Purpose: Acoustic noise has always been a problem for fMRI, especially if acoustic stimuli
are used. However, moving to higher field strength enhances this problem drastically.
The acoustic noise can mask the stimulus, and the stimulus-bound fMRI-signal might
be contaminated or even eliminated by the BOLD-signal of the scanner noise. Using
non-continuous acquisition schemes by separating the acquisition of successive volumes
by a couple of seconds (so called “SPARSE“ designs) avoids the acoustic masking of
the stimuli and allows for deconvolution of the haemodynamic response function (hrf)
induced by the acoustic noise itself and the stimulus bound hrf. However, these paradigms
are extremely inefficient (regarding time and statistical power) and inflexible in
the view of timing. The optimal inter-stimulus interval is about 9s, and it is not
always possible to deliver the stimuli and to acquire the volunteer response and the
MRI data within this short time slot. Parallel imaging techniques might help in this
situation by increasing the acquisition speed by a factor of 2–3. However, applying
parallel imaging techniques as sensitivity encoding (SENSE) lead to a reduction of
SNR, and it is questionable if the combination of SENSE with SPARSE designs provides
sufficient power to detect brain activations during an fMRI experiment. Goal of this
study was to investigate if SENSE and SPARSE-designs can be combined at high field
strength and still have sufficient power to detect main effects and differential effects
in fMRI-paradigms.
Methods: 16 right-handed volunteers were examined using an acoustic paradigm. The paradigm
was designed to investigate multi-sensory integration by presenting visual and acoustic
stimuli. The visual stimuli consisted of short video sequences of a vertically jumping
person. The acoustic stimuli represented the jump acoustically by modulation of the
amplitude of the standard pitch A with the pressure applied to the ground during the
jump (='sonification'). The volunteers had to judge the height of the jump based either
on the visual information (Cond A), visual and acoustic information (Cond B) or the
acoustic information alone (Cond C). The Cond B contained two subconditions using
either concordant acoustic stimuli (Cond B1) or discordant acoustic stimuli (Cond
B2), the latter condition presenting a sound that did not fit to the video seen on
the screen. The MRI-examinations were performed on a Achieva 3.0T whole body MRI (Philips,
Best, Netherlands), using a 8 channel SENSE head coil (MRI-Devices) and a GE-Single
Shot EPI sequence with a SENSE-Factor of two (TE/TR/Flip=35/1.54/90°, 3.6×3.6×3.6mm3, 121 dyn. scans / total scan time: ~ 18min).
We calculated three contrasts: 1. Cond A – Cond C, 2. Cond C – Cond A and 3. Cond
B1– Cond B2. The first contrast should reveal the visual stream, whereas the second
contrast should show the primary and secondary auditory areas as well as the acoustic
association cortex. The third contrast was used to test if even subtle differences
can be detected. In particular, Cond B1– Cond B2 should show an activation of the
multimodal integration areas around the superior temporal sulcus (STS). All contrasts
were examined at a threshold of p<0.05 corrected.
Results: The contrast (Cond A – Cond C) did show the activation of the complete dorsal visual
stream, whereas the opposite contrast (Cond C – Cond A), revealed the activation of
the auditory system including the auditory association areas. The differential contrast,
Cond B1- Cond B2, showed activation of the STS as expected.
Conclusion: Using a sonification paradigm we were able to show that at high field strength even
extremely inefficient fMRI-designs, as the combination of SENSE and SPARSE-designs,
can be successfully combined and are capable of detecting even subtle activation changes.