Improved quantification in CEST-MRI by joint spatial total generalized variation.

Area of Science:

• Medical Imaging
• Biophysics
• Spectroscopy

Background:

• Chemical Exchange Saturation Transfer (CEST) is a sensitive MRI technique for detecting low-concentration metabolites.
• Accurate spectral fitting of CEST data is crucial for reliable quantification.
• Existing fitting methods can be unstable and sensitive to noise, limiting diagnostic confidence.

Purpose of the Study:

• To investigate the use of joint Total Generalized Variation (TGV) regularization for improving Multipool-Lorentzian fitting of CEST spectra.
• To enhance the stability and parameter signal-to-noise ratio (SNR) of CEST spectral fitting.

Main Methods:

• Integrated joint TGV regularization into the nonlinear parameter fitting problem.
• Employed preconditioning via voxel-wise singular value decomposition for improved convergence.
• Solved the problem using an iteratively regularized Gauss-Newton method with a Primal-Dual splitting algorithm.
• Evaluated the TGV method on simulated phantoms, 3T phantom data, and 7T in vivo data, comparing it against standard nonlinear fitting, nonlocal-means filtering, and pyramid schemes.

Main Results:

• The proposed TGV regularization method demonstrated significantly improved robustness compared to reference methods.
• TGV fitting outperformed other methods across a range of SNR values, providing accurate results even with high noise levels.
• Parameter values obtained were closer to or comparable with ground truth.
• For in vivo datasets, TGV regularization increased parameter map SNR and prevented instabilities.

Conclusions:

• The TGV regularization fitting method yields improved results for various datasets and noise levels.
• This method is applicable to all Z-spectrum data with different pool configurations.
• Enhanced SNR and stability can increase diagnostic confidence in CEST imaging.