The development of new hyperpolarization (HP) ¹³C methods has allowed unprecedented characterization of metabolism and perfusion changes in various diseases processes using MRI^1,2. To characterize these temporal processes, dynamic hyperpolarized ¹³C imaging is desirable, although a trade-off between high temporal resolution and high spatial resolution is typically required³. The bSSFP sequence offers the highest SNR per unit time and can be readily combined with compressed sensing to achieve high spatial and temporal resolution dynamic imaging⁴. A low rank plus sparse (L+S) reconstruction approach for undersampled data has been previously demonstrated in ¹H perfusion imaging⁵, whereby the low rank component describes the temporally and spatially correlated background, and the sparse component describes the remaining dynamic signal changes, both of which are typically seen in hyperpolarized ¹³C imaging. In this project, we utilized the L+S reconstruction to achieve 3D high resolution (1.5-2 mm isotropic; 0.003-0.008 cm³) ¹³C images of [¹³C,¹⁵N₂]urea and [2-¹³C]pyruvate in vivo with 5 s temporal resolution.Simulated reconstructions using the L+S method (total variation along the time dimension for the sparse component) were performed by retrospectively undersampling a previously acquired 3D high resolution dataset that was transformed into a 4D dataset by adding signal dynamics based on previously published in vivo hyperpolarized ¹³C studies². A different undersampling variable-density pattern was used for each time point and structural similarity index (SSIM) was used to assess the reconstruction. The in vivo experiments were performed with a custom dynamic 3D bSSFP pulse sequence in normal Sprague-Dawley rats. For each of the 13 time points for [¹³C,¹⁵N₂]urea imaging, ~192/480 phase encodes (80 x 40 x 12 matrix size) were acquired based on a simulated 60% undersampling pattern chosen for initial studies. A 1.6 ms sinc pulse (no slice select gradients) with a flip angle θ = 30°, after a θ = 15° preparatory pulse, was used with a TR/TE of 7.5/3.75 ms, leading to a scan time of 1.44 s per time point. A delay of 3.56 s was used between time points, leading to a temporal resolution of 5 s, and a total scan time of 65 s. [2-¹³C]pyruvate imaging was acquired with similar pulse and TR/TE parameters, but ~108/270 phase encodes (60 x 30 x 9 matrix size) were acquired, leading to a scan time 810 ms per time point, and a 4.19 s delay to achieve 5 s temporal resolution. Both compounds were imaged with a 12 x 6 x 1.8 cm³ field of view, leading to 1.5 mm isotropic resolution for [¹³C,¹⁵N₂]urea and 2 mm isotropic for [2-¹³C]pyruvate. The ¹³C images of [2-¹³C]pyruvate and [¹³C,¹⁵N₂]urea were each acquired independently starting at the beginning of injection (t = 0 s). The experiments were conducted on a 3T GE MR scanner. DNP experiments used a HyperSense polarizer, and 3 mL of 110 mM [¹³C,¹⁵N₂]urea and 80 mM [2-¹³C]pyruvate was injected via tail vein catheters in two different animals.The L+S reconstruction of retrospectively simulated data was in good agreement with the fully sampled data for various acceleration factors, with the resulting SSIMs being above 0.9. Figure 1 shows the L+S reconstructed and fully sampled images, as well as the SSIM map for a representative dynamic slice, for the 60% undersampling pattern used in in vivo studies. The resulting 3D dynamic images showed distribution of urea (Figure 2) and pyruvate (Figure 3) within the kidneys, aorta, and heart, with the resolution being high enough to visualize uptake in the renal cortex, medulla, and pelvis. The figures show the full 3D view of each compound at 20 s after start of injection, as well as the time course for a representative slice of both urea and pyruvate, with the SNR being high enough to detect both compounds in multiple time frames. The signal dynamics for each compound from the left kidney are also shown. While the RF pulse was not spectrally selective, any signal from metabolites formed from [2-¹³C]pyruvate can be considered relatively negligible compared to the signal level of [2-¹³C]pyruvate⁶. The application of the L+S reconstruction to the bSSFP sequence allowed for 3D dynamic high resolution (0.003-0.008 cm³) imaging of HP ¹³C substrates. This approach also provided high temporal resolution and a large temporal window that showed in-flow and out-flow of the substrates in different anatomical structures. The 1.5-2 mm isotropic resolution achieved here is similar to that used for ¹H imaging and application of this approach can be extended to future studies of cancer and other disease models, as well as ultimately for clinical imaging.