Conventional contrast-enhanced MRA is generally performed without cardiac gating due to the time constraint imposed by breath-hold and the first passage of the gadolinium based agent and therefore provides limited delineation on blood vessels that are subject to cardiac motion. To address this issue, we recently proposed the MUSIC¹ approach that acquires cardiac-phase-resolved 4D images without breath-holds during the steady-state distribution of ferumoxytol, an intra-vascular contrast agent. Using ECG and airway pressure signal for motion gating, the MUSIC has been routinely applied in our institution on pediatric patients with congenital heart disease (CHD) and provides exquisitely detailed information on both intra- and extra-cardiac vascular anatomy¹. However, the airway pressure signal is only available when general anesthesia (GA) is performed and the ECG signal is not always reliable especially in high field strength due to interference from the RF pulses and gradients. To overcome these limitations, we propose the self-gated MUSIC (sgMUSIC) where a retrospective motion self-gating strategy is used to eliminate the need for external physiological signal and make the MUSIC potentially available on general patient population. In this study, we tested the sgMUSIC on pediatric CHD patients underwent GA so that the derived motion signal can be validated against the recorded physiological signal and the sgMUSIC images compared with original MUSIC. (a)Data Acquisition: A 3D GRE sequence is modified using ROtating Cartesian K-space (ROCK) where kykz views of 3D Cartesian grid were reordered using quasi-spiral pattern with each arm starts from the outer k-space and ends at the center(Fig.1a). The arms are rotated using segmented golden ratio² in which the entire k-space is divided into N=7 angular segments, each segment is sampled using golden ratio and different segments are sampled in a pseudo-random order. The extra degree of randomness ensures more uniform k-space sampling even after data sorting. (b)Data Sorting: Cardiac and respiratory self-gating signal was derived from the repeatedly sampled centerline (kykz=0) using cross-correlation³ and further separated by band-pass filter. The acquired data is then retrospectively soft-gated⁴ based on the derived respiratory self-gating signal and sorted into different cardiac phase based on the derived cardiac self-gating signal. (c)Image Reconstruction: 3D images were reconstructed independently for each cardiac phase using ESPIRiT, a compressed sensing and parallel imaging combined approach⁵. The data sorting and image reconstruction algorithms were implemented in our GPU-accelerated image reconstruction framework⁵ so that 4D images are available at the scanner console 5-8 minutes after the scan. (d)Phantom Experiment: Static phantom experiment was performed to validate the ROCK pattern in different settings: 1) direction of quasi-spiral arms (inward vs. outward); 2) rotating strategy (segmented golden ratio vs. golden angle). Data was sorted using physiological signal recorded from previous patient scans. (e)In-vivo Experiment: 7 clinically indicated pediatric CHD patients (aged 1 month to 8 years) underwent GA with controlled mechanical ventilation were included in this study. 4mg-Fe/kg ferumoxytol was administrated. The sgMUSIC was performed after standard clinical protocol, which includes the original MUSIC with matching resolution. Sequence parameters included: 0.9/2.9ms, FA=25°, 500x300x150, 0.8-1.0mm³ isotropic resolution, 9 phases, TA=5±1.5min. The self-gated images were reconstructed without using ECG or ventilator pressure signal. (f)Image Evaluation: Subjective image quality scores were visually assessed on original MUSIC and sgMUSIC in 3 different anatomical regions by a cardiologist using criteria listed in Table 1.The time series of SG projections from static phantom (Fig.1b) shows that the inward quasi-spiral arm could significantly reduce the eddy-current interference in self-gating signal because an adjacent k-space line is acquired in previous TR instead of a distant one in the case of outward arm. The k-space patterns (Fig.1c) show the segmented golden ratio provides more uniform k-space sampling even after data sorting. Fig.2 shows the in-vivo result from a 3-month-old female where the derived SG signal corresponds well with the recorded physiological signal and the sgMUSIC images using self-gating signal have similar quality with the ones using recorded physiological signal. When compared with the original MUSIC acquired using with 4X GRAPPA, the sgMUSIC acquisition took less time due to the use of variable density sampling pattern and non-linear image reconstruction algorithm. Scores in Table.2 shows that the sgMUSIC maintains or exceeding the image quality of original MUSIC, with significant improvement in 1 of 3 regions. (P<0.05) The volumetric images allows reformatting into arbitrary planes for visualization of detailed vascular structures (Fig.3).The proposed cardiac and respiratory self-gated MUSIC eliminates the need of external physiological signal for motion gating, has slightly increased scan efficiency, while maintaining or exceeding the image quality of the original MUSIC.