All methods to image hyperpolarized agents in vivo are subject to potential bias, artifact and blurring due to the time-dependence of magnetization as RF pulses are applied and as polarization is lost through spin-lattice relaxation. These problems are magnified when the longer pulses or acquisition periods, characteristic of spectroscopic imaging, are used. In this abstract, we discuss a method to detect and correct for artifacts using chemical shift imaging.To compare excitation and analysis methods, a healthy mouse was imaged using four injections of 80mM 300 μL hyperpolarized 1-¹³C pyruvate. Imaging was performed using a Bruker Avance 400 vertical bore imaging spectrometer and a 25mm ID ¹H/¹³C saddle coil (Bruker). The neutral, isotonic agent was prepared using a Hypersense (Oxford Instruments). A cylindrical sodium lactate phantom (OD=3mm, 4M) was placed at one side of the animal bed. The sedated animal (isoflurane) was monitored for respiration and controlled for temperature (37°) using a rectal probe. Acquisitions 1 and 3 utilized chemical shift imaging with a progressive flip angle as per [1], beginning with α=3.58° and increasing to 90° over 256 acquisitions in an effort to maintain constant signal intensity. Acquisitions 2 and 4 used a constant flip angle of α=8° for each excitation. Other parameters were common to all acquisitions: SW=5000 Hz, AT=51.3ms, TR/TE=53/0.47ms, FOV=25x25mm², ST=8mm, axial, covering both kidneys, MS=16x16 for an in-plan resolution of 1.56x1.56mm². Images were separated by approximately 1 hour to allow for sample polarization. k-space was sampled using a custom outward spiral trajectory which was modified such an additional k=0 (in-plane nonselective) spectrum was acquired after every 10 k≠0 encodings. This allowed agent and metabolite time-dependences to be observed independently. Multislice reference proton images with ST=1mm and the same FOV were acquired between ¹³C images. Metabolite maps were reconstructed by integrating individual peaks in each voxel, and interpolated to the 256x256 grid of the proton image. Fig. 1 shows the pyruvate and lactate peak integrals for the 24 k=0 acquisitions spaced throughout the 13.6s of imaging time. The progressive flip-angle spectra (top traces) maintained more consistent peak integrals, although due to in/outflow and ongoing label exchange, significant trends were observed that differed by chemical species. This highlights the fact that no acquisition scheme can account for signal dynamics when multiple species are being studied. Figs. 2 and 3 show the reconstructed CSI spectra reconstructed from constant and progressive flip-angle acquisitions overlaid on a 1mm proton image positioned at the center of the ¹³C slice. Signal-to-noise of the agent in an individual voxel is at most ~30, as compared to ~12 and 8 for the metabolites lactate and alanine. The most robust metabolism was observed in the small intestine, with lesser activity seen in the kidneys and skeletal muscle. Because of the larger signal near k=0, the constant flip-angle images display superior signal-to-noise, but visible blurring due to loss of signal for large k acquisitions. This effect is more clearly observed in figs. 4 and 5, which show pyruvate and lactate maps, respectively. The top row is reconstructed directly from the localized spectra of figs. 2 and 3. In the bottom row, the time-dependence of signal amplitude is eliminated by dividing each acquisition by the metabolite-specific curves of fig. 1 (“de-blurring”). In this way, the underlying structure is clarified (see, e.g., the ring-like metabolite structure of the left kidney, and the highly heterogeneous metabolism in the intestine; both are observed in independent images, and signal-to-noise is sufficient to ensure that neither is artifact. We have introduced a method to observe and correct for signal variability during an extended imaging protocol using hyperpolarized agents. Notably, the observed signal dynamics requires an independent correction for each chemical species, and this correction minimizes blurring due to motion or relaxation-induced dynamics. This technique may be generalized to other sequences; e.g., multi-echo sequences in which the dynamics of T₂ relaxation must be accounted for or understood. It may be that investigators will choose to tolerate some blurring to increase signal near k=0, but this method of periodic resampling will nonetheless allow an understanding of blurring artifacts and aid in image interpretation.