Multi-band EPI has primarily been applied to DTI and resting-state fMRI and there are an increasing number of examples of task-based fMRI studies. The benefit of multi-band EPI for task-based fMRI in general has been explored¹, however the specific benefits of multi-band EPI for event-related(EvR) paradigms have not been considered. With multi-band EPI, it should be possible to shorten the Tr to the point that the BOLD response of each event can be adequately sampled, thus significantly improving the efficiency of EvR fMRI studies. In addition densely sampled EvR responses should enhance the exploration of variations in the hemodynamic response, both as a function of stimulus variation and brain region. Here we have used two visual paradigms to explore the benefits of multi-band EPI for EvR fMRI.

To assess the benefit of decreased Tr for EvR fMRI a series of experiments were preformed on 4 subjects (35±4 yrs. old) using combinations of in-plane (r) and slice (s) acceleration. Five Trs were used 3.252s (r1s1), 2.0s (r2s1), 1.626s (r1s2), 1.0s (r2s2), and 0.813s (r1s4). For r1s4 FA was 35°, which was below the Ernst angle (56°), due to peak power limitations of the RF amplifier. For all other accelerations the FA was set to the Ernst angle. For all combinations of acceleration, the reconstructed matrix size was 64x64x40 with 3mm isotropic resolution. These data were acquired using a 16-channel receive array (Nova Medical) covering the entire brain. The stimulus paradigm used for each run was 14s rest followed by 19 repetitions of an 8Hz flickering checkerboard (1s on – 14s off).

An additional study was performed on one subject (36 yrs. old), using a previously published contrast adaptation paradigm². By using a multi-band factor of 4 and a receive array(Life Services LLC) targeting the visual cortex (16-channels over the posterior half of the brain) the temporal resolution was improved from 800ms to 500ms while increasing spatial coverage from 8 to 24 slices, covering the entire visual cortex with a 3mm isotropic spatial resolution. The stimulus consisted of 4 checkerboard circles, each 8° in diameter, one placed in each visual quadrant, 7° from the fixation point. Following a 60s period of adaptation to the baseline contrast (12.5% or 25%), the contrast would randomly increase or decrease by one or two octaves every 11~15s, for 3s, with a total run time of 600s. There were two runs for each baseline contrast, for a total of four runs.

Slice separation was performed using the slice GRAPPA algorithm³ and in-plane accelerations were reconstructed with GRAPPA⁴. All experiments were conducted on an Agilent 4T MRI system. The functional data was analyzed with mrTools⁵. For the Tr comparison study, a GLM design matrix consisting of the stimulus paradigm convolved with a double Gamma function and its temporal derivative was used. For the contrast adaptation study, a Finite Impulse Response⁶ (FIR) analysis was used.

Figure 1 shows example functional maps at each acceleration factor for one subject. Figure 2 shows the average number of activated voxels over the entire visual cortex at each Tr and acceleration. Both figures show that for Tr values of 2s or more (slice acceleration factor of 1) the extend of activation is substantially reduced. At these low sampling rates the peak of the BOLD response could be missed by as much as 1.0~1.6s and only 2~3 points are sampled during each response, leading to reduced contrast to noise and statistical power. The activation extent at the shortest Tr(0.813s, r1s4) isomewhat below those for Tr = 1.0 & 1.6s, in most subjects. This is likely due to the less than optimal flip angle. This limitation can be resolved by implementing VERSE³ or parallel transmit methods⁷.

Figure 3, shows responses to the contrast change plotted on the retinotopic map of the subject's right hemisphere. Figure 4, shows FIR determined hemodynamic responses to contrast changes in V1 and hV4. In Figure 4, V1 responses to decrements in contrast are negative as previously reported². The dense sampling of the BOLD response clearly shows that negative responses in V1 peak earlier (1.4s) than the positive responses. In hV4, the BOLD responses to all contrast changes are positive, as previously reported².

The higher temporal resolution afforded by multi-band EPI allows more efficient sampling of the BOLD response in EvR studies which can lead to dramatically enhanced statistical power and improved characterization of the hemodynamic response, even when using only modest levels of slice acceleration, such as those employed in this study.