Limits...
Topographic factor analysis: a Bayesian model for inferring brain networks from neural data.

Manning JR, Ranganath R, Norman KA, Blei DM - PLoS ONE (2014)

Bottom Line: The neural patterns recorded during a neuroscientific experiment reflect complex interactions between many brain regions, each comprising millions of neurons.However, the measurements themselves are typically abstracted from that underlying structure.Specifically, TFA casts each brain image as a weighted sum of spatial functions.

View Article: PubMed Central - PubMed

Affiliation: Princeton Neuroscience Institute, Princeton University, Princeton, New Jersey, United States of America; Department of Computer Science, Princeton University, Princeton, New Jersey, United States of America.

ABSTRACT
The neural patterns recorded during a neuroscientific experiment reflect complex interactions between many brain regions, each comprising millions of neurons. However, the measurements themselves are typically abstracted from that underlying structure. For example, functional magnetic resonance imaging (fMRI) datasets comprise a time series of three-dimensional images, where each voxel in an image (roughly) reflects the activity of the brain structure(s)-located at the corresponding point in space-at the time the image was collected. FMRI data often exhibit strong spatial correlations, whereby nearby voxels behave similarly over time as the underlying brain structure modulates its activity. Here we develop topographic factor analysis (TFA), a technique that exploits spatial correlations in fMRI data to recover the underlying structure that the images reflect. Specifically, TFA casts each brain image as a weighted sum of spatial functions. The parameters of those spatial functions, which may be learned by applying TFA to an fMRI dataset, reveal the locations and sizes of the brain structures activated while the data were collected, as well as the interactions between those structures.

Show MeSH
Predicting the covariance structure of an fMRI dataset.A. Each dot reflects the covariance between a pair of images from a single participant (-axis: observed, -axis: estimated) using  sources. The correlation reported in the panel is between entries in the two covariance matrices. B. We also used TFA to estimate the covariance structure of held-out data, using a 6-fold cross validation procedure. The panel displays the median correlations ( bootstrap-estimated 95% confidence intervals) between the observed and estimated covariance matrices (of held out data), as a function of the number of sources we fit. The medians are taken across the 6 folds and 9 participants, and the error bars reflect across-participant variability.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4012983&req=5

pone-0094914-g006: Predicting the covariance structure of an fMRI dataset.A. Each dot reflects the covariance between a pair of images from a single participant (-axis: observed, -axis: estimated) using sources. The correlation reported in the panel is between entries in the two covariance matrices. B. We also used TFA to estimate the covariance structure of held-out data, using a 6-fold cross validation procedure. The panel displays the median correlations ( bootstrap-estimated 95% confidence intervals) between the observed and estimated covariance matrices (of held out data), as a function of the number of sources we fit. The medians are taken across the 6 folds and 9 participants, and the error bars reflect across-participant variability.

Mentions: In addition to examining the quality of individual image reconstructions, we may also wish to know the extent to which TFA preserves the covariance structure across all of a participant’s images. As shown in Figure 6A, we computed the observed across-image covariance matrix for one participant and compared it to the TFA-estimated across-image covariance matrix (using sources). Each dot in the figure reflects a single entry in one of these by covariance matrices (correlation between entries in the observed and estimated covariance matrices: ).


Topographic factor analysis: a Bayesian model for inferring brain networks from neural data.

Manning JR, Ranganath R, Norman KA, Blei DM - PLoS ONE (2014)

Predicting the covariance structure of an fMRI dataset.A. Each dot reflects the covariance between a pair of images from a single participant (-axis: observed, -axis: estimated) using  sources. The correlation reported in the panel is between entries in the two covariance matrices. B. We also used TFA to estimate the covariance structure of held-out data, using a 6-fold cross validation procedure. The panel displays the median correlations ( bootstrap-estimated 95% confidence intervals) between the observed and estimated covariance matrices (of held out data), as a function of the number of sources we fit. The medians are taken across the 6 folds and 9 participants, and the error bars reflect across-participant variability.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4012983&req=5

pone-0094914-g006: Predicting the covariance structure of an fMRI dataset.A. Each dot reflects the covariance between a pair of images from a single participant (-axis: observed, -axis: estimated) using sources. The correlation reported in the panel is between entries in the two covariance matrices. B. We also used TFA to estimate the covariance structure of held-out data, using a 6-fold cross validation procedure. The panel displays the median correlations ( bootstrap-estimated 95% confidence intervals) between the observed and estimated covariance matrices (of held out data), as a function of the number of sources we fit. The medians are taken across the 6 folds and 9 participants, and the error bars reflect across-participant variability.
Mentions: In addition to examining the quality of individual image reconstructions, we may also wish to know the extent to which TFA preserves the covariance structure across all of a participant’s images. As shown in Figure 6A, we computed the observed across-image covariance matrix for one participant and compared it to the TFA-estimated across-image covariance matrix (using sources). Each dot in the figure reflects a single entry in one of these by covariance matrices (correlation between entries in the observed and estimated covariance matrices: ).

Bottom Line: The neural patterns recorded during a neuroscientific experiment reflect complex interactions between many brain regions, each comprising millions of neurons.However, the measurements themselves are typically abstracted from that underlying structure.Specifically, TFA casts each brain image as a weighted sum of spatial functions.

View Article: PubMed Central - PubMed

Affiliation: Princeton Neuroscience Institute, Princeton University, Princeton, New Jersey, United States of America; Department of Computer Science, Princeton University, Princeton, New Jersey, United States of America.

ABSTRACT
The neural patterns recorded during a neuroscientific experiment reflect complex interactions between many brain regions, each comprising millions of neurons. However, the measurements themselves are typically abstracted from that underlying structure. For example, functional magnetic resonance imaging (fMRI) datasets comprise a time series of three-dimensional images, where each voxel in an image (roughly) reflects the activity of the brain structure(s)-located at the corresponding point in space-at the time the image was collected. FMRI data often exhibit strong spatial correlations, whereby nearby voxels behave similarly over time as the underlying brain structure modulates its activity. Here we develop topographic factor analysis (TFA), a technique that exploits spatial correlations in fMRI data to recover the underlying structure that the images reflect. Specifically, TFA casts each brain image as a weighted sum of spatial functions. The parameters of those spatial functions, which may be learned by applying TFA to an fMRI dataset, reveal the locations and sizes of the brain structures activated while the data were collected, as well as the interactions between those structures.

Show MeSH