MAGORINO: Magnitude-only fat fraction and R2* estimation with Rician noise modelling
Purpose: Magnitude-based fitting of chemical shift-encoded data enables proton density fat fraction (PDFF) and R2* estimation where complex-based methods fail or when phase data is inaccessible or unreliable. However, traditional magnitude-based fitting algorithms do not account for Rician noise, creating a source of bias. To address these issues, we propose an algorithm for Magnitude-Only PDFF and R2* estimation with Rician Noise modelling (MAGORINO). Methods: Simulations of multi-echo gradient echo signal intensities are used to investigate the performance and behavior of MAGORINO over the space of clinically-plausible PDFF, R2* and signal-to-noise ratio (SNR) values. Fitting performance is assessed through detailed simulation, including likelihood function visualization, and in a multi-site, multi-vendor and multi-field-strength phantom dataset and in vivo. Results: Simulations show that Rician noise-based magnitude fitting outperforms existing Gaussian noise-based fitting and reveals two key mechanisms underpinning the observed improvement. Firstly, the likelihood functions exhibit two local optima; Rician noise modelling increases the chance the global optimum corresponds to the ground truth. Secondly, when the global optimum corresponds to ground truth for both noise models, the optimum from Rician noise modelling is closer to ground truth. Multisite phantom experiments show good agreement of MAGORINO PDFF with reference values, and in vivo experiments replicate the performance benefits observed in simulation. Conclusion: MAGORINO reduces Rician noise-related bias in PDFF and R2* estimation, thus addressing a key limitation of existing magnitude-only fitting methods. Our results offer an insight into the importance of the noise model for selecting the correct optimum when multiple plausible optima exist.
Code (1)
Similar Papers 제목 키워드 기반
Rician likelihood loss for quantitative MRI using self-supervised deep learning
Purpose: Previous quantitative MR imaging studies using self-supervised deep learning have reported biased parameter estimates at low SNR. Such systematic errors arise from the choice of Mean Squared Error (MSE) loss fun…
parameter estimationQuantitative MRIRealistic noise synthesis reduces bias and improves tissue microstructure estimation with supervised machine learning
Diffusion MRI enables non-invasive probing of tissue microstructure, but accurate parameter estimation is challenged by noise-related effects. In supervised machine learning frameworks trained on simulated data, discrepa…
Noise EstimationFast Cauchy-Rician Modelling of SAR Images with Method of Algebraic Moments Estimator
SAR technology has been intensively implemented for geo-sensing and mapping purposes due to its advantages of high azimuthal resolution and weather-independent operation compared to other remote sensing technologies. Mod…
parameter estimationA novel and efficient parameter estimation of the Lognormal-Rician turbulence model based on k-Nearest Neighbor and data generation method
In this paper, we propose a novel and efficient parameter estimator based on $k$-Nearest Neighbor ($k$NN) and data generation method for the Lognormal-Rician turbulence channel. The Kolmogorov-Smirnov (KS) goodness-of-fi…
parameter estimationOn 1-Laplacian Elliptic Equations Modeling Magnetic Resonance Image Rician Denoising
Modeling magnitude Magnetic Resonance Images (MRI) rician denoising in a Bayesian or generalized Tikhonov framework using Total Variation (TV) leads naturally to the consideration of nonlinear elliptic equations. These i…
Denoising