Compressed sensing effects on quantitative analysis of undersampled human brain sodium MRI

Area of Science:

• Medical Imaging
• Biophysics

Background:

• Clinical sodium MRI faces challenges with low image quality and long scan times.
• Compressed sensing (CS) can reduce acquisition time but may introduce quantitative bias in low SNR x-Nuclei imaging.

Purpose of the Study:

• To analyze CS for quantitative human brain sodium MRI using undersampled data.
• To provide recommendations for accurate tissue sodium concentration (TSC) estimation.

Main Methods:

• Investigated CS reconstructions from 3D radial acquisitions in 5 healthy volunteers.
• Evaluated varying undersampling factors (USFs) and CS penalty weights across Wavelet, DCT, and Identity sparsity domains.
• Compared reconstructions with highly sampled and undersampled NUFFT-based images for quality and TSC bias.

Main Results:

• Wavelet-based CS reconstructions yielded the highest image quality and stable TSC estimates across most USFs.
• Good structural detail was observed up to an USF of 4.
• DCT and Identity CS showed good image quality but introduced TSC bias, reducing estimates.

Conclusions:

• Wavelet-based CS reconstructions minimize image intensity bias, enabling up to a fourfold speed-up.
• This acceleration, with maintained structural detail, can help integrate sodium MRI into clinical practice.