Learning Shared Neural Manifolds from Multi-Subject FMRI Data

Huang, Jessie; Busch, Erica L.; Wallenstein, Tom; Gerasimiuk, Michal; Benz, Andrew; Lajoie, Guillaume; Wolf, Guy; Turk‐Browne, Nicholas B.; Krishnaswamy, Smita

doi:10.1109/mlsp55214.2022.9943383

Cited by 7 publications

(11 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This allows PHATE manifolds to reflect local and global manifold structure and perform effective denoising and makes PHATE well suited for the high dimensionality and intrinsic noise of fMRI activity. Despite the low reliability 36 and aforementioned challenges of developmental task-based neuroimaging data, we suspected that individual differences in brain function would be highlighted in low-dimensional PHATE manifold and would, in turn, better relate to cognition and mental health relative to voxel data 22,24 .…”

Section: Methodsmentioning

confidence: 99%

“…This resulted in a participants-by-voxels matrix for each region and contrast, which was then embedded using PHATE into 20 dimensions to get a participants-by-20 PHATE dimensions matrix. Dimensionality was selected based on prior literature 21,23 and was kept consistent across regions, contrasts, and embedding methods.…”

Section: Methodsmentioning

confidence: 99%

“…PHATE captures local neighborhoods in the data using a diffused Markov transition matrix and also maintains global data context using a potential distance metric. This allows PHATE manifolds to reflect local and global manifold structure and perform effective denoising and makes PHATE well suited for the high dimensionality and intrinsic noise of fMRI activity 21,23 .…”

Section: Methodsmentioning

confidence: 99%

“…Further information about these measures and the E-PHATE algorithm are outlined in Supplemental Methods. For all analyses, PHATE and E-PHATE embeddings were 20 dimensional, which was selected based on prior literature 22,24 . Embeddings were performed separately for each region/network and task contrast.…”

Section: Manifold Embeddingsmentioning

confidence: 99%

“…Manifold learning, a nonlinear approach to dimensionality reduction, is increasingly popular for highlighting complex latent structure in high-dimensional biological data 19 . In particular, PHATE, an unsupervised manifold learning algorithm designed for high-dimensional, noisy biomedical data, has been used to discover local and global latent structure in functional magnetic resonance imaging (fMRI) data [20][21][22] . Prior research showed that combining PHATE with additional data (e.g., stimulus structure or temporal dynamics of brain responses) can enhance the relevance of the latent structure of brain activation for understanding complex cognitive processing, for example during movie-viewing 23 .…”

mentioning

confidence: 99%

See 4 more Smart Citations

Manifold learning uncovers nonlinear interactions between the adolescent brain and environment that predict emotional and behavioral problems

Busch,

Conley,

Baskin-Sommers

2024

Preprint

View full text Add to dashboard Cite

Advanced statistical methods that capture the complex interplay between adolescents and their social environments are essential for improving our understanding of how differences in brain function contribute to mental health problems. To move the study of adolescent mental health beyond what we have achieved so far- a complex account of brain and environmental risk factors without understanding the neurobiological embedding of the social environment- we need to find ways of studying the complex, nonlinear relationships between brain function and adolescents' experiences in the real-world. Manifold learning techniques can discover and highlight latent structure from high-dimensional, complex biomedical data, such as fMRI. Here, we develop a novel manifold learning technique, exogenous PHATE (EPHATE), to capture the interplay between brain function and adolescents' social environments. By applying EPHATE, we demonstrate that harmonizing cutting-edge computational methods with longstanding developmental theory can advance efforts to detect and predict mental health problems during the transition to adolescence.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Manifold Embeddingsmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Manifold learning uncovers nonlinear interactions between the adolescent brain and environment that predict emotional and behavioral problems

Busch,

Conley,

Baskin-Sommers

2024

Preprint

View full text Add to dashboard Cite

show abstract

Manifold Learning Uncovers Nonlinear Interactions Between the Adolescent Brain and Environment That Predict Emotional and Behavioral Problems

Busch,

Conley,

Baskin-Sommers

2024

Biological Psychiatry: Cognitive Neuroscience and Neuroimaging

View full text Add to dashboard Cite

A novel subject-wise dictionary learning approach using multi-subject fMRI spatial and temporal components

Khalid,

Nauman

2023

Sci Rep

View full text Add to dashboard Cite

The conventional dictionary learning (DL) algorithms aim to adapt the dictionary/sparse code to individual functional magnetic resonance imaging (fMRI) data. Thus, lacking the capability to consolidate the spatiotemporal diversities offered by other subjects. Considering that subject-wise (sw) data matrix can be decomposed into the sparse linear combination of multi-subject (MS) time courses and MS spatial maps, two new algorithms, sw sequential DL (swsDL) and sw block DL (swbDL), have been proposed. They are based on the novel framework, defined by the mixing model, where base matrices prepared by operating a computationally fast sparse spatiotemporal blind source separation method over multiple subjects are employed to adapt the mixing matrices to sw training data. They solve the optimization models formulated using $$l_0$$ l 0 /$$l_1$$ l 1 -norm penalization/constraints through dictionary/sparse code pair update and alternating minimization approach. They are unique because no existing sparse DL method can incorporate MS spatiotemporal components while updating sw atoms/sparse codes, which can eventually be assembled using neuroscience knowledge to extract group-level dynamics. Various fMRI datasets are used to evaluate and compare the performance of the proposed algorithms with existing state-of-the-art algorithms. Specifically, overall, a $$14\%$$ 14 % increase in the mean correlation value and $$39\%$$ 39 % reduction in the mean computation time exhibited by swsDL and swbDL, respectively, over the adaptive consistent sequential dictionary algorithm.

show abstract

Learning Shared Neural Manifolds from Multi-Subject FMRI Data

Cited by 7 publications

References 22 publications

Manifold learning uncovers nonlinear interactions between the adolescent brain and environment that predict emotional and behavioral problems

Manifold learning uncovers nonlinear interactions between the adolescent brain and environment that predict emotional and behavioral problems

Manifold Learning Uncovers Nonlinear Interactions Between the Adolescent Brain and Environment That Predict Emotional and Behavioral Problems

A novel subject-wise dictionary learning approach using multi-subject fMRI spatial and temporal components

Contact Info

Product

Resources

About