We recently reported the detection of resting state correlations within gray matter of healthy human spinal cords based on blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) at 7 Tesla [1]. Our observations of significant correlations between ventral horns and between dorsal horns revealed functional connectivity within spinal cord motor and sensory networks in a resting state. We subsequently have investigated whether such resting state correlations can be reliably detected in individual subjects at 3 Tesla with the goal of translating functional connectivity studies for practical clinical applications. We hypothesize that while BOLD contrast is decreased at lower field, the improved homogeneity of B₀ and B₁ fields, more robust coils, and the smaller impact of physiological noise, may permit measures of spinal cord connectivity to be practically achievable in a clinically relevant manner. The detection of these networks via temporal cross-correlations is influenced by factors such as degrees of freedom and range of frequencies examined, so this study evaluated two different 20-min acquisition sequences with three bandpass filter ranges to develop an optimized acquisition and processing protocol for clinical resting state spinal cord studies.Experiments were performed on a Philips Achieva 3T scanner with a dual-channel transmit body coil and a 16-channel neurovascular coil for signal reception covering the brain and cervical spinal cord. Ten healthy volunteers (5 male) with no history of spinal cord injury were scanned under a protocol approved by the institutional review board. Two consecutive 20-min resting state runs were performed with a 3D gradient-echo sequence: field of view = 150×150 mm, slice thickness = 5 mm, voxel size = 1×1×5 mm³, 12 slices, repetition time = 40 ms, echo time = 8.0 ms, flip angle = 8°, k-space lines per radiofrequency pulse = 7, and sensitivity encoding [2] = 2.0 (left-right). To investigate the trade-off between temporal signal-to-noise ratio (TSNR) and degrees of freedom in detecting correlated BOLD signal fluctuations, one fMRI sequence (“full-k”) did not employ a partial Fourier scheme and had an acquisition time of 2.77 sec (434 volumes) whereas the other sequence (“part-k”) used a partial Fourier acquisition that resulted in a reduced acquisition time of 2.08 sec (580 volumes). Functional data were preprocessed using a refined version of the procedure in [1]. Gray matter masks were subdivided into quadrants to identify left and right ventral and dorsal horns and morphologically eroded to remove outermost voxels and mitigate partial volume effects. Ventral horn connectivity is the partial correlation between time series within left and right ventral masks while controlling for the averaged time series in dorsal masks, and vice versa for dorsal horn connectivity.Figure 1 presents data from a healthy volunteer. Axial slices were planned perpendicular to the cord to obtain coverage of vertebrae C2 to C5 (Fig. 1A). High-resolution (0.65×0.65 mm²) averaged multi-echo gradient echo (mFFE) [3] T₂*-weighted axial images (Fig. 1B) clearly show the characteristic gray matter butterfly, and a similar pattern is observed in T₂*-weighted fMRI data (Fig. 1C). Across all subjects, median TSNR in gray matter is 22.0 for full-k and 19.6 for part-k. Figure 2 displays mean within-slice ventral and dorsal horn connectivity across subjects for both acquisition sequences and three bandpass filtering ranges (0.01-0.08 Hz, 0.01-0.13 Hz, and 0.01-0.17 Hz) for truncated time series between 2 and 20 mins.We acquired 20-min runs to investigate temporal correlations over what may be considered the longest feasible time to acquire a resting state run during a clinical examination. The results (Fig. 2) suggest that the detectability of functional connectivity (z-score) increases as run length increases, although there are diminishing returns after approximately 6 mins. Dorsal horn connectivity is consistently lower than ventral horn connectivity, which is consistent with previous findings at 7T [1]. Interestingly, an important factor in increasing sensitivity for detecting connectivity is the inclusion of frequencies above 0.08 Hz. At 6 mins, which is acceptable for a clinical examination, high z-scores are achieved with full-k and filtering up to 0.13 Hz or 0.17 Hz, or part-k and filtering up to 0.17 Hz. Higher frequencies appear to reliably increase detection of bilateral dorsal horn connectivity at 3T. Future work will investigate dynamic functional connectivity and use these methods to characterize differences in resting state spinal cord functional connectivity between healthy controls and cohorts of patients with diseases of the central nervous system.