CEST Mapping from Undersampled Z-spectra in the Brain Using Deep Learning

Chemical exchange saturation transfer (CEST) imaging uses radio frequency pulses at different frequency offsets to generate CEST maps. In this work, we used deep learning to calculate CEST maps from steady-state CEST (ss-CEST) images at undersampled frequency offsets, reducing the total scan time by a factor of 3.5. The Z-spectrum was undersampled by selecting the top 15 frequency offsets from Fisher information gain analysis. Fitting results from the proposed method were compared with those from multi-pool fitting with fully sampled Z-spectrum. We showed that it is feasible to reconstruct CEST maps from undersampled, field uncorrected ss-CEST images.