Solid tumors have increased glucose uptake as compared to normal tissues, and preferentially metabolize glucose to lactate by anaerobic glycolysis. This phenomenon, known as the Warburg effect, is a less efficient pathway for ATP production and leads to elevated lactate levels in solid tumors^1-2. Thus, lactate accumulation is a key indicator of tumor hypoxia and altered metabolism. Magnetic resonance spectroscopic imaging (MRSI) is a useful technique for imaging lactate metabolism in vivo, introducing the possibility of employing non-invasive lactate imaging as a powerful prognostic marker in the clinic. However, the long scan time associated with MRSI is a deterrent to its inclusion in current clinical protocols due to associated costs and patient discomfort. Acceleration strategies like compressed sensing (CS) permit faithful reconstructions even when the k-space is undersampled well below the established Nyquist limit^3-4, and could potentially reduce the scan time in MRSI⁵. The objective of this study was to speed-up the acquisition of spectrally-edited MRSI using compressed sensing, for rapid imaging of the lactate resonance. CS-accelerated lactate maps were also acquired for tumor mice treated with the prodrug combretastatin A4 phosphate (CA4P)⁶, which rapidly disrupts and shuts down the tumor vasculature, to assess the changes in lactate metabolism in response to treatment.MRSI experiments were conducted on a Bruker BioSpec 7T preclinical MRI scanner. A MRSI sequence containing spectral editing components for the selective excitation of lactate⁷ was developed in the ParaVision 5.1 environment. A built-in sampling mask enabled the pseudo-random undersampling of the k-space ‘on the fly’, facilitating CS acquisitions. The developed lactate-CS-MRSI sequence was tested on phantoms and in vivo in mice subcutaneously implanted with H1975 tumors in the right thigh. Baseline lactate CS-MRSI datasets were acquired (1X-5X) prior to administering CA4P (83 mg/kg body weight), and 24 hours following the injection of the prodrug. MRSI acquisition parameters – 16x16x2048 grid, TE/TR = 144/1500 ms, 4 mm slice thickness, 4 averages, FOV 3x3 cm², total scan time for the 1X dataset = 25 min 36 s. Undersampled MRSI datasets were reconstructed offline in Matlab using an in-house reconstruction algorithm. The reconstruction was cast as a convex optimization problem, which involved minimizing the following cost function: ∥F_um−y∥₂ + λ_L1∥Wm∥₁ + λ_TVTV(m). Minimal post processing was applied to the reconstructed datasets in jMRUI⁸, namely apodization, phase correction, and removal of residual water. The generated lactate maps for various acceleration factors, both pre- and post-CA4P administration, were quantitatively compared with the reference 1X dataset on a voxel-by-voxel basis using metrics like the peak amplitude and SNR to assess the fidelity of the CS reconstructions.Lactate was readily detected in the phantom and in tumor mice, with effective suppression of the water, fat, and other resonances by the lactate-CS-MRSI sequence. Figure 1(a) shows the MRSI pulse sequence incorporating lactate editing with pseudo-random undersampling of the phase encodes. Figure 1(b) depicts lactate maps in a tumor mouse for prospectively acquired CS-MRSI datasets 2X-5X. There were no statistically significant differences between the undersampled and reference reconstructions, except for the 4X case (p<0.05). Maps showing lactate metabolism in response to CA4P treatment are illustrated in Figure 2. A decrease in lactate levels was observed 24 hours after administration of CA4P, as measured by the lactate-CS-MRSI sequence (1X pre/post = 294.58/206.18 a.u. and 5X pre/post = 311.25/200.91 a.u., mean integrated intensity). Figure 3 depicts lactate maps from the administration of dextrose to a H1975 tumor mouse, which served as the control. As expected, no significant difference was found in the total lactate level over the tumor volume (1X pre/post = 278.39/263.07 a.u. and 5X pre/post =269.37/265.49 a.u.) 24 hours after dextrose injection. In both cases, CS reconstructions demonstrated high fidelity with the 1X dataset, both pre- and post CA4P therapy/dextrose administration. The combination of spectral editing and CS undersampling enabled rapid mapping of the changes in lactate metabolism in response to therapy and otherwise. The undersampled reconstructions maintained fidelity with the 1X dataset. Clinical implementation of the lactate-CS-MRSI sequence is in progress, to facilitate fast non-invasive lactate imaging in cancer patients.