Original contributionLinear signal combination T2 spectroscopy
Introduction
In pure water, transverse relaxation takes the form of a monoexponential decay characterized by a single time constant “T2” (Fig. 1):
In complex systems such as tissue, however, a spectrum of T2 values may be present:
The signal decay is then a sum of the individual T2 components:
Changes in the T2 spectrum may reflect underlying structure and/or pathology within the tissue. For example, the T2 spectrum of white matter contains a component at ~15 ms associated with myelin, and a second component ~90 ms associated with intra-/extra-cellular water [[1], [2], [3], [4], [5], [6], [7]] (Fig. 1). In diseases such as multiple sclerosis, myelin loss may be detected directly through a reduction in the short T2 component [[8], [9], [10], [11], [12]]. Similar approaches haven been proposed for other anatomy [2,13,14].
In MRI, data is acquired in the time domain (Eq. (3)). To determine the T2 spectrum, an inversion is required. Based on the form of Eq. (3), one plausible approach could be Inverse Laplace Transformation:
Unfortunately, MR signals of the form in Eq. (3) aren't appropriate for a strict use of the inverse Laplace Transform [15,16]. Rather, Eq. (3) can be considered as a Fredholm integral equation of the first kind [15]. To solve Eq. (3), most approaches employ some form of non-negative least squares (NNLS) fitting [17]. The most straightforward method is to directly apply the NNLS algorithm. Unfortunately, a difficulty arises when the number of free parameters significantly exceeds the number of data points – a situation typically encountered with in vivo T2 decay experiments. Under these circumstances, the problem becomes ill-posed [16]. As a result, most approaches to inverting Eq. (3) employ some form of regularization [18,19]. This involves adding a smoothing term to the NNLS fitting equation [20]. The advantage of regularization is that it reduces the ill-possedness of the problem, and stabilizes the fit. However, the disadvantage is that the derived solution is not necessarily identical to the underlying true T2 spectrum [21]. Techniques have been developed to optimize the smoothness-accuracy tradeoff [16,20]. Nonetheless, there is limited range of possible solutions within the constraints of a regularization term.
In this study, a new technique for T2 spectroscopy is presented. It is based on a weighted linear combination of time domain data. Like regularization, this approach provides a spectrum that is smoothed relative to the true T2 spectrum. However, the process of linear combination can be cast as a filter design problem. Therefore, unlike regularization, one has a large degree of control over the nature of the smoothing. Furthermore, we will demonstrate that the effects of smoothing may be at least partially removed through deconvolution.
In the sections below, the basic theory underlying this linear combination T2 spectroscopy technique will be derived. Simulations will be presented to characterize its behaviour. Finally, the technique will be applied to experimental MRI phantom data to demonstrate feasibility.
Section snippets
Theory
The spectroscopic technique to be developed in this study builds on the concept of linear signal combination filtering. The theory underlying linear signal combination filtering will be first reviewed. The extension to T2 spectroscopy will be then presented.
Simulations
Fig. 8 illustrates some sample results from the simulation. Following linear signal combination, the overall shape of the calculated T2 spectrum approximately resembles the original – although with some obvious distortions. Following deconvolution, the distortion is reduced. In fact, the T2 spectrum estimate using the filter with SNR factor = 10−3 and suppression factor = 10−2 (Fig. 8a) is almost identical to the true T2 spectrum. On the other hand, the T2 spectrum estimate from the filter with
Discussion
The linear combination technique can provide accurate estimates of T2 spectra. However, the effectiveness of the technique was found to depend strongly on the filter design parameters. Specifically, the narrower the filter bandwidth and/or the more strict the suppression factor, the better the spectrum estimate. This is due to the fact that narrower filters and stricter suppression both minimize the amount of distortion in the spectrum following linear combination. In turn, the less distortion,
Conclusions
This study presented a new technique for T2 spectroscopy. It is based on the linear combination of signals. It was demonstrated how this procedure can be cast as a filter design problem. In this context, the quality of the spectrum estimate improves as the filter passband is narrowed, and the stopband suppression factors are tightened. Conversely, while the spectrum estimates improve, the SNR performance decreases. Thus, the filter design requires a tradeoff between T2 spectral distortion and
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
References (36)
- et al.
Longitudinal changes in myelin water fraction in two MS patients with active disease
J Neurol Sci
(2009) - et al.
Numerical estimation of relaxation and diffusion distributions in two dimensions
Prog Nucl Magn Reson Spectrosc
(2012) - et al.
Analysis of biological NMR relaxation data with continuous distributions of relaxation times
J Magn Reson
(1986) - et al.
Quantitative interpretation of NMR relaxation data
J Magn Reson
(1989) - et al.
Insights into brain microstructure from the T-2 distribution
Magn Reson Imaging
(2006) - et al.
Using 3D spatial correlations to improve the noise robustness of multi component analysis of 3D multi echo quantitative T2 relaxometry data
Neuroimage
(2018) - et al.
Magnetization-transfer and T2 relaxation components in tissue
Magn Reson Med
(1995) - et al.
Anisotropy of NMR properties of tissues
Magn Reson Med
(1994) - et al.
A pathology-MRI study of the short-T2 component in formalin-fixed multiple sclerosis brain
Neurology
(2000) - et al.
In vivo measurement of T-2 distributions and water contents in normal human brain
Magn Reson Med
(1997)
In-vivo visualization of myelin water in brain by magnetic resonance
Magn Reson Med
Spin-spin relaxation in experimental allergic encephalomyelitis - analysis of CPMG data using a nonlinear least-squares method and linear inverse-theory
Magn Reson Med
The short T2 component is absent in chronically demyelinated plaques in formalin-fixed multiple sclerosis brain
Neurology
Magnetization transfer and multicomponent T2 relaxation measurements with histopathologic correlation in an experimental model of MS
J Magn Reson Imaging
Correlation of apparent myelin measures obtained in multiple sclerosis patients and controls from magnetization transfer and multicompartmental T-2 analysis
Magn Reson Med
Is multicomponent T-2 a good measure of myelin content in peripheral nerve?
Magn Reson Med
Characterizing white matter with magnetization transfer and T-2
Magn Reson Med
Linear combination filtering for T-2-selective imaging of the knee
Magn Reson Med
Cited by (1)
A method for estimating the statistical error of the solution in the inverse spectroscopy problem
2021, Izvestiya Instituta Matematiki i Informatiki Udmurtskogo Gosudarstvennogo Universiteta