“School of Cognitive Sciences”
Back to Papers HomeBack to Papers of School of Cognitive Sciences
| Paper IPM / Cognitive Sciences / 14227 |
|
||||||||||
| Abstract: | |||||||||||
|
Introduction: Due to the mechanism of contrast origin in BOLD fMRI, generally only the magnitude data is used for activation detection. Through the use of SSFP sequences, we are able to detect fMRI activation directly based on phase variation induced by frequency shift. BOSS fMRI utilizes this phase information in addition to magnitude changes for activation detection1, 2. The contribution of phase, however, is highly dependent on the position of the center frequency on the SSFP profile2 and therefore may not be quantified accurately. Averaged-BOSS (A-BOSS)3, which has recently been introduced for fMRI by compressing the SSFP profile into a single voxel, also indicates that activation patterns should exist in both phase and magnitude images. Figure 1 graphically shows this concept. Because of its averaging nature, this method provides a novel approach for quantitative analysis of fMRI induced phase changes while removing its sensitivity to the off-resonance frequency. Method: A-BOSS signal is essentially similar to BOSS, but the SSFP profile is averaged across each single voxel. In this context fMRI activation is a change of the voxel signal in the complex domain. Figure 2 shows this idea. Here S and ô show the fMRI magnitude and phase change due to activation, respectively.
Bloch equation simulation was performed for a voxel inside the blood veins where an
oxygenation change from 70fMRI data was acquired from a single subject on a 3T Siemens TIM Trio scanner.
Data was collected during a right-hand finger tapping experiment with a block design of 3
cycles of on (30sß0s), TRms\4.9ms, FA=20º, matrix size 128x128 and resolution 2x2x5 mm3.
Four axial slices and a total of 30 volumes were collected in 3 minutes. Both magnitude and phase images were
acquired from scanner. Preprocessing was performed using SPM 8 including head motion correction, slice
timing correction and smoothing with a Gaussian kernel of twice the voxel size. Activation detection was
performed for magnitude and phase data separately, based on the cross correlation of amplitude and phase time
series with the temporal pattern of task paradigm. The cross correlation values were then converted to z-score.
Download TeX format |
|||||||||||
| back to top | |||||||||||
