Improving parametric estimation in the brain from multispin-echo sequences using a fusion bootstrap moves solver

To simultaneously estimate the $$ field (along with the T₂ ) in the brain with multispin-echo (MSE) sequences and dictionary matching.

T₂ mapping provides clinically relevant information such as in the assessment of brain degenerative diseases. It is commonly obtained with MSE sequences, and accuracy can be further improved by matching the MSE signal to a precomputed dictionary of echo-modulation curves. For additional T₁ quantification, transmit $$ field knowledge is also required. Preliminary work has shown that although simultaneous brain $$ estimation along with T₂ is possible, it presents a bimodal distribution with the main peak coinciding with the true value. By taking advantage of this, the $$ maps are expected to be spatially smooth by applying an iterative method that takes into account each pixel neighborhood known as the fusion bootstrap moves solver (FBMS). The effect of the FBMS on $$ accuracy and piecewise smoothness is investigated and different spatial regularization levels are compared. Total variation regularization was used for both $$ and T₂ simultaneous estimation because of its simplicity as an initial proof-of-concept; future work could explore non edge-preserving regularization independently for $$ .

Improvements in $$ accuracy (up to 45.37% and 16.81% $$ error decrease) and recovery of spatially homogeneous maps are shown in simulations and in vivo 3.0T brain data, respectively.

Accurate $$ estimated values can be obtained from widely available MSE sequences while jointly estimating T₂ maps with the use of echo-modulation curve matching and FBMS at no further cost.