least squares with application to brain-behavior associations},
	volume = {7},
	copyright = {2024 The Author(s)},
	issn = {2399-3642},
	url = {https://www.nature.com/articles/s42003-024-05869-4},
	doi = {10.1038/s42003-024-05869-4},
	abstract = {Associations between datasets can be discovered through multivariate methods like Canonical Correlation Analysis (CCA) or Partial Least Squares (PLS). A requisite property for interpretability and generalizability of CCA/PLS associations is stability of their feature patterns. However, stability of CCA/PLS in high-dimensional datasets is questionable, as found in empirical characterizations. To study these issues systematically, we developed a generative modeling framework to simulate synthetic datasets. We found that when sample size is relatively small, but comparable to typical studies, CCA/PLS associations are highly unstable and inaccurate; both in their magnitude and importantly in the feature pattern underlying the association. We confirmed these trends across two neuroimaging modalities and in independent datasets with n ≈ 1000 and n = 20,000, and found that only the latter comprised sufficient observations for stable mappings between imaging-derived and behavioral features. We further developed a power calculator to provide sample sizes required for stability and reliability of multivariate analyses. Collectively, we characterize how to limit detrimental effects of overfitting on CCA/PLS stability, and provide recommendations for future studies.},
	language = {en},
	number = {1},
	urldate = {2025-01-06},
	journal = {Communications Biology},
	author = {Helmer, Markus and Warrington, Shaun and Mohammadi-Nejad, Ali-Reza and Ji, Jie Lisa and Howell, Amber and Rosand, Benjamin and Anticevic, Alan and Sotiropoulos, Stamatios N. and Murray, John D.},
	month = feb,
	year = {2024},
	note = {Publisher: Nature Publishing Group},
	keywords = {Cognitive neuroscience, Computational neuroscience, Statistical methods},
	pages = {1--15},
}

Associations between datasets can be discovered through multivariate methods like Canonical Correlation Analysis (CCA) or Partial Least Squares (PLS). A requisite property for interpretability and generalizability of CCA/PLS associations is stability of their feature patterns. However, stability of CCA/PLS in high-dimensional datasets is questionable, as found in empirical characterizations. To study these issues systematically, we developed a generative modeling framework to simulate synthetic datasets. We found that when sample size is relatively small, but comparable to typical studies, CCA/PLS associations are highly unstable and inaccurate; both in their magnitude and importantly in the feature pattern underlying the association. We confirmed these trends across two neuroimaging modalities and in independent datasets with n ≈ 1000 and n = 20,000, and found that only the latter comprised sufficient observations for stable mappings between imaging-derived and behavioral features. We further developed a power calculator to provide sample sizes required for stability and reliability of multivariate analyses. Collectively, we characterize how to limit detrimental effects of overfitting on CCA/PLS stability, and provide recommendations for future studies.

@article{niso_open_2022,
	title = {Open and reproducible neuroimaging: {From} study inception to publication},
	volume = {263},
	issn = {1053-8119},
	shorttitle = {Open and reproducible neuroimaging},
	url = {https://www.sciencedirect.com/science/article/pii/S1053811922007388},
	doi = {10.1016/j.neuroimage.2022.119623},
	abstract = {Empirical observations of how labs conduct research indicate that the adoption rate of open practices for transparent, reproducible, and collaborative science remains in its infancy. This is at odds with the overwhelming evidence for the necessity of these practices and their benefits for individual researchers, scientific progress, and society in general. To date, information required for implementing open science practices throughout the different steps of a research project is scattered among many different sources. Even experienced researchers in the topic find it hard to navigate the ecosystem of tools and to make sustainable choices. Here, we provide an integrated overview of community-developed resources that can support collaborative, open, reproducible, replicable, robust and generalizable neuroimaging throughout the entire research cycle from inception to publication and across different neuroimaging modalities. We review tools and practices supporting study inception and planning, data acquisition, research data management, data processing and analysis, and research dissemination. An online version of this resource can be found at https://oreoni.github.io. We believe it will prove helpful for researchers and institutions to make a successful and sustainable move towards open and reproducible science and to eventually take an active role in its future development.},
	urldate = {2024-10-09},
	journal = {NeuroImage},
	author = {Niso, Guiomar and Botvinik-Nezer, Rotem and Appelhoff, Stefan and De La Vega, Alejandro and Esteban, Oscar and Etzel, Joset A. and Finc, Karolina and Ganz, Melanie and Gau, Rémi and Halchenko, Yaroslav O. and Herholz, Peer and Karakuzu, Agah and Keator, David B. and Markiewicz, Christopher J. and Maumet, Camille and Pernet, Cyril R. and Pestilli, Franco and Queder, Nazek and Schmitt, Tina and Sójka, Weronika and Wagner, Adina S. and Whitaker, Kirstie J. and Rieger, Jochem W.},
	month = nov,
	year = {2022},
	keywords = {EEG, MEG, MRI, Open science, PET, Reproducibility},
	pages = {119623},
}

