Liver dynamic contrast enhancement MRI (DCE-MRI) is a promising non-invasive imaging technique that can be used to qualitatively and quantitatively assess hepatocellular carcinoma (HHC) and other liver diseases [1]. Accurate liver perfusion quantification has been challenging largely due to the respiratory motion. A 3D radial “stack-of-stars” (SOS) trajectory has great potential to be used in liver DCE-MRI because of its inherent motion robustness, and furthermore, self-gating during free breathing can be accomplished due to the acquisition of the center of k-space every TR [2-4]. However, the self-gated signal (SGS) from the center of k-space includes both respiratory motion and contrast uptake in DCE-MRI, and it is not trivial to perfectly separate two from SGS [5], leading to inaccuracies of respiratory motion. In this work, we propose a method to extract the respiratory motion only from SGS using golden angle radial acquisition with two-point Dixon separation. The proposed method utilizes the fact the fat-only SGS does not include contrast uptake while including the same respiratory motion. Two echoes (S₁: in-phase and S₂: out-of-phase) were acquired using a 3D Golden Angle (GA) Radial VIBE Dixon WIP sequence on a 3T Siemens Skyra scanner (Siemens Healthcare, Erlangen, Germany). The proposed fat-only SGS method consists of the following steps; 1) the fat-only signal (S_F) was computed as S_F = 0.5×|S₁ – S₂| [6]. 2) The SGS was obtained by taking the FFT of the center (kx = ky = 0) kz-axis of the data and arranged in a time series [5] (Fig. 1), and the respiratory motion was extracted using a weighted sum, Z intensity-weighted position (ZIP) [4]. The coil closest to the diaphragm was chosen to determine the respiratory motion. 3) A low-pass hamming filter was applied to smooth out the noise. The end-expiration state was reconstructed; by binning the data into that motion state and then using an NUFFT. To evaluate the effectiveness of the proposed fat-only SGS, four in-vivo MRI studies were performed during free-breathing without contrast injection. The scans had the following parameters: TR = 3.85ms, TE1 = 1.23ms, TE2 = 2.46ms, FOV 380 x 380 x 144 mm³, 256 x 256 x 48 matrix size, 860 radial spokes, flip angle = 7°, and TA ≈ 2:40min. Then we compared the fat-only (S_F) SGS with the conventional SGS from S₁. S₁ is assumed to be the standard signal for SGS and to include both respiratory motion and contrast uptake in DCE-MRI. To further validate the respiratory signal, each subject was instructed to breath normally for the first 90 seconds, then to breath-in deeply and exhale all the way, then to breath more rapidly for the rest of the scan. Figure 2 compares the fat-only respiratory signal with S₁ respiratory signal. The two signals are comparable and have maximums and minimums at similar times. In figure 2b the fat signal follows the instructed breathing patterns that were given to the subjects. Figure 3 shows the end-expiration reconstructed image of S₁ using the binning from S_F and S₁ respiratory motion and no motion correction. There are more streaking artifacts in the S_F and S₁ binned images compared to no motion compensation due to the undersampling. The streaking artifacts can be removed using a more complex reconstruction, such as compressed sensing. Despite the streaking artifacts, there is more blurring in the image with no motion compensation. The images reconstructed by using S_F and S₁ SGS’s are comparable in resolution, which is to be expected.The study shows that the changes in the respiratory signal could be clearly seen and were at proper timing on both conventional and fat-only self-gating signals. The proposed method can be advantageous in liver DCE-MRI due to its inherent separation between respiratory motion and contrast uptake in fat. If there were contrast injected then it would be expected that the S₁ SGS in figure 2a would have a sharp change in signal during contrast uptake, and the S_F SGS would remain the same. The limitations of this study are that there was no direct comparison of the respiratory motion with ground truth, and no validation scan using DCE-MRI. Future studies will include DCE-MRI scans and accurate monitoring of respiratory motion with a bellows system.We have demonstrated that the respiratory motion can be extracted using the fat-only self-gating signal acquired from a two-point 3D GA Radial VIBE Dixon, which has implications of a more robust motion correction for liver DCE-MRI.