Blood oxygen level dependent (BOLD)-functional magnetic resonance imaging (fMRI) is a non-invasively accessible tool to investigate brain function¹. The reliability of fMRI as measured by test/retest reproducibility as well as its ability to detect subtle changes have not been established conclusively. Time-dependent changes may arise from either random or systematic processes. Random processes are non-physiologic originated from motion artifact, instability of the MR scanner, data processing, or neurophysiologic originated from random cognitive processes and/or arousal level. Systematic processes are related to the difference in performance of a specified task such as habituation, sensitization and learning. It is known that responses in the sensory systems show habituation over time, leading to a reduction in functional activation or conversely to an enlargement of activation when learning occurs. Confirming the reproducibility for stimulation-induced BOLD response is important on longitudinal studies when investigating a time course of functional recovery. Therefore, the purpose of this study is to evaluate the reproducibility or time-dependent changes of fMRI activation in somatosensory cortex in rats under isoflurane anesthesia.

Sprague-Dawley rats (weight, 350±20g, n=6) were initially anesthetized using 2-5% isoflurane, then maintained by using an artificial ventilation system that supplied 1.4% isoflurane in a mixture of O₂ and air gases for MRI experiments. The MRI data were obtained at baseline, 1 day, 1 week, 2 weeks, 3 weeks and 6 weeks with a 7T MR System (Bruker-Biospin, Fallanden, Switzerland). We derived electrical stimulus pulses of 12 Hz into both forepaws alternatively. Each stimulus run consisted of a 20 sec pre-stimulus, 20 sec stimulus, and 40 sec post-stimulus period. BOLD-fMRI was performed using a single-shot gradient-echo EPI sequence (TR/TE=1000/60ms, resolution=469×469µm², slice thickness=1.5mm, number of repetition=80). A minimum of 10 runs were performed for each forepaw with the inter-run period longer than 3 min. The activation map was generated from a voxel-wise cross-correlation between the signal time course and boxcar reference convolved with a canonical hemodynamic response function. The correlation coefficients (r) of the activation maps greater than 0.5 were identified as BOLD activation. For detailed analysis, the activated voxel number was determined in each subject and then coefficient of variation (CV) between different time sessions was calculated. Reproducibility (i.e., overlap) maps showing the number of times that a given voxel was classified as activated were generated. A reproducibility index (RI) was estimated to compare the test-retest reproducibility for different sessions and different subjects, according to the following equation: $$RI = \frac{1}{(N-1)}\left(\frac{\sum_{i=1}^Nini}{\sum_{i=1}^Nni}-1\right)$$

Where N is the total number of runs of a task, ni is the number of voxels that were classified as activated during i runs.

BOLD responses were observed in the sensorimotor cortex contralateral to each stimulated forepaw. The CV of the numbers of activated voxels was 11.62±4.02% (range, 4.34~17.07%). There were no definite time-dependent changes in terms of the activated voxel number. The reproducibility index was 0.58±0.08 (range, 0.41~0.68, Fig.1). The value of 0 indicates that no voxels were activated during all runs, whereas a value of 1 indicates that any voxels were activated during all runs. Fig. 2 shows the averaged BOLD percent signal changes in primary somatosensory cortex at baseline to 6 weeks. The BOLD percent signal change shows decreasing tendency from baseline to 3 weeks, however at 6 weeks, the signal change increased even higher than the signal change at baseline.Our study shows that BOLD patterns are qualitatively and quantitatively robust for rat under isoflurane anesthesia. The CV was about 11% for activated voxel number, so if the following study did not show voxel number change less than 11%, and it may not have true meaning. However, our experiments which have CV of 11% and RI of 0.6 are robust and comparable to the other previous study using other anesthetics^{2, 3}. In addition, we observed decreasing tendency of signal change from baseline to 3 weeks, however at 6 weeks, the signal change was accentuated, even greater than the signal change at baseline. From baseline to 3 weeks, the phenomenon may be called habituation. A previous human study showed no time effect and no habituation within one session but test-retest effect between session separated by 5 hours or by 1 or 2 months¹. Similarly, our studies shows habituation may prolonged to about 1 weeks in rat. However, the 3 weeks interval may delete the habituation effect and even exaggerate the response. The phenomenon possibly means sensitization, a type of non-associative learning for noxious stimulus such as pain or angry face.