Spatiotemporal analysis of event-related fMRI data using partial least squares

doi:10.1016/j.neuroimage.2004.05.018

NeuroImage

Volume 23, Issue 2, October 2004, Pages 764-775

https://doi.org/10.1016/j.neuroimage.2004.05.018 Get rights and content

Abstract

Partial least squares (PLS) has proven to be a important multivariate analytic tool for positron emission tomographic and, more recently, event-related potential (ERP) data. The application to ERP incorporates the ability to analyze space and time together, a feature that has obvious appeal for event-related functional magnetic resonance imaging (fMRI) data. This paper presents the extension of spatiotemporal PLS (ST-PLS) to fMRI, explaining the theoretical foundation and application to an fMRI study of auditory and visual perceptual memory. Analysis of activation effects with ST-PLS was compared with conventional univariate random effects analysis, showing general consensus for both methods, but several unique observations by ST-PLS, including enhanced statistical power. The application of ST-PLS for assessment of task-dependent brain–behavior relationships is also presented. Singular features of ST-PLS include (1) no assumptions about the shape of the hemodynamic response functions (HRFs); (2) robust statistical assessment at the image level through permutation tests; (3) protection against outlier influences at the voxel level through bootstrap resampling; (4) flexible analytic configurations that allow assessment of activation difference, brain–behavior relations, and functional connectivity. These features enable ST-PLS to act as an important complement to other multivariate and univariate approaches used in neuroimaging research.

Introduction

Partial least squares (PLS, McIntosh et al., 1996) has proven sensitive to the detection of task-related activity changes, brain–behavior relations, and large-scale patterns of functional connectivity in applications to PET data (Della-Maggiore et al., 2000). More recently, we introduced the spatiotemporal extension of PLS (ST-PLS) for the analysis of event-related potentials (ERP) (Lobaugh et al., 2001). In the current paper, we demonstrate the application of ST-PLS to event-related functional magnetic resonance imaging (fMRI) data. ST-PLS is able to identify time-varying distributed activity patterns that differentiate experimental conditions. Moreover, as with the application of PLS to PET data, ST-PLS enables the measurement of brain–behavior relations and whether such relationships vary between conditions or groups.

The effectiveness of ST-PLS is demonstrated through application to a visual-auditory perceptual memory experiment. The mechanics of ST-PLS for fMRI data are similar to the version used for ERP, with the obvious difference being the temporal resolution of the data. The primary advantages of ST-PLS over other current methods are (1) no assumptions about the shape of the hemodynamic response functions (HRFs); (2) robust statistical assessment at the image level through permutation tests; (3) protection against outlier influences at the voxel level through bootstrap resampling; (4) flexible configurations that allow assessment of activation difference, brain–behavior relations, and functional connectivity.

Section snippets

Partial least squares

The term “partial least squares” refers to the computation of the optimal least-squares fit to part of a correlation or covariance matrix (Wold, 1982). The part is the “cross-block” correlation between the exogenous and dependent measures. PLS is similar to PCA, but one important feature of PLS is that the solutions are constrained to the part of the covariance structure that is attributable to experimental manipulations or that relate to behavior. Moreover, PLS is ideal for data sets where the

fMRI experiments

Data were obtained from a perceptual memory study comparing auditory and visual processes. The full report of this experiment is in preparation. We report a subset of this study to illustrate the use of ST-PLS for event-related designs.

Task-PLS

Two significant LVs were identified from the task analysis. The first (Fig. 2) was the main effect of task versus baseline (P < 0.001). The spatiotemporal activity pattern for LV1 is shown in Fig. 2A. From this pattern, the peak differentiation appears to occur at 6–8 s after stimulus onset (four to five TRs), which is corroborated by the temporal scores (Fig. 2B). The response for individual voxels, selected from the maxima in the singular image, is shown in Fig. 3 and illustrates that the

Discussion

We presented the application of spatiotemporal PLS (ST-PLS) to the analysis of functional MRI data. ST-PLS is an effective multivariate analytic tool for data with spatial and temporal information and was originally proposed for ERP (Lobaugh et al., 2001). In the present paper, ST-PLS has been shown to be sensitive to distributed brain activity patterns associated with experimental design, including providing some valuable information about the delay and duration of the responses. ST-PLS can

Acknowledgements

This work was supported by CIHR, ORDCF, and JS McDonnell Foundation grants to AR McIntosh.

