Introduction

Although echo-volume imaging (EVI)^1,2 can provide fast image acquisition, it is sensitive to image distortion and blurring due to its long echo train length (ETL) ^3,4, and thus it has not been widely used in fMRI studies. This study aims to develop an EVI-based pulse sequence (named k-t rFOV-EVI) to mitigate the aforementioned problems and further improve the acquisition efficiency. These were achieved by employing three-dimension restricted field-of-view (rFOV) imaging⁵ and random (k, t)-space undersampling⁶. Human visual fMRI experiments were conducted to demonstrate the capability of k-t rFOV-EVI in fast fMRI data acquisition over a focused imaging region. The performance of k-t rFOV-EVI was also compared to that of the simultaneous multi-slice EPI (SMS-EPI), which is the mainstream pulse sequence used in fast fMRI.

Methods

k-t rFOV-EVI pulse sequence: As shown in Fig. 1A, the k-t rFOV-EVI sequence was built upon a gradient-echo multi-shot EVI sequence³. To reduce the ETL, a slab-selective 2D RF pulse⁵ was used to excite a partial 3D field-of-view. The RF pulse consisted of 11 sub-pulses that were modulated by an envelope pulse with pulse-width = 14.7 ms. The time-bandwidth product of sub-pulse/envelope pulse was 3.01/3.53. After one RF excitation (shot), blipped phase-encoding gradients were applied in the through-slab direction (or z-direction) to acquire M k_x-k_y planes in k-space using EPI readouts. A total of P (= N_z/M) shots are needed to acquire a 3D k-space matrix with size = N_x × N_y × N_z. To accelerate the image acquisition, the central (k_z,c) and outer (k_z,o) k_z regions were respectively fully sampled and undersampled by varying the amplitudes of the blipped gradients in z. The sampling patterns of k_z,o was randomly changed across different image volumes (time frames) (Fig. 1B). Compared to the conventional multi-shot EVI, the random (k, t)-space undersampling scheme reduced the number of acquired k_x-k_y planes from N_z to N_s at a given time frame, and the number of shots and the volume TR were reduced by a factor of R (= N_z/N_s) accordingly. No acceleration was applied to the k_x-k_y planes. The k-space time series acquired with k-t rFOV-EVI were reconstructed to full k-space data using the PS-Sparse algorithm^6,7. Finally, magnitude images were obtained with a 3D-FFT (Fig. 1B).

fMRI data acquisition and analysis: fMRI experiments were conducted in healthy subjects on a GE MR750 3T scanner with a 32-channel head coil. A 4-min paradigm with interleaved 16-s blank (black) screen and 16-s contrast-reversing (8-Hz) checkerboard was used for visual stimulation. fMRI images covering the visual cortex were acquired using the k-t rFOV-EVI sequence: FOV=19.2×8.0×6.0 cm³, matrix=76×32×24 (voxel size = 2.5×2.5×2.5 mm³), 3 shots with M = 2, the number of samples in k_z,c/k_z,o = 2/4 (R = 4), TR/volume-TR/TE = 80/240/30 ms, and flip angle = 20°. For comparison, an additional fMRI scan was performed with a commercial SMS-EPI sequence and the same spatial resolution: FOV=19.2×19.2cm², 24 slices with slice acceleration factor = 8 and without in-plane acceleration, TR/TE = 240/30ms, and flip angle = 34°. The fMRI data were analyzed in AFNI. Motion correction and spatial smoothing with FWHM = 3 mm were applied to the data prior to the general linear model analysis. The t-maps were thresholded with p<0.0001 and cluster corresponding to FPR<0.05 to detect the brain activation. The number of activated voxels and average t-value were measured in the visual cortex.

Results

Fig. 2 illustrates SMS-EPI and k-t rFOV-EVI images from a representative subject. It can be seen that severe g-factor noise appeared in the SMS-EPI images due to the short distance between simultaneously excited slices (=7.5 mm). Compared to the SMS-EPI, k-t rFOV-EVI provided better image quality for visualizing the brain structures. In the fMRI activation maps (Fig. 3), there were 36% more activated voxels in the visual cortex in the k-t 3D-rFOVI (2062 voxels) than in the SMS-EPI (1521 voxels) scan, and the average t-value of activated foci was elevated in the k-t rFOV-EVI scan (t-value = 7.6) relative to the SMS-EPI scan (t-value = 6.9). k-t rFOV-EVI also improved the temporal SNR by about 42% (temporal SNR = 52.8/37.2 for k-t rFOV-EVI/SMS-EPI).

Discussion and Conclusions

In this study, we demonstrated a k-t rFOV-EVI pulse sequence for acquiring fMRI data in a restricted field-of-view at 2.5-mm spatial resolution and an ultrashort volume TR (240 ms). The use of rFOV reduces the ETL, hence mitigating the image quality problems while increasing acquisition efficiency. The scan efficiency is further improved by the use of (k, t)-space undersampling and PS-Sparse reconstruction. The fMRI experiments have shown that this new sequence provides higher image quality than SMS-EPI and considerably improves the detection sensitivity of brain activations. These results indicate that k-t rFOV-EVI can be a strong contender for fast fMRI studies over a focused brain area. In the present study, the acquisition of each k_x-k_y plane was completed in a single TR. If a higher spatial resolution (< 2.5 mm) is required for an fMRI study, a k_x-k_y plane can be acquired using interleaved/multi-segment EPI modules to increase the number of phase encodings in the k_y-direction³. We will investigate this case in future studies.