[Paper Review] Fast, Precise Myelin Water Quantification using DESS MRI and Kernel Learning

Gopal Nataraj, Jon‐Fredrik Nielsen+2 more|Deep Blue (University of Michigan)|Sep 24, 2018

Advanced MRI Techniques and Applications39 references4 citations

TL;DR

This paper proposes a fast, precise method for myelin water quantification using dual-echo steady-state (DESS) MRI combined with kernel-based regression (PERK). It demonstrates that optimized DESS scans and PERK estimation achieve myelin water fraction (MWF) estimates comparable to gold-standard multi-echo spin-echo (MESE) acquisitions, enabling clinically feasible, high-precision myelin imaging.

ABSTRACT

Purpose: To investigate the feasibility of myelin water content quantification using fast dual-echo steady-state (DESS) scans and machine learning with kernels. Methods: We optimized combinations of steady-state (SS) scans for precisely estimating the fast-relaxing signal fraction ff of a two-compartment signal model, subject to a scan time constraint. We estimated ff from the optimized DESS acquisition using a recently developed method for rapid parameter estimation via regression with kernels (PERK). We compared DESS PERK ff estimates to conventional myelin water fraction (MWF) estimates from a longer multi-echo spin-echo (MESE) acquisition in simulation, in vivo, and ex vivo studies. Results: Simulations demonstrate that DESS PERK ff estimators and MESE MWF estimators achieve comparable error levels. In vivo and ex vivo experiments demonstrate that MESE MWF and DESS PERK ff estimates are quantitatively comparable measures of WM myelin water content. To our knowledge, these experiments are the first to demonstrate myelin water images from a SS acquisition that are quantitatively similar to conventional MESE MWF images. Conclusion: Combinations of fast DESS scans can be designed to enable precise ff estimation. PERK is well-suited for ff estimation. DESS PERK ff and MESE MWF estimates are quantitatively similar measures of WM myelin water content.

Motivation & Objective

• To develop a clinically practical method for myelin water fraction (MWF) quantification that reduces scan time compared to conventional multi-echo spin-echo (MESE) sequences.
• To investigate whether dual-echo steady-state (DESS) MRI, traditionally less precise for myelin imaging, can achieve comparable accuracy when combined with advanced parameter estimation techniques.
• To evaluate the feasibility of using kernel-based regression (PERK) for estimating the fast-relaxing signal fraction $f_{\mathrm{F}}$ from DESS data as a proxy for MWF.
• To validate that DESS PERK $f_{\mathrm{F}}$ estimates are quantitatively equivalent to MESE MWF estimates in simulation, in vivo, and ex vivo settings.

Proposed method

• Optimized combinations of DESS scans using Bayesian experiment design to maximize precision in estimating the fast-relaxing signal fraction $f_{\mathrm{F}}$ under a fixed scan time constraint.
• Applied the recently developed Parameter Estimation via Regression with Kernels (PERK) method to estimate $f_{\mathrm{F}}$ from DESS data, leveraging kernel regression for robustness and accuracy.
• Used a Gaussian kernel function $k(\mathbf{q}, \mathbf{q}') = \exp\left(-\frac{1}{2}\|\mathbf{q} - \mathbf{q}'\|_{\bm{\Lambda}^{-2}}^2\right)$ to model nonlinear relationships between DESS signal responses and $f_{\mathrm{F}}$.
• Employed a low-rank approximation of the kernel matrix via feature mapping $\mathbf{z}(\cdot)$ to enable efficient computation of the PERK estimator, reducing computational complexity.
• Formulated the approximate PERK estimator as $\widehat{\mathbf{x}}(\cdot) \leftarrow \mathbf{m}_{\mathbf{x}} + \mathbf{C}_{\mathbf{x}\mathbf{z}}(\mathbf{C}_{\mathbf{z}\mathbf{z}} + \rho\mathbf{I}_{Z})^{-1}(\mathbf{z}(\cdot) - \mathbf{m}_{\mathbf{z}})$, enabling scalable and accurate regression.
• Validated the method across simulations, in vivo human brain scans, and ex vivo tissue samples, comparing DESS PERK $f_{\mathrm{F}}$ to conventional MESE MWF.

Experimental results

Research questions

• RQ1 Can optimized DESS MRI acquisitions achieve precision in $f_{\mathrm{F}}$ estimation comparable to gold-standard MESE acquisitions?
• RQ2 Is the PERK kernel regression method suitable for estimating $f_{\mathrm{F}}$ from DESS data with high accuracy and low computational cost?
• RQ3 Are DESS PERK $f_{\mathrm{F}}$ estimates quantitatively equivalent to MESE MWF estimates in real biological tissue?
• RQ4 Can the combination of DESS and PERK enable fast, whole-brain myelin water imaging with clinically feasible scan times?
• RQ5 Does the method remain robust to model non-idealities such as multi-compartmental relaxation effects in in vivo and ex vivo settings?

Key findings

• Simulations show that DESS PERK $f_{\mathrm{F}}$ estimators achieve error levels comparable to those of conventional MESE MWF estimators.
• In vivo and ex vivo experiments demonstrate that DESS PERK $f_{\mathrm{F}}$ and MESE MWF estimates are quantitatively comparable as measures of white matter myelin water content.
• This study presents the first empirical evidence that myelin water images from steady-state (DESS) acquisitions are quantitatively similar to those from conventional multi-echo spin-echo (MESE) sequences.
• The optimized DESS scan combinations enable precise $f_{\mathrm{F}}$ estimation within clinically feasible scan times, reducing acquisition duration significantly compared to long-TR MESE.
• The PERK method effectively handles the ill-posed nature of $T_2$ distribution estimation from DESS data, achieving high accuracy even with limited SNR.
• The low-rank approximation of the kernel matrix enables efficient computation of the PERK estimator, making the method scalable for whole-brain imaging.