References (33)

R.W. Cox
AFNI: software for analysis and visualization of functional magnetic resonance neuroimages
Comput. Biomed. Res
(1996)
K.J. Friston et al.
Multisubject fMRI studies and conjunction analyses
NeuroImage
(1999)
K.J. Friston et al.
Dynamic causal modelling
NeuroImage
(2003)
C.L. Grady et al.
An examination of the effects of stimulus type, encoding task, and functional connectivity on the role of right prefrontal cortex in recognition memory
NeuroImage
(2001)
L. Harrison et al.
Multivariate autoregressive modeling of fMRI time series
NeuroImage
(2003)
R.A. Harshman et al.
PARAFAC—Parallel factor analysis
Comput. Stat. Data Anal
(1994)
R.N. Henson et al.
Detecting latency differences in event-related BOLD responses: application to words versus nonwords and initial versus repeated face presentations
NeuroImage
(2002)
A.R. McIntosh et al.
Spatial pattern analysis of functional brain images using Partial Least Squares
NeuroImage
(1996)
J. Raz et al.
Statistical tests for fMRI based on experimental randomization
NeuroImage
(2003)
P.D. Sampson et al.
Neurobehavioral effects of prenatal alcohol: Part II. Partial least squares analysis
Neurotoxicol. Teratol
(1989)

S.C. Strother et al.

The quantitative evaluation of functional neuroimaging experiments: the NPAIRS data analysis framework

NeuroImage

(2002)

K.J. Worsley et al.

Characterizing the response of PET and fMRI data using multivariate linear models

NeuroImage

(1997)

C. Buchel et al.

Modulation of connectivity in visual pathways by attention: cortical interactions evaluated with structural equation modeling and fMRI

Cereb. Cortex

(1997)

R. Cabeza et al.

Age-related differences in neural activity during memory encoding and retrieval: a positron emission tomography study

J. Neurosci

(1997)

V. Della-Maggiore et al.

Corticolimbic interactions associated with performance on a short-term memory task are modified by age

J. Neurosci

(2000)

E.S. Edgington

Randomization Tests

(1980)

Cited by (266)

