Intrinsic brain networks seen in humans, including the default-mode network (DMN), have been demonstrated in non-human primates and rodents [1, 2] using resting-state functional fMRI (rs-fMRI). Characteristics of these brain networks, such as frequency specificity, have been assessed in humans [3], but are much less known in animal models. These characteristics are of importance when translating findings from preclinical models to clinical applications. The frequency range used in a human rs-fMRI analysis is typically ≤ 0.1 Hz [4]; however, an appropriate frequency range in rodents remains unclear. In this study, we investigated the resting-state functional connectivity (rsFC) of rat brains in three frequency ranges: 1) 0.01 – 0.1 Hz, 2) 0.1 – 0.25 Hz, and 3) 0.25 – 0.5 Hz, and compared the result with that in human brains.

Human study: Thirty-four health human subjects from Human Connectome Project (HCP) were selected (16 males; age range: 22-35 years; mean±SD of age: 29 ± 3.4 years). All the MRI information of functional (TR = 720 ms, using a multiband factor of 8, FA = 52°—reduced from 90° to match the Ernst angle, maximizing SNR; time points = 450) and structural (3D MPRAGE) data could be found in [5].

Rat study: Thirty-four male Spragur-Dawley rats (275±25 grams) were used in this study. Animals were anesthetized with a combination of 2-0.5% isoflurane and 0.015 mg/kg/hr dexmedetomidine hydrochloride. All MRI data were acquired using a Bruker Biospin 9.4T scanner. MRI scans contained the high-resolution anatomical images and two rs-fMRI (TE = 15 ms; TR = 1,000 ms; FOV = 30 × 30 mm²; matrix size = 64×64; time points = 520).

Image processing: Motion correlation, spatial smoothing, and detrend were used in preprocessing. The ICA analysis by MELODIC were applied in human and rat rs-scans. The component of DMN and sensorimotor network in human and rat images were manually selected. After the processing procedures, three band-pass filters were applied in rat and human rs-fMRI data: 1) 0.01 – 0.1 Hz, 2) 0.1 – 0.25 Hz, and 3) 0.25 – 0.5 Hz. The rsFC map of two selected networks was reconstructed by dual regression approaching after the three band-pass filtering.

For comparison, we conducted the spatial similarity to demonstrate the similarity between the rsFC maps, and the similarity was estimated by the correlation of the rsFC strength across voxels.

The rsFC signal in rats attenuated more slowly than that in humans as frequency increases (Fig.1 left). The rsFC signal in the frequency range of 0.01 – 0.1 Hz consists of approximately 60% spectral power in humans, but only about 40% in rats (Fig.1 right). To reach the same power composition as in humans, a wider frequency range (approximately 0.01 – 0.2 Hz) needs to be selected in rat rs-fMRI data. Fig.2 shows the DMN and sensorimotor network after the rs-fMRI signals were decomposed into three frequency bands and reconstructed by the dual regression. The spatial similarity of each network across the three frequency bands are shown in Fig. 3. Significantly higher similarity was found in the rat networks than in human networks.In this study, we demonstrated that rs-fMRI signals in rat default-mode and sensorimotor networks extend far beyond 0.1 Hz, which is typically used as a frequency cut for human rs-fMRI data. The fractional power in the frequency range of 0.01 - 0.1 Hz in humans is about 1.5 times higher than that in rats. These results suggest that the frequency range used in rats should be wider than that traditionally used in humans. However, Keilholz el al. mentioned that the spectral distribution of resting state MRI signal in rats appears to be anesthetics-dependent [6]. It’s still unclear whether the wider frequency spectrum is caused by anesthetics. Although the underlying neural mechanism remains unclear, our findings provide a practical guidance for the analysis of rs-fMRI data of rats.