Functional connectivity (FC) is typically assessed using blood-oxygen-level-dependent (BOLD) signal contrast (1,2). BOLD signal contrast dependents on many factors including cerebral metabolic rate for oxygen (CMRO2) and blood flow (3). Literature tends to show that multi-echo echo-planar-imaging (MEPI) may be useful to decouple CMRO2 and flow changes (4-6). In this work we estimate transverse relaxation rate (R2*) using MEPI and compare whole brain FC estimates from R2* resting-sate fMRI (rs-fMRI) with BOLD rs-fMRI.Imaging: Thirty-two healthy subjects were imaged using three-echo MEPI on GE 3T Discovery MR750 scanner using a 32 channel brain coil under institutional guidelines. The MEPI acquisition had the following parameters: repetition-time (TR) 2.56s, first echo-time of 12ms and echo-spacing of 16.9ms, ASSET factor 2.0, flip-angle 90^o, 36 slices per TR, image matrix of 64×64, field-of-view 220mm, slice thickness 3mm, gap of 0.4mm between slices, and at total of 184 time-points for rs-fMRI. A 1mm isotropic resolution T1-weighted MRI was also obtained. Pre-processing: R2* and M (initial magnetization approximated by BOLD signal at TE=0) were estimated from the natural logarithm of multi-echo data using least-squares estimation. The preprocessing of BOLD, M and R2* time-courses included motion correction, registration to MNI atlas, physiological nuisance removal including global signal regression, spatial smoothing using a 7 mm FWHM Gaussian filter and temporal band-pass filtering (0.01 to 0.1 Hz). Seed-based FC Map: Thirteen 6-mm radius spherical regions were drawn around seed-points associated with functional networks obtained from previously published work (7) and manually placed in specific Brodmann areas. The mean time-course for each seed-region was computed. Then, for each voxel in the brain, the Pearson correlation-coefficient between the voxel and the seed time-course was computed to create the FC map. Correlation-coefficient was converted to z-score using Fisher transform. Parcel-based FC Graph: FreeSurfer brain regions (8) were used to label 86 structural parcels in the T1-weighted image. The correlation-coefficient between two parcel mean rs-fMRI time-courses was computed and converted to z-scores as the FC measure for the parcel pair. Z-scores were computed between every pair of 86 FreeSurfer parcel time-courses producing an 86×86 FC graph (matrix). Similarly, a 90×90 FC matrix was computed using 90 functional parcels (9). Once, the FC matrices were computed for each subject in the cohort, the group averages were computed to generate the average connectivity matrix for the cohort. Jaccard Similarity: Jaccard similarity between the connectivity matrices from BOLD, R2* and M was computed. Similarity was measured separately for positive (>=0.2), negative (<=-0.2) and near zero (>-0.2 and <0.2) FC z-score values. Jaccard similarity for two matrices A and B was computed as n(A && B)/n(A || B), where n() denotes the cardinality.Typical functional networks were observed in Echo 2 (TE » 30ms, which is the conventional BOLD echo time) and R2* based rs-fMRI. Figures 1 shows functional connectivity measured using BOLD signal at echo 1 and echo 2, M and R2* in a representative subject for the default mode network (DMN). Figure 2 shows the same comparison for the primary visual network. Echo 2 and R2* derived FC maps though similar were not identical (shown by blue arrows in Figure 1). Figure 3 compares the group average FC matrices computed using echo1 and echo 2 with those from M and R2* using 86 FreeSurfer structural parcels. Figure 4 shows the same comparison with 90 functional parcels. Table 1 (Figure 5) shows the Jaccard similarity between the FC matrices from BOLD, R2* and M. Contrast in the BOLD signal at shorter echo-times is expected to be primarily driven by changes in flow (M). The echo 2 FC matrix has combination of flow and CMRO2 effects as shown by the Jaccard similarity with M and R2*. Results tend to show that M might be helpful to decouple flow effects. Connectivity matrix computed using structural parcels showed positive connectivity for parcels with indices less than 19 (sub cortical regions) on BOLD, M and R2*. However, the structural parcels of the cortical regions were weakly connected on R2*. This could be because of the inhomogeneity of the R2* time-courses in the cortical structural parcels. The connectivity matrix for R2* using functional parcels, however, showed significant within network connectivity (clusters along the diagonal) for cortical and sub-cortical parcels. Positive between network connectivity also survives in R2* derived matrices, however, the anti-correlations observed between networks in echo 1, echo 2 and M were significantly lesser in R2* derived matrices. R2* rs-fMRI may be useful in decoupling flow and CMRO2 effects and improve the understanding of network connectivity.