Variations in Cortical Functional Gradients Relate to Dimensions of Psychopathology in Preschool Children
2024, Journal of the American Academy of Child and Adolescent Psychiatry
It is unclear how the functional brain hierarchy is organized in preschool-aged children, and whether alterations in the brain organization are linked to mental health in this age group. Here, we assessed whether preschool-aged children exhibit a brain organizational structure similar to that of older children, how this structure might change over time, and whether it might reflect mental health.
This study derived functional gradients using diffusion embedding from resting state functional magnetic resonance imaging data of 4.5-year-old children (N = 100, 42 male participants) and 6.0-year-old children (N = 133, 62 male participants) from the longitudinal Growing Up in Singapore Towards healthy Outcomes (GUSTO) cohort. We then conducted partial least-squares correlation analyses to identify the association between the impairment ratings of different mental disorders and network gradient values.
The main organizing axis of functional connectivity (ie, principal gradient) separated the visual and somatomotor regions (ie, unimodal) in preschool-aged children, whereas the second axis delineated the unimodal−transmodal gradient. This pattern of organization was stable from 4.5 to 6 years of age. The second gradient separating the high- and low-order networks exhibited a diverging pattern across mental health severity, differentiating dimensions related to attention-deficit/hyperactivity disorder and phobic disorders.
This study characterized, for the first time, the functional brain hierarchy in preschool-aged children. A divergence in functional gradient pattern across different disease dimensions was found, highlighting how perturbations in functional brain organization can relate to the severity of different mental health disorders.
Functional brain network identification and fMRI augmentation using a VAE-GAN framework
2023, Computers in Biology and Medicine
Recently, deep learning models have achieved superior performance for mapping functional brain networks from functional magnetic resonance imaging (fMRI) data compared with traditional methods. However, due to the lack of sufficient data and the high dimensionality of brain volume, deep learning models of fMRI tend to suffer from overfitting. In addition, existing methods rarely studied fMRI data augmentation and its application. To address these issues, we developed a VAE-GAN framework that combined a VAE (variational auto-encoder) with a GAN (generative adversarial net) for functional brain network identification and fMRI augmentation. As a generative model, the VAE-GAN models the distribution of fMRI so that it enables the extraction of more generalized features, and thus relieve the overfitting issue. The VAE-GAN is easier to train on fMRI than a standard GAN since it uses latent variables from VAE to generate fake data rather than relying on random noise that is used in a GAN, and it can generate higher quality of fake data than VAE since the discriminator can promote the training of the generator. In other words, the VAE-GAN inherits the advantages of VAE and GAN and avoids their limitations in modeling of fMRI data. Extensive experiments on task fMRI datasets from HCP have proved the effectiveness and superiority of the proposed VAE-GAN framework for identifying both temporal features and functional brain networks compared with existing models, and the quality of fake data is higher than those from VAE and GAN. The results on resting state fMRI of Attention Deficit Hyperactivity Disorder (ADHD)-200 dataset further demonstrated that the fake data generated by the VAE-GAN can help improve the performance of brain network modeling and ADHD classification.
The iron-dopamine D1 coupling modulates neural signatures of working memory across adult lifespan
2023, NeuroImage
Brain iron overload and decreased integrity of the dopaminergic system have been independently reported as brain substrates of cognitive decline in aging. Dopamine (DA), and iron are co-localized in high concentrations in the striatum and prefrontal cortex (PFC), but follow opposing age-related trajectories across the lifespan. DA contributes to cellular iron homeostasis and the activation of D1-like DA receptors (D1DR) alleviates oxidative stress-induced inflammatory responses, suggesting a mutual interaction between these two fundamental components. Still, a direct in-vivo study testing the iron-D1DR relationship and their interactions on brain function and cognition across the lifespan is rare. Using PET and MRI data from the DyNAMiC study (n=180, age=20-79, %50 female), we showed that elevated iron content was related to lower D1DRs in DLPFC, but not in striatum, suggesting that dopamine-rich regions are less susceptible to elevated iron. Critically, older individuals with elevated iron and lower D1DR exhibited less frontoparietal activations during the most demanding task, which in turn was related to poorer working-memory performance. Together, our findings suggest that the combination of elevated iron load and reduced D1DR contribute to disturbed PFC-related circuits in older age, and thus may be targeted as two modifiable factors for future intervention.
A deep learning method for autism spectrum disorder identification based on interactions of hierarchical brain networks
2023, Behavioural Brain Research
It has been recently shown that deep learning models exhibited remarkable performance of representing functional Magnetic Resonance Imaging (fMRI) data for the understanding of brain functional activities. With hierarchical structure, deep learning models can infer hierarchical functional brain networks (FBN) from fMRI. However, the applications of the hierarchical FBNs have been rarely studied.
In this work, we proposed a hierarchical recurrent variational auto-encoder (HRVAE) to unsupervisedly model the fMRI data. The trained HRVAE encoder can predict hierarchical temporal features from its three hidden layers, and thus can be regarded as a hierarchical feature extractor. Then LASSO (least absolute shrinkage and selection operator) regression was applied to estimate the corresponding hierarchical FBNs. Based on the hierarchical FBNs from each subject, we constructed a novel classification framework for brain disorder identification and test it on the Autism Brain Imaging Data Exchange (ABIDE) dataset, a world-wide multi-site database of autism spectrum disorder (ASD). We analyzed the hierarchy organization of FBNs, and finally used the overlaps of hierarchical FBNs as features to differentiate ASD from typically developing controls (TDC).
