A single‐shot T2 mapping protocol based on echo‐split gradient‐spin‐echo acquisition and parametric multiplexed sensitivity encoding based on projection onto convex sets reconstruction

    loading  Checking for direct PDF access through Ovid

Excerpt

Quantitative T2 mapping has important clinical applications, ranging from diagnosis of multiple sclerosis 1, epilepsy 2, Parkinson's disease 3, to detection of biophysical and biochemical changes due to cartilage diseases 4, to measurement of liver iron concentration in patients with iron‐loading disorders 5.
Although T2 relaxation time can be fitted from multiple spin‐echo imaging data sets acquired at different echo times (TEs), the long scan time of spin‐echo imaging makes it challenging to perform T2 mapping clinically. For example, the scan time for four sets of spin‐echo imaging, with repetition time (TR) = 2 seconds and matrix size = 128 × 128, is around 17 minutes. The long scan time also makes spin‐echo MRI‐based T2 mapping highly vulnerable to motion artifacts.
The use of the Carr‐Purcell‐Meiboom‐Gill (CPMG) scheme in fast spin‐echo (FSE) pulse sequence 6 enables T2 mapping at shorter scan time, as compared with spin‐echo imaging. With FSE‐based T2 mapping protocols, whole‐brain multi‐TE images (eg, of four different TEs) may be obtained in around 3 minutes. However, FSE‐based T2 mapping has several limitations. First, FSE data acquired in 3 minutes may still be susceptible to potential motion‐related artifacts in challenging subject populations. Second, the radiofrequency (RF) pulse imperfections in CPMG scans lead to stimulated echoes and other high‐order echoes 7, which, in turn, make FSE‐based T2 mapping inaccurate.
A number of methods have been developed to address the above‐mentioned issues. First, in order to reduce the scan time and potential motion‐related artifact, T2 mapping data may be obtained more quickly with Cartesian subsampling 8 or compressed sensing 9 schemes and then undergo model‐based or constrained reconstruction 8 that incorporates temporal correlation in k‐space data and other prior knowledge into T2 mapping. Second, in order to reduce the impact of stimulated and high‐order echoes on CPMG‐based T2 mapping, the gradient waveforms in pulse sequences may be modified to eliminate some of the high‐order echo pathways 18. In those modified multiecho pulse sequences 21, the produced CPMG signals could be analyzed with an extended phase graph (EPG) method 18. In a recent paper, Huang et al 27 reported a fast T2 mapping protocol that integrated highly undersampled multiecho MRI data acquisition and EPG‐model–based T2 calculation, producing a T2‐mapping protocol with scan time <30 seconds.
Built upon the previously developed methods, here we report a single‐shot T2‐mapping protocol the integrates 1) a new single‐shot echo‐split gradient‐spin‐echo (GRASE) acquisition method 28 and 2) parametric multiplexed sensitivity encoding based on projection onto convex sets (parametric‐POCSMUSE) reconstruction. Our new approach has a number of strengths as compared with existing T2‐mapping protocols. First, using the new method, a single set of multiecho images can be acquired in a single shot, with scan time of around 0.2 seconds (eg, for sampling TEs at 30, 60, 90, and 120 msec) per slice. As compared with previously reported T2 mapping protocols that have significant time gaps (eg, at least a few seconds) between data acquisition for different TEs, our single‐shot scan strategy can more effectively reduce the negative impact of subject motion on T2 mapping accuracy. Second, our new echo‐split GRASE sequence is designed to eliminate stimulated echoes and most high‐order echoes in CPMG signals, and the residual high‐order echoes are taken into consideration in an EPG‐model–based parametric‐POCSMUSE reconstruction, enabling highly accurate T2 measurement from single‐shot data. Third, the imperfections in RF pulses and slice profiles are taken into consideration when performing parametric‐POCSMUSE reconstruction, minimizing hardware‐induced errors in T2 mapping.

Related Topics

    loading  Loading Related Articles