 |
|
|
|
|
|
|
| Principal investigators: | James J. Pekar, Ph.D., Associate Professor of Radiology |
| The past decade has seen the
advance of functional MRI (fMRI) methodologies useful for non-invasive study of
brain activity. These developments have revolutionized cognitive neuroscience,
and MRI has become a popular tool for investigations in human neuroscience.
However, MRI is fundamentally a low-sensitivity technique; the same factors
(e.g., low energy per photon) that make MRI non-invasive also render it
insensitive, and so the spatial resolution, temporal resolution, and
signal-to-noise of fMRI are relatively poor. One way to increase the
sensitivity of MRI is to increase the static magnetic field. Our research center
opened with a 1.5 Tesla scanner, then added a 3.0 Tesla scanner. We are now
preparing for a 7.0 Tesla scanner, which will further increase signal-to-noise.
Optimization of 7.0 Tesla scanning is required, to “trade” these
signal-to-noise increases for increases in spatial or temporal resolution, as
appropriate. Functional MRI yields large data sets in which the time courses of voxels have been sensitized to the hemodynamic sequelae of brain activation. These “brain movies” must then be analyzed to yield spatial and temporal summaries. Standard analytic approaches employ voxel-wise tests of a priori hypotheses. Exploratory data analysis, on the other hand, approaches these large data sets without specific prior hypotheses, and aims to discover within the data features reporting upon the organization of brain activity. In particular, spatial independent component analysis (s-ICA, or just ICA) seeks to express an fMRI data set as a sum of products of spatial maps and their respective time courses, such that the maps are drawn from statistically independent distributions, consistent with the princple of modular organization of brain function. ICA of fMRI data often reveals brain activations which, because they do not smoothly follow paradigm events, may not be predicted by a “bottom-up” approach, and are therefore omitted from standard hypothesis-testing analyses. However, such unanticipated components must be evaluated for specificity: Are they hemodynamic sequelae of neuronal activity, or are they artifacts (e.g., respiration, cardiac pulsations, head motion)? Also, ICA can be applied to fMRI data acquired during rich naturalistic behaviors - such as simulated automobile driving [see figure], movie-watching, sleep, or the “resting state” – which do not lend themselves to analysis with standard approaches. This may be especially valuable in clinical populations who would have difficulty complying with conventional brain imaging paradigms. The developments proposed (optimization of 7.0 T scanning; evaluation of independent components of fMRI data; development of approaches to rich naturalistic behaviors) serve the ultimate aim of improving the utility of fMRI for investigations in basic and clinical neuroscience. (Related Publications)
Hypothesized neural substrates of simulated driving. Functional MRI data were acquired during epochs of simulated driving, and analyzed using ICA. From: VD Calhoun, JJ Pekar, VB McGinty, T Adali, TD Watson, & GD Pearlson. "Different Activation Dynamics in Multiple Neural Systems During Simulated Driving." Human Brain Mapping 16:158 (2002). |
|
|
|
|
|
© 2001 The
Kennedy Krieger Institute. Baltimore, Maryland.
All rights reserved. Last Updated 02/17/00; Send Email |
