Introduction

Blood oxygen-level dependent (BOLD) functional MRI (fMRI) based on gradient-echo echo-planar imaging (EPI) is the most commonly used fMRI method due to its high sensitivity and robustness. However, its fidelity in reflecting the blood flow response to neuronal activity is compromised by the generally dominant contribution of large veins that drain the activated area. This limits the spatial accuracy of typical fMRI experiments to a few millimeters. While special stimulation tasks and processing approaches may ameliorate this problem somewhat, alternative fMRI contrasts less sensitive to venous drainage may be desirable for high resolution studies. For example, both perfusion and blood volume (BV) based fMRI methods have been demonstrated to improve spatial accuracy.^1,2,3 These methods suffer from reduced sensitivity however. Here we investigated the utility of Single Shot Perfusion Labeling (SSPL)⁴ to perform high resolution fMRI experiments, with as ultimate aim laminar fMRI⁵. SSPL has been demonstrated to improve sensitivity over conventional pulsed labeling based fMRI experiments about 2-fold by using two inversion pulses to suppress stationary tissue signal and eliminate the need for a reference scan. It may also have advantages over BV based methods as heuristically according to Grubb's law,⁶ perfusion changes substantially exceed BV changes.

Materials and Methods

Experiments were performed on a Siemens Magnetom 7T whole-body scanner with 32-channel Nova Medical receive coil using an in-house developed pulse sequence that allowed conventional BOLD-EPI as well as SSPL-EPI. To optimize parameters for 7 T, data were acquired using 2³ mm³ (8 μl) voxels, rate-2 SENSE, 3 s TR and 17.5 ms TE. A multi-slice implementation was investigated in which several imaging slices were acquired around the nominal second inversion time (TI2). Here 5-slice SSPL was acquired; BOLD-EPI with similar resolution and echo time allowed the acquisition of 45 slices. A FAIR-like option was added to the SSPL sequence in which inversion was non-selective every other TR to allow contrast interpretation and perfusion quantification. Two slice-selective hyperbolic secant pulses (10240 μs duration; 800 Hz amplitude; 30/120 mm thickness) were used for the inversions. Functional MRI used a 7.5 Hz full-field checkerboard task with 30 s off/30 s on blocks, ~5.5 min scan duration (110 repetitions). A limited number of high-resolution scans with 1³ mm³ (1 μl) resolution were also acquired (n=3, rate-3 SENSE; 38 ms TE; 5 slices; 110 or 210 repetitions; BOLD: 33 slices).

Results and Discussion

The combination of TI1 and TI2 was optimized for simultaneous suppression of gray and white matter. TI1 was fixed at 1400 ms, optimal background signal suppression was then obtained around 450-475 ms TI2 (Figure 1). Slice excitation and EPI readout took almost 50 ms, therefore most 5-slice SSPL experiments were performed using 350, 400, 450, 500 and 550 ms TI2. Effects of additional RF pulses to excite neighboring slices were not observed when comparing the 5-slice approach to comparable conventional 1-slice SSPL (Figure 2). This was supported by FAIR-like data from similar 5-slice experiments for various TI2 ranges (data not shown). Robust activation was found for TI2 ≤450 ms, longer TI2s showed substantially reduced activation (Figure 3 and Table 1). Note that the sign of the activation was inverted for longer TI2 values since the sign of the background signal had flipped. Very low background signal (~450 ms TI2) was found to hamper analysis, as either real, task-induced signal changes, or noise, could change signal sign, an effect lost in the analysis of magnitude data that can only be partially addressed by analysis based on complex signal (under development). Overall the level of activation observed in SSPL was reduced compared to BOLD, both when comparing activation size and contrast-to-noise ratio (CNR) (Table 1). Note however that the reduced activation size could in part reflect improved localization of the activation. SSPL's reduced sensitivity to physiological fluctuations is evident from the ~two-fold higher temporal stability (tSNR) at 8 μl resolution (Table 1). Preliminary data acquired at the higher 1 μl resolution reflect these findings (e.g. see Figure 5), showing that activation can be measured using SSPL fMRI at 1 μl resolution, albeit with BOLD showing more robust activation. Additional high-resolution data will be required to address the question of improved spatial localization of SSPL when compared to BOLD.

Conclusion

The feasibility of high-resolution SSPL for perfusion-based fMRI at high field, at up to 1 μl resolution, was demonstrated. Whereas BOLD fMRI at the same resolution provides both higher sensitivity and coverage, spatial specificity concerns make non-BOLD approaches of interest for high-fidelity laminar fMRI.

Acknowledgements

This work was supported by the Intramural Research Program of NINDS, National Institutes of Health.