Characterization of brain tumours with spin–spin relaxation: pilot case study reveals unique T 2 distribution profiles of glioblastoma, oligodendroglioma and meningioma

Abstract

Prolonged spin-spin relaxation times in tumour tissue have been observed since some of the earliest nuclear magnetic resonance investigations of the brain. Over the last three decades, numerous studies have sought to characterize tumour morphology and malignancy using quantitative assessment of T relaxation times, although attempts to categorize and differentiate tumours have had limited success. However, previous work must be interpreted with caution as relaxation data were typically acquired using a variety of multiple echo sequences with a range of echoes and T decay curves and were frequently fit with monoexponential analysis. We defined the distribution of T components in three different human brain tumours (glioblastoma, oligodendroglioma, meningioma) using a multi-echo sequence with a greater number of echoes and a longer acquisition window than previously used (48 echoes, data collection out to 1120 ms) with no a priori assumptions about the number of exponential components contributing to the T decay. T relaxation times were increased in tumour tissue and each tumour showed a distinct T distribution profile. Tumours have complex and unique compartmentalization characteristics. Quantitative assessment of T relaxation in brain cancer may be useful in evaluating different grades of brain tumours on the basis of their T distribution profile, and has the potential to be a non-invasive diagnostic tool which may also be useful in monitoring therapy. Further study with a larger sample size and varying grades of tumours is warranted.

Transferring principles of solid-state and Laplace NMR to the field of in vivo brain MRI

“…A discrete P (R 2 , D) distribution is then estimated by solving a discretized version of Eq. (2) via