Estimation of the intrinsic dimensionality of fMRI data

Cordes, Dietmar; Nandy, Rajesh

doi:10.1016/j.neuroimage.2005.07.054

Cited by 61 publications

(67 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, the additional application of ICA on a subspace of reduced dimensionality helps reduce the r LeVan and Gotman r r 2028 r degradation that occurs when overfitting the ICA model with too many components in the original decompositions [Li et al, 2007]. It should also be noted that the dimensionality used for the final application of ICA was higher than some values reported in the literature [Cordes and Nandy, 2006;Li et al, 2007;Rodionov et al, 2007], which were less than 100, although we had larger datasets. Extracting a large number of components may prevent relevant signals of very low amplitude from being missed by the decomposition.…”

Section: Component Identificationmentioning

confidence: 85%

“…Methods based on the application of information-theoretic criteria to the eigenspectrum can provide a good estimate of the dimensionality of the data, but only if the covariance of the noise is known [Cordes and Nandy, 2006]. Otherwise, the estimated number of sources tends to be proportional to the number of time points in the fMRI acquisition, an unrealistic assumption.…”

Section: Independent Component Analysismentioning

confidence: 99%

“…However, the results may depend on the number of components selected for the decomposition. Currently, there is no consensus on the most appropriate way of determining the dimensionality of the fMRI dataset, and several methods have been proposed [Beckmann and Smith, 2004;Cordes and Nandy, 2006;Li et al, 2007;Yang et al, 2008]. Nevertheless, identifying reproducible components in multiple ICA decompositions at least eliminates noisy components due to the stochastic nature of the algorithm.…”

Section: Component Identificationmentioning

confidence: 99%

See 2 more Smart Citations

Independent component analysis as a model‐free approach for the detection of BOLD changes related to epileptic spikes: A simulation study

LeVan¹,

Gotman²

2008

Human Brain Mapping

View full text Add to dashboard Cite

EEG-fMRI in epileptic patients is commonly analyzed using the general linear model (GLM), which assumes a known hemodynamic response function (HRF) to epileptic spikes in the EEG. In contrast, independent component analysis (ICA) can extract Blood-Oxygenation Level Dependent (BOLD) responses without imposing constraints on the HRF. This technique was evaluated on data generated by superimposing artificial responses on real background fMRI signals. Simulations were run using a wide range of EEG spiking rates, HRF amplitudes, and activation regions. The data were decomposed by spatial ICA into independent components. A deconvolution method then identified component time courses significantly related to the simulated spikes, without constraining the shape of the HRF. Components matching the simulated activation regions ("concordant components") were found in 84.4% of simulations, while components at discordant locations were found in 12.2% of simulations. These false activations were often related to large artifacts that coincidentally occurred simultaneously with some of the random simulated spikes. The performance of the method depended closely on the simulation parameters; when the number of spikes was low, concordant components could only be identified when HRF amplitudes were large. Although ICA did not depend on the shape of the HRF, data processed with the GLM did not reveal the appropriate activation region when the HRF varied slightly from the canonical shape used in the model. ICA may thus be able to extract BOLD responses from EEG-fMRI data in epileptic patients, in a way that is robust to uncertainty and variability in the shape of the HRF.

show abstract

Section: Component Identificationmentioning

confidence: 85%

Section: Independent Component Analysismentioning

confidence: 99%

Section: Component Identificationmentioning

confidence: 99%

See 1 more Smart Citation

Independent component analysis as a model‐free approach for the detection of BOLD changes related to epileptic spikes: A simulation study

LeVan¹,

Gotman²

2008

Human Brain Mapping

View full text Add to dashboard Cite

show abstract

“…(d) A set of time-dependent weights W s for each subject was calculated by projecting the centered dynamic FC matrix onto a few eigenconnectivities. 2000), overestimate the dimensionality in the presence of correlated noise (Cordes and Nandy, 2006). Comparing the eigenvalues to a null distribution obtained using surrogate data avoids making parametric assumption about the noise.…”