Empirical observations of how labs conduct research indicate that the adoption rate of open practices for transparent, reproducible, and collaborative science remains in its infancy. This is at odds with the overwhelming evidence for the necessity of these practices and their benefits for individual researchers, scientific progress, and society in general. To date, information required for implementing open science practices throughout the different steps of a research project is scattered among many different sources. Even experienced researchers in the topic find it hard to navigate the ecosystem of tools and to make sustainable choices. Here, we provide an integrated overview of community-developed resources that can support collaborative, open, reproducible, replicable, robust and generalizable neuroimaging throughout the entire research cycle from inception to publication and across different neuroimaging modalities. We review tools and practices supporting study inception and planning, data acquisition, research data management, data processing and analysis, and research dissemination. An online version of this resource can be found at https://oreoni.github.io. We believe it will prove helpful for researchers and institutions to make a successful and sustainable move towards open and reproducible science and to eventually take an active role in its future development.

@article{tedersoo_data_2021,
	title = {Data sharing practices and data availability upon request differ across scientific disciplines},
	volume = {8},
	copyright = {2021 The Author(s)},
	issn = {2052-4463},
	url = {https://www.nature.com/articles/s41597-021-00981-0},
	doi = {10.1038/s41597-021-00981-0},
	abstract = {Data sharing is one of the cornerstones of modern science that enables large-scale analyses and reproducibility. We evaluated data availability in research articles across nine disciplines in Nature and Science magazines and recorded corresponding authors’ concerns, requests and reasons for declining data sharing. Although data sharing has improved in the last decade and particularly in recent years, data availability and willingness to share data still differ greatly among disciplines. We observed that statements of data availability upon (reasonable) request are inefficient and should not be allowed by journals. To improve data sharing at the time of manuscript acceptance, researchers should be better motivated to release their data with real benefits such as recognition, or bonus points in grant and job applications. We recommend that data management costs should be covered by funding agencies; publicly available research data ought to be included in the evaluation of applications; and surveillance of data sharing should be enforced by both academic publishers and funders. These cross-discipline survey data are available from the plutoF repository.},
	language = {en},
	number = {1},
	urldate = {2025-01-06},
	journal = {Scientific Data},
	author = {Tedersoo, Leho and Küngas, Rainer and Oras, Ester and Köster, Kajar and Eenmaa, Helen and Leijen, Äli and Pedaste, Margus and Raju, Marju and Astapova, Anastasiya and Lukner, Heli and Kogermann, Karin and Sepp, Tuul},
	month = jul,
	year = {2021},
	note = {Publisher: Nature Publishing Group},
	keywords = {Genetic databases, Molecular ecology},
	pages = {192},
}

Data sharing is one of the cornerstones of modern science that enables large-scale analyses and reproducibility. We evaluated data availability in research articles across nine disciplines in Nature and Science magazines and recorded corresponding authors’ concerns, requests and reasons for declining data sharing. Although data sharing has improved in the last decade and particularly in recent years, data availability and willingness to share data still differ greatly among disciplines. We observed that statements of data availability upon (reasonable) request are inefficient and should not be allowed by journals. To improve data sharing at the time of manuscript acceptance, researchers should be better motivated to release their data with real benefits such as recognition, or bonus points in grant and job applications. We recommend that data management costs should be covered by funding agencies; publicly available research data ought to be included in the evaluation of applications; and surveillance of data sharing should be enforced by both academic publishers and funders. These cross-discipline survey data are available from the plutoF repository.

@article{botvinik-nezer_variability_2020,
	title = {Variability in the analysis of a single neuroimaging dataset by many teams},
	volume = {582},
	url = {https://www.nature.com/articles/s41586-020-2314-9},
	number = {7810},
	urldate = {2024-10-09},
	journal = {Nature},
	author = {Botvinik-Nezer, Rotem and Holzmeister, Felix and Camerer, Colin F. and Dreber, Anna and Huber, Juergen and Johannesson, Magnus and Kirchler, Michael and Iwanir, Roni and Mumford, Jeanette A. and Adcock, R. Alison},
	year = {2020},
	note = {Publisher: Nature Publishing Group UK London},
	pages = {84--88},
}

@misc{noauthor_large_nodate,
	title = {Large language models surpass human experts in predicting neuroscience results {\textbar} {Nature} {Human} {Behaviour}},
	url = {https://www.nature.com/articles/s41562-024-02046-9},
	urldate = {2025-01-06},
}