Author: Muhammed AR Anjum, Andrew J. Fagan 👨🔬
Affiliation: Mayo Clinic 🌍
Purpose:
This study describes a novel post-processing method to boost image SNR for deuterium-MRI acquired using chemical shift imaging (CSI) at 7T. Deuterium MRI via exogenous administration of a deuterated-glucose solution has emerged as a promising technique for imaging tumors in the body. Traditionally, techniques such as AMARES which require prior biochemical knowledge and initial parameter estimations are used to post-process CSI data to produce an image. Here, we describe a novel approach using the sub-band Steiglitz-McBride algorithm (SSMA), which utilizes sub-band decomposition and autoregressive moving-average modeling of CSI data to enhance SNR and estimation accuracy respectively, and does not require biochemical and interactive prior knowledge as required for existing methods.
Methods:
Performance of AMARES and the newly-developed SSMA-based method were compared using the mean RRMSE and Cramer-Rao lower bound (CRLB) computed for a simulated 31P MRS signal specified in the original AMARES study. Phantom experiments were performed at 7T (Terra, Siemens) using an 8Tx/Rx (2H) body array coil with CSI parameters: 12.5mm isotropic voxels, TR/TE=1500/1ms, pulse-duration=2ms, samples=1024, TA=130mins. Scans were repeated at the 10 transmit-voltage levels (32.2-193.5 V) to compute B1+ maps. The data was processed in MATLAB (MathWorks, USA) using OXSA-AMARES toolbox and in-house scripts.
Results:
SSMA outperformed AMARES in terms of accuracy by consistently maintaining a relatively close mean RMSE to the theoretically-optimal CRLB at all noise levels. B1+ maps derived from the magnitude estimates of AMARES had significant artifacts particularly in areas of low SNR, whereas SSMA generated magnitude maps without these artifacts, and provided more consistent, reproducible and bias-free B1+ distributions without the need for ‘good’ initial parameter estimations or human interaction.
Conclusion:
The SSMA approach can provide a significant increase in SNR with superior robustness-accuracy properties for spectroscopic imaging techniques of low-SNR nuclei such as deuterium.