Neuroimaging
Laboratory
Idiographic Interpolation of Hippocampal fMRI Data
Julie Petersen, J. Michael Williams, PhD, & Karol Osipowicz, PhD
Department of Psychology, Drexel University
Introduction
Results from Standard Analysis
fMRI activation of the hippocampus is elusive due to the depth of its
location in the brain and tonic activity. Traditional approaches to
fMRI data processing normalize the data to an MNI atlas, with a
resultant resolution of 2mm3. While this approach is good for most
brain areas, it does not fully utilize the potential interpolated
resolution of fMRI data. Whereas standard preprocessing can
produce voxel sizes of 2mm3, the native resolution of Anatomical
data can reliably reach .5mm3. This increase in resolution would
make finding activation of small, deep, and difficult to image
structures, like the hippocampus, far less susceptible to beta-error.
All results were reported at FWE corrected p<.05, with a cluster
extent consistent with image smoothness. Participants had the best
recall for pictures. Emotional stimuli were remembered better than
neutral stimuli. The right hippocampus was more active than the left
for encoding of images. See Fig. 1.
Objective
Department of Psychology
Secondary Results from Idiographic
Interpolation Analysis
Fig. 3 Activation of the hippocampus on an individual-subject level
Fig. 1 Standard fMRI analyses
revealed an effect of posterior
hippocampal activation
associated with both neutral and
emotional pictures
In this analysis, we compare the results of two processing schemes:
standard and idiographic interpolated. We hypothesize that
activation of the hippocampus will be evident in the idiographic
interpolated and not in the standard data.
Methods
21 healthy, normal adults (10 males, Mage = 25 yrs, SDage = 4) were
scanned using a 3.0 T scanner. Patterns of hippocampal activations
were examined over the course of six runs. Emotional and neutral
pictures, words, and environmental sounds were alternated during a
single session. Each run lasted 4.2 minutes for a total of 84
volumes each run. Memory for the stimuli was assessed following
the scanning. Data was analyzed using SPM (Flandin & Friston,
2008) and CONN software.
Standard Analysis
Data preprocessing included slice time correction, realignment,
coregistration (MNI) (Schönecker et al., 2009), normalization (MNI)
(Rodionov et al., 2009), and smoothing (5mm). Data underwent first
and second level modeling
Primary Results from Idiographic
Interpolation Analysis
All results were reported at FWE corrected p<.05, with a cluster
extent consistent with image smoothness. Idiographic interpolation
analyses revealed bilateral activation of both the anterior and
posterior hippocampi for emotional v. neutral pictures (see Fig. 2
and Fig. 3), as opposed to bilateral posterior hippocampal activation
present only with emotional and neutral pictures (see Fig. 1).
Fig. 2 Group idiographic
interpolation fMRI analyses
Idiographic Interpolated Analysis
revealed an effect of bilateral
posterior and anterior
Data preprocessing included the standard method of slice time
hippocampal activation
correction and realignment. Normalization utilized the standard
algorithm, but each participant’s T2 data was normalized to his own associated with emotional
T1 space, instead of normalizing the data to AAL Atlas. This allowed v. neutral pictures
resolution to increase from the standard 2 x 2 x 2 mm per voxel to
0.5 x 0.5 x 1 mm per voxel. The AAL Atlas was then coregistered
and normalized to each participant’s T1 data, giving each participant
their own atlas. Data within the hippocampus, parahippocampal
gyrus, and amygdala were smoothed. Each of these regions then
underwent first level modeling, as data for each region was
extracted from the functional data and modeled across time for
each event. A sample-specific template was generated. Second
level comparisons were made according to the standard method.
Discussion
Of our 21 participants, 6 participants were excluded from the
idiographic interpolation analyses due to problematic data. Similar
analyses could be run for these participants, but would require
subject-specific drawing of ROIs. The idiographic interpolation
analyses did not rely on any anatomical determination, but instead
was purely atlas driven. The bilateral hippocampal activation of all
15 participants indicate this analytical technique as a great method
for clinical hippocampal imaging.
Conclusion
Our results provide some support for the technique, but group level
comparisons with this data are problematic, thus further research is
necessary to establish the utility of this technique.
References
Flandin, G., & Friston, K. J. (2008). Statistical parametric mapping (SPM).
Scholarpedia, 3(4), 6232.
Rodionov, R., Chupin, M., Williams, E., Hammers, A., Kesavadas, C., & Lemieux, L.
(2009). Evaluation of atlas-based segmentation of hippocampi in
healthy humans. Magnetic resonance imaging, 27(8), 1104-1109.
Schönecker, T., Kupsch, A., Kühn, A., Schneider, G.-H., & Hoffmann, K.-T. (2009).
Automated Optimization of Subcortical Cerebral MR Imaging− Atlas
Coregistration for Improved Postoperative Electrode Localization in
Deep Brain Stimulation. American Journal of Neuroradiology, 30(10),
1914-1921.