Performance Comparison of Compressed Sensing Algorithms for Accelerating T_1ρ Mapping of Human Brain

Abstract

Background 3D‐T1ρ mapping is useful to quantify various neurologic disorders, but data are currently time‐consuming to acquire. Purpose To compare the performance of five compressed sensing (CS) algorithms—spatiotemporal finite differences (STFD), exponential dictionary (EXP), 3D‐wavelet transform (WAV), low‐rank (LOW) and low‐rank plus sparse model with spatial finite differences (L + S SFD)—for 3D‐T1ρ mapping of the human brain with acceleration factors (AFs) of 2, 5, and 10. Study Type Retrospective. Subjects Eight healthy volunteers underwent T1ρ imaging of the whole brain. Field Strength/Sequence The sequence was fully sampled 3D Cartesian ultrafast gradient echo sequence with a customized T1ρ preparation module on a clinical 3T scanner. Assessment The fully sampled data was undersampled by factors of 2, 5, and 10 and reconstructed with the five CS algorithms. Image reconstruction quality was evaluated and compared to the SENSE reconstruction of the fully sampled data (reference) and T1ρ estimation errors were assessed as a function of AF. Statistical Tests Normalized root mean squared errors (nRMSE) and median normalized absolute deviation (MNAD) errors were calculated to compare image reconstruction errors and T1ρ estimation errors, respectively. Linear regression plots, Bland–Altman plots, and Pearson correlation coefficients (CC) are shown. Results For image reconstruction quality, at AF = 2, EXP transforms had the lowest mRMSE (1.56%). At higher AF values, STFD performed better, with the smallest errors (3.16% at AF = 5, 4.32% at AF = 10). For whole‐brain quantitative T1ρ mapping, at AF = 2, EXP performed best (MNAD error = 1.62%). At higher AF values (AF = 5, 10), the STFD technique had the least errors (2.96% at AF = 5, 4.24% at AF = 10) and the smallest variance from the reference T1ρ estimates. Data Conclusion This study demonstrates the use of different CS algorithms that may be useful in reducing the scan time required to perform volumetric T1ρ mapping of the brain. Level of Evidence 2. Technical Efficacy Stage 1.

Regularized SUPER‐CAIPIRINHA: Accelerating 3D variable flip‐angle T1 mapping with accurate and efficient reconstruction

“…The L + S reconstruction led to severe blurring at 16‐fold acceleration. This observation is in consistency with a recent work by Menon et al, 50 where a loss of fine features was also observed with L + S reconstruction for an acceleration rate of 10. The ghosting artifacts associated with L + S may be caused by insufficient suppression of the aliasing artifacts at high acceleration rates 16 .…”

Data‐driven optimization of sampling patterns for MR brain T1ρ mapping

“…The monotone fast iterative shrinkage algorithm with variable acceleration (MFISTA‐VA) 23 is used to minimize x . Although not tested here, other CS techniques such as using wavelets may be tested 24–27 …”

“…Although not tested here, other CS techniques such as using wavelets may be tested. [24][25][26][27]…”

“…The experiments were used to test the prospective data acquisition with VD, PD, VD-optimized sampling patterns (VD-OSP), and PD-OSP. We tested a fully sampled prospective acquisition with 6 TSL durations (2,4,8,15,25,45 the 3D data for the prospective FS, accelerated PD, and VD acquisitions, multiple echoes were acquired following each phase encode, and shots were acquired following each partition encode. The FS prospective data was acquired with 128 shots and 128 echoes for each spin lock duration with a center-out acquisition in the phase and partition encode directions.…”

Systematic review of reconstruction techniques for accelerated quantitative MRI

“…Review studies on qMRI reconstruction techniques are typically specific to a class of techniques such as compressed sensing, 3 MR fingerprinting [4][5][6] or deep-learning. 7,8 Alternatively, they are specific to a particular class of parameters of interest, such as relaxometry, 7,[9][10][11][12] or diffusion 13 parameters, or to a particular application domain, such as the knee, 14 or cardiac and abdominal imaging. 15 To facilitate the comparison among different reconstruction techniques used for accelerated qMRI, the current work presents a systematic review that provides a technical overview of the existing techniques, without confining to any specific parameter or application domain, and categorizes them based on methodology.…”