fMRI studies in rodents have mainly adopted the same experimental parameters as typically used in human fMRI (i.e., single-shot EPI with a TR of 2s), without addressing putatively physiological and instrumental differences. First, the higher magnetic fields used in rodents lead to stronger susceptibility artefacts, especially in lower parts of the brain. Second, the physiology of the animals needs to be taken into account, as typical heart and respiration rates are much higher, causing physiological noise to alias into the bandwidth of interest [1]. In this work, we have explored various modalities of image acquisition to achieve the best compromise between temporal resolution, geometric distortions, artefacts and SNR. For this purpose, we compared single-shot EPI, segmented EPI and PRESTO, which uses echo shifting to allow rapid acquisition of images that require a long TE and an echo train to reduce distortions [2].

Studies were conducted on Sprague Dawley rats (male, 400-450g) and were carried out on a BioSpec 9.4T MR system (Bruker BioSpinMRI, Germany) equipped with a volume resonator for transmission and a room temperature surface coil for reception. Isoflurane (2-3%) mixed with oxygen enriched air was used for inducing and maintaining the animals under anesthesia. Body temperature was kept constant at 37°C throughout the entire experiment. After acquiring an anatomical reference scan (TR/TE= 3200/11ms), we compared single-shot with two and three shots GRE-EPI and PRESTO (TR=15ms). The following parameters were kept constant between acquisitions: TE_eff = 22ms (optimal BOLD contrast), FOV = 35x35mm, matrix size = 96x96, bandwidth= 300kHz, whereas the slice thickness varied between 1.0, 0.8, 0.6 and 0.4mm; flip angles were set to the Ernst angle for T1= 1900ms [3]. To evaluate the influence of temporal resolution, we compared three different sampling rates (T_vol): 2000, 1000 and 500ms. TRs were adapted according to the formula: TR = T_vol / N_seg, to keep the sampling rate constant for each of the three cases. Noise profiles were acquired by repeating each experiment with FA= 0°. All stability parameters summarized here are explained in [4] and were extracted from an ROI in the left somatosensory cortex. In addition, we ran simulations to evaluate the effect of sampling rate and the number of segments on SNR efficiency by using the following formula [5], and compared it to our experimental data:

$$Eff = \frac{SNR}{\sqrt{T_{vol}}} = \frac{M_{0}}{2a} \cdot \frac{T_2^* \cdot (1-e^{-T_{ACQ}} \cdot e^{1 / ({N_{seg} \cdot T_2^* })})}{\sqrt{T_{ACQ} / N_{seg}}} \cdot \frac{(1-e^{-TR/T_{1}})}{\sqrt{1-e^{-2 \cdot TR / T_{1}}}} \cdot \sqrt{\frac{N_{seg}}{T_{vol}}}$$

PRESTO allows for high temporal resolution and is more robust to susceptibility distortions due to its shorter acquisition time. However, its low efficiency results in insufficient SNR for fMRI studies (Fig.1d). Similarly, segmented EPI uses echo trains to minimize off-resonance distortions to an extent dependent on the number of segments (Fig.1), but without imposing such a radical TR constraint. Off-resonance effects are reduced to their minimum, when the number of segments is equal to the number of k-space lines (i.e., FLASH: the extreme case of segmented EPI, acquiring one k-space line per segment). But, this scheme drastically decreases the temporal resolution and, as shown in figure 2 for constant T_vol, decreases the SNR efficiency (with increasing number of segments). Using two or three segments, on the other hand, maintains the SNR efficiency in an acceptable range (Fig.2), keeps the temporal resolution sufficient for fMRI and improves image quality in terms of in-plane susceptibility distortion (Fig.1). Signal losses due to dephasing also appear through-plane. These losses can be diminished by reducing slice thickness (Fig.3) at the price of SNR. Additionally, the temporal resolution can be further improved by shortening TR, as for a fixed number of segments, decreasing TR does not induce a major penalty in SNR efficiency (Fig.2). In fact, shorter TRs even give the advantage of a more accurate physiological noise sampling, thus avoiding artefacts from respiration aliasing in to the bandwidth of interest. Although, segmented EPI improves general image quality, it comes at the cost of an enlarged number of ghosts and an amplified sensitivity to scanner instabilities, highlighted in figure 4, by decreased SNR_t and increased percent fluctuations.These initial results show the potential advantages of using segmented EPI, with two or three shots, depending on minimum TR and total volume coverage, in the study of functional brain networks in rodents. It allows for rapid acquisition of multiple slices, while reducing susceptibility artefacts and maintaining sufficient SNR. However, an efficient strategy for minimizing ghosting needs to be implemented.