Validation of a T1 and T2 mapping software for quantitative MRI

Methods The T1 & T2 mapping modules were implemented in Segment [3] and support the pulse sequence types shown in Table 1. T1 estimates were initialized by a lookup table search in an interval of 0-4000 ms and a 5 ms gap between entries. T2 estimates were initialized by linear regression to the signal logarithm. Relaxation time estimates were refined using a C-implementation of the Nelder-Mead Simplex method. Twelve Gadolinium-Agarose phantoms were used for validation at a 1.5T and a 3T MR-scanner (Siemens, Erlangen). Reference T1 values were determined using an Inversion Recovery (IR) Spin Echo sequence with TR = 10 seconds. Reference T2 values were determined using a Spin Echo sequence with TR = 10 seconds. Nelder-Mead Simplex optimization available in Matlab (Math Works, Natick, MA; 2013a) was used for reference relaxation time values. The proposed modules generated relaxation time maps from two free-breathing, single-shot bSSFP sequences. Saturation-Recovery (SR) and T2-preparation prepulses were used to create T1 & T2 weighting. Apart from T2-prepared images, a saturation prepared image was used in the T2 estimates, as previously proposed [4]. In addition, T1 was estimated with MOLLI at 1.5T using 5(3b)3 and 4(1b)3(1b)2 schemes. Acquisition times for the three bSSFP sequences were 13 min 15 sec for T1-SR, 11 sec for T1-MOLLI and 14 min 3 sec for the T2 sequence. Bias and Variability were determined using modified Bland-Altman analysis. Error percentages were computed by dividing the difference between the evaluated method and the reference with the reference value.