NOVIFAST: A fast algorithm for accurate and precise VFA MRI

A fast non-linear least squares method for accurate and precise estimation of T1 from SPGR signals.

In quantitative magnetic resonance T1 mapping, the Variable Flip Angle (VFA) steady state spoiled gradient recalled echo (SPGR) imaging technique is popular as it provides a series of high resolution T1 weighted images in a clinically feasible time. Fast, linear methods that estimate T1 maps from these weighted images have been proposed, such as DESPOT1 and iterative reweighted linear least squares (IRWLLS). More accurate, nonlinear least squares (NLLS) estimators are in play, but these are generally much slower and require careful initialization. In this work, we present NOVIFAST, a novel NLLS-based algorithm specifically tailored to VFA SPGR T1 mapping. By exploiting the particular structure of the SPGR model, a computationally efficient, yet accurate and precise T1 map estimator is derived. Simulation and in vivo human brain experiments demonstrate a twenty-fold speed gain of NOVIFAST compared to conventional gradient-based NLLS estimators, while maintaining a high precision and accuracy. Moreover, NOVIFAST is eight times faster than efficient implementations of the VARPRO method. Furthermore, NOVIFAST is shown to be robust against initialization.
Main reference:
[1] Ramos-Llordén, G., Vegas-Sánchez-Ferrero, G., Björk, M., Vanhevel, F., Parizel, P. M., San José Estépar, R., den Dekker, A. J., and Sijbers, J., "NOVIFAST: a fast algorithm for accurate and precise VFA MRI T1 mapping,"
IEEE Trans. Med. Imag., in press, DOI: 10.1109/TMI.2018.2833288
http://dx.doi.org/10.1109/TMI.2018.2833288

This Matlab package contains the following Matlab scripts/functions:

1) novifast_1D.m (Function)
Description:
This code is the 1D version of the original NOVIFAST implementation [1]. It is meant to estimate the T1 value from an SPGR signal. To estimate a complete T1 map from an SPGR T1-weighted image set, we do not recommend to apply novifast_1D in a voxel-wise manner. Instead, we advocate for using 'novifast_image', also included in this NOVIFAST package.
2) novifast_image.m (Function)
Description:
This code is a vectorized version of the original NOVIFAST implementation [1]. It works on the VFA T1-weighted image set, and provides the estimated T1 map (NLLS sense) directly. It is not a voxel-wise, for-loop, implementation of 1D NOVIFAST. This is the code we recommend to use for fast VFA T1 mapping.

3) demo_novifast_image.m (script)
Description:
In this script, Novifast is run with real 3D SPGR MRI data.

4) demo_novifast_SNR.m (script)
Description:
In this script, the behavior of NOVIFAST I as a function of SNR and number of iterations is presented. Experiments are performed with simulated SPGR MRI data.

5) Simulated and actual in vivo SPGR data

For more information, we refer the user to:

https://visielab.uantwerpen.be/research-topics/novifast