The experimental results on 871 subjects from ABIDE dataset showed that the HRVAE model can effectively derive hierarchical FBNs including many well-known resting state networks (RSN). Moreover, the classification result improved the state-of-the-art by achieving a very high accuracy of 82.1 %.
This work presents a novel data-driven deep learning method using fMRI data for ASD identification, which could provide valuable reference for clinical diagnosis. The classification results suggest that the interactions of hierarchical FBNs have association with brain disorder, which promotes the understanding of FBN hierarchy and could be applied to other brain disorder analysis.
Sex differences in longitudinal changes of episodic memory-related brain activity and cognition in cognitively unimpaired older adults with a family history of Alzheimer's disease
2023, NeuroImage: Clinical
Episodic memory decline is an early symptom of Alzheimer’s disease (AD) – a neurodegenerative disease that has a higher prevalence rate in older females compared to older males. However, little is known about why these sex differences in prevalence rate exist. In the current longitudinal task fMRI study, we explored whether there were sex differences in the patterns of memory decline and brain activity during object-location (spatial context) encoding and retrieval in a large sample of cognitively unimpaired older adults from the Pre-symptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease (PREVENT-AD) program who are at heightened risk of developing AD due to having a family history (+FH) of the disease. The goal of the study was to gain insight into whether there are sex differences in the neural correlates of episodic memory decline, which may advance knowledge about sex-specific patterns in the natural progression to AD. Our results indicate that +FH females performed better than +FH males at both baseline and follow-up on neuropsychological and task fMRI measures of episodic memory. Moreover, multivariate data-driven task fMRI analysis identified generalized patterns of longitudinal decline in medial temporal lobe activity that was paralleled by longitudinal increases in lateral prefrontal cortex, caudate and midline cortical activity during successful episodic retrieval and novelty detection in +FH males, but not females. Post-hoc analyses indicated that higher education had a stronger effect on +FH females neuropsychological scores compared to +FH males. We conclude that higher educational attainment may have a greater neuroprotective effect in older +FH females compared to +FH males.
Learning brain representation using recurrent Wasserstein generative adversarial net
2022, Computer Methods and Programs in Biomedicine
To understand brain cognition and disorders, modeling the mapping between mind and brain has been of great interest to the neuroscience community. The key is the brain representation, including functional brain networks (FBN) and their corresponding temporal features. Recently, it has been proven that deep learning models have superb representation power on functional magnetic resonance imaging (fMRI) over traditional machine learning methods. However, due to the lack of high-quality data and labels, deep learning models tend to suffer from overfitting in the training process.
In this work, we applied a recurrent Wasserstein generative adversarial net (RWGAN) to learn brain representation from volumetric fMRI data. Generative adversarial net (GAN) is widely used in natural image generation and is able to capture the distribution of the input data, which enables the extraction of generalized features from fMRI and thus relieves the overfitting issue. The recurrent layers in RWGAN are designed to better model the local temporal features of the fMRI time series. The discriminator of RWGAN works as a deep feature extractor. With LASSO regression, the RWGAN model can decompose the fMRI data into temporal features and spatial features (FBNs). Furthermore, the generator of RWGAN can generate high-quality new data for fMRI augmentation.
The experimental results on seven tasks from the HCP dataset showed that the RWGAN can learn meaningful and interpretable temporal features and FBNs, compared to HCP task designs and general linear model (GLM) derived networks. Besides, the results on different training datasets showed that the RWGAN performed better on small datasets than other deep learning models. Moreover, we used the generator of RWGAN to yield fake subjects. The result showed that the fake data can also be used to learn meaningful representation compared to those learned from real data.
To our best knowledge, this work is among the earliest attempts of applying generative deep learning for modeling fMRI data. The proposed RWGAN offers a novel methodology for learning brain representation from fMRI, and it can generate high-quality fake data for the potential use of fMRI data augmentation.

View all citing articles on Scopus

View full text

Technical ReportSpatiotemporal analysis of event-related fMRI data using partial least squares

Abstract

Introduction

Section snippets

Partial least squares

fMRI experiments

Task-PLS

Discussion

Acknowledgements

Comput. Biomed. Res

NeuroImage

NeuroImage

NeuroImage

NeuroImage

Comput. Stat. Data Anal

NeuroImage

NeuroImage

NeuroImage

Neurotoxicol. Teratol

NeuroImage

NeuroImage

Modulation of connectivity in visual pathways by attention: cortical interactions evaluated with structural equation modeling and fMRI

Cereb. Cortex

Age-related differences in neural activity during memory encoding and retrieval: a positron emission tomography study

J. Neurosci

Corticolimbic interactions associated with performance on a short-term memory task are modified by age

J. Neurosci

Randomization Tests

Technical Report
Spatiotemporal analysis of event-related fMRI data using partial least squares