A Unified Maximum Likelihood Framework for Simultaneous Motion and $T_{1}$ Estimation in Quantitative MR $T_{1}$ Mapping

Abstract

In quantitative MR T mapping, the spin-lattice relaxation time T of tissues is estimated from a series of T -weighted images. As the T estimation is a voxel-wise estimation procedure, correct spatial alignment of the T -weighted images is crucial. Conventionally, the T -weighted images are first registered based on a general-purpose registration metric, after which the T map is estimated. However, as demonstrated in this paper, such a two-step approach leads to a bias in the final T map. In our work, instead of considering motion correction as a preprocessing step, we recover the motion-free T map using a unified estimation approach. In particular, we propose a unified framework where the motion parameters and the T map are simultaneously estimated with a Maximum Likelihood (ML) estimator. With our framework, the relaxation model, the motion model as well as the data statistics are jointly incorporated to provide substantially more accurate motion and T parameter estimates. Experiments with realistic Monte Carlo simulations show that the proposed unified ML framework outperforms the conventional two-step approach as well as state-of-the-art model-based approaches, in terms of both motion and T map accuracy and mean-square error. Furthermore, the proposed method was additionally validated in a controlled experiment with real T -weighted data and with two in vivo human brain T -weighted data sets, showing its applicability in real-life scenarios.

Multiscale brain MRI super-resolution using deep 3D convolutional networks

“…The use of SR techniques has been studied in many works in the context of brain MRI analysis: structural MRI (Manjón et al 2010b;Rousseau et al 2010a;Manjón et al 2010a;Rueda et al 2013;Shi et al 2015), diffusion MRI (Scherrer et al 2012;Poot et al 2013;Fogtmann et al 2014;Steenkiste et al 2016), spectroscopy MRI (Jain et al 2017), quantitative T 1 mapping (Ramos-Llordén et al 2017;Van Steenkiste et al 2017), fusion of orthogonal scans of moving subjects (Gholipour et al 2010;Rousseau et al 2010b;Kainz et al 2015;Jia et al 2017). The development of efficient and accurate SR techniques for 3D MRI data could be a major step forward for brain studies.…”

High‐fidelity, accelerated whole‐brain submillimeter in vivo diffusion MRI using gSlider‐spherical ridgelets (gSlider‐SR)

“…In-vivo real data that was used in the experiment section required a motion correction scheme, which we smoothly integrated in the gSlider-SR as an iterative registration step with the popular FLIRT algorithm (34). While this approach provided very good results, motion and eddy correction can be explicitly modeled within the forward-model of Eq.2 as is done in (14,36,37,20). This will ensure that the super-resolution DWI dataset and the motion parameters, which vary not only for each diffusion direction but also along with RF-encoding profiles, can be simultaneously estimated within an integrated framework, improving the performance of gSlider-SR (38,36).…”

NOVIFAST: A Fast Algorithm for Accurate and Precise VFA MRI <inline-formula> <tex-math notation="LaTeX">${T}_{1}$ </tex-math> </inline-formula> Mapping

“…(1) with TR = 5 ms and with FAs given by the AN =10 FA set) were created based on ground-truth T 1GT and K GT maps. Those maps were estimated from a simulated IR gradient recalled echo sequence, with similar settings as those given in [39]. The size of both 3D maps was 111 × 93 × 71 with an isotropic voxel size of 1.5 mm.…”

“…A non-central χ distribution typically applies when complex images from several coils are combined with the Sum of Squares (SoS) method [13], [36]. It has been shown recently that solving an ML estimation problem with non-central χ or Rician distributed data is equivalent to iteratively solving a collection of NLLS subproblems [39], [50]. As a result, NOVIFAST can be integrated into this approach, by solving each of the NLLS subproblems, thereby greatly boosting the speed of the overall ML estimation procedure.…”