Section: Comparing Hc Subjects and Rrms Patientsmentioning

confidence: 99%

Principal components of functional connectivity: A new approach to study dynamic brain connectivity during rest

et al. 2013

View full text Add to dashboard Cite

Functional connectivity (FC) as measured by correlation between fMRI BOLD time courses of distinct brain regions has revealed meaningful organization of spontaneous fluctuations in the resting brain. However, an increasing amount of evidence points to non-stationarity of FC; i.e., FC dynamically changes over time reflecting additional and rich information about brain organization, but representing new challenges for analysis and interpretation. Here, we propose a data-driven approach based on principal component analysis (PCA) to reveal hidden patterns of coherent FC dynamics across multiple subjects. We demonstrate the feasibility and relevance of this new approach by examining the differences in dynamic FC between 13 healthy control subjects and 15 minimally disabled relapse-remitting multiple sclerosis patients. We estimated whole-brain dynamic FC of regionally-averaged BOLD activity using sliding time windows. We then used PCA to identify FC patterns, termed "eigenconnectivities", that reflect meaningful patterns in FC fluctuations. We then assessed the contributions of these patterns to the dynamic FC at any given time point and identified a network of connections centered on the default-mode network with altered contribution in patients. Our results complement traditional stationary analyses, and reveal novel insights into brain connectivity dynamics and their modulation in a neurodegenerative disease.© 2013 Elsevier Inc. All rights reserved. IntroductionSpontaneous fluctuations of the functional MRI (fMRI) bloodoxygen-level-dependent (BOLD) signal are not random but temporally coherent between distinct brain regions. While these fluctuations were long considered as "noise", Biswal et al. (1995) showed that fluctuations of motor areas were correlated even in the absence of a motor task. Several other networks of coherent BOLD activity between remote brain regions have since been identified, including visual, auditory, language and attention networks, and a network called the "default mode network" (DMN) which reduces its activity during attentiondemanding tasks. These networks of regions with coherent activity during rest are consistent across subjects and closely resemble the brain's functional organization of evoked responses (Damoiseaux et al., 2006;Fox and Raichle, 2007;Laird et al., 2011;Smith et al., 2009). Coherent BOLD activity persists during sleep and in anesthetized monkeys, suggesting that it reflects a fundamental property of the brain's functional organization (Larson-Prior et al., 2009;Vincent et al., 2007).Coherent BOLD activity, known as "functional connectivity" (FC), is modulated by learning (Bassett et al., 2011), cognitive and affective states (Cribben et al., 2012;Ekman et al., 2012;Eryilmaz et al., 2011;Richiardi et al., 2011;Shirer et al., 2012) and also spontaneously Chang and Glover, 2010; Kitzbichler et al., 2009). Chang andGlover (2010) showed that FC between the posterior cingulate cortex, a key region of the default mode network, and various other brain regions was highly dy...

show abstract

“…Recall also that it is s n that characterizes the features of the HRF that are "really there". In practice, however, s n , termed the "intrinsic dimensionality of fMRI data" (Cordes and Nandy, 2006), is unknown for real fMRI data. As far as we know, all published work assumes that…”

Section: Figure 1 An Illustrative Plot Of the Hrf H(t J ) With T J =mentioning

confidence: 99%

Regularized estimation of hemodynamic response function for fMRI data

Zhang

2010

Statistics and Its Interface

View full text Add to dashboard Cite

One of the primary goals in analyzing fMRI data is to estimate the Hemodynamic Response Function (HRF), which is a large-dimensional parameter vector possessing some form of sparsity. This paper introduces a varyingdimensional model for the HRF, and develops novel regularization methods for estimating the HRF from fMRI time series via incorporating the sparsity feature. Particularly, we present three types of penalty choice methods: the Lasso, the adaptive Lasso and the SCAD. Simulation studies demonstrate the advantages of regularization methods, in terms of sparsity recovery, over conventional non-regularized approaches which restrict the HRF to be fixed low dimensional without capturing the sparsity structure. We illustrate the regularized methods for estimating the HRF using a real fMRI data set and compare with results offered by a popular imaging analysis tool AFNI.

show abstract

Estimation of the intrinsic dimensionality of fMRI data

Cited by 61 publications

References 21 publications

Independent component analysis as a model‐free approach for the detection of BOLD changes related to epileptic spikes: A simulation study

Independent component analysis as a model‐free approach for the detection of BOLD changes related to epileptic spikes: A simulation study

Principal components of functional connectivity: A new approach to study dynamic brain connectivity during rest

Regularized estimation of hemodynamic response function for fMRI data

Contact Info

Product

Resources

About