Login / Signup

Synthesis of higher-B 0 CEST Z-spectra from lower-B 0 data via deep learning and singular value decomposition.

Mengdi YanChongxue BieWentao JiaChuyu LiuXiaowei HeXiaolei Song
Published in: NMR in biomedicine (2024)
Chemical exchange saturation transfer (CEST) MRI at 3 T suffers from low specificity due to overlapping CEST effects from multiple metabolites, while higher field strengths (B 0 ) allow for better separation of Z-spectral "peaks," aiding signal interpretation and quantification. However, data acquisition at higher B 0 is restricted by equipment access, field inhomogeneity and safety issues. Herein, we aim to synthesize higher-B 0 Z-spectra from readily available data acquired with 3 T clinical scanners using a deep learning framework. Trained with simulation data using models based on Bloch-McConnell equations, this framework comprised two deep neural networks (DNNs) and a singular value decomposition (SVD) module. The first DNN identified B 0 shifts in Z-spectra and aligned them to correct frequencies. After B 0 correction, the lower-B 0 Z-spectra were streamlined to the second DNN, casting into the key feature representations of higher-B 0 Z-spectra, obtained through SVD truncation. Finally, the complete higher-B 0 Z-spectra were recovered from inverse SVD, given the low-rank property of Z-spectra. This study constructed and validated two models, a phosphocreatine (PCr) model and a pseudo-in-vivo one. Each experimental dataset, including PCr phantoms, egg white phantoms, and in vivo rat brains, was sequentially acquired on a 3 T human and a 9.4 T animal scanner. Results demonstrated that the synthetic 9.4 T Z-spectra were almost identical to the experimental ground truth, showing low RMSE (0.11% ± 0.0013% for seven PCr tubes, 1.8% ± 0.2% for three egg white tubes, and 0.79% ± 0.54% for three rat slices) and high R 2 (>0.99). The synthesized amide and NOE contrast maps, calculated using the Lorentzian difference, were also well matched with the experiments. Additionally, the synthesis model exhibited robustness to B 0 inhomogeneities, noise, and other acquisition imperfections. In conclusion, the proposed framework enables synthesis of higher-B 0 Z-spectra from lower-B 0 ones, which may facilitate CEST MRI quantification and applications.
Keyphrases
  • density functional theory
  • deep learning
  • electronic health record
  • big data
  • machine learning
  • artificial intelligence
  • contrast enhanced
  • computed tomography
  • molecular dynamics
  • data analysis
  • virtual reality