Myocardial T1 mapping using an instantaneous signal loss simulation modeling and a Bayesian estimation method: A robust T1 extraction method free of tuning parameters (RF_FR_301)

Background:


The Instantaneous Signal Loss Simulation (InSiL) model is a promising alternative to the classical mono- exponential fitting of the Modified Look-Locker Inversion-recovery (MOLLI) sequence in cardiac 𝑇1 mapping applications, which achieves better accuracy and is less sensitive to heart rate (HR) variations. Classical non- linear least squares (NLLS) estimation methods require some parameters of the model to be fixed a priori in order to give reliable 𝑇1 estimations and avoid outliers. This introduces further bias in the estimation, reducing the advantages provided by the InSiL model. 







Methods: In this paper, a novel Bayesian estimation method using a hierarchical model is proposed to fit the parameters of the InSiL model. The hierarchical Bayesian modeling has a shrinkage effect that works as a regularizer for the estimated values, by pulling spurious estimated values toward the group-mean, hence reducing greatly the number of outliers. Simulations, physical phantoms, and in-vivo human cardiac data have been used to show that this approach estimates accurately all the InSiL parameters, and achieve high precision estimation of the 𝑇1 compared to the classical MOLLI model and NLLS InSiL estimation. T1 maps were computed from original MOLLI images with: the MOLLI model, InSiL model with a least square method for all 4 parameters (LSQ(4p)), least square method with one InSiL parameter fixed and 3 parameters  (LSQ(3p)), Bayesian method with a shrinkage prior on the 3  parameters -HBI(3p)), and on 2 parameters (HBI(2P)).

Results:


Contrary to previously proposed approaches based on non- linear least square methods, all the model parameters are estimated without fixing some of them to a priori values. The parameters of the multivariate Gaussian prior used in the Bayesian model are estimated directly from the data, together with the model parameters. This makes the method free of tuning parameters, hence more objective, with increase accuracy and precision. Numerical simulation, physical phantoms, and clinical data show that 𝑇1 esti- mation uncertainty is considerably reduced thanks to the shrinkage effect of the priors, and falls in the range of what is obtained with the classical MOLLI estimation method. Hence, this estimation method allows for the first time to exploit the full potential of the InSiL model in clinical practice, and to get 𝑇1 estimates that are more accurate, and less sensitive to confounding factors such as noise level, tissue properties, HR variability, sequence design, and readout sequence. 







Conclusion: Despite better accuracy and precision, in-depth clinical studies are needed to evaluate the added value of the InSiL model for the patient. The algorithm would benefit from being retrospectively applied to numerous cohorts, especially those for which a conventional T1 has not revealed differences. This could help study whether the added value of this new diagnostic precision and accuracy would make T1 an even better imaging biomarker. This could be explored in specific physiopathological contexts, such as pre- and post-therapy assessments, or longitudinal follow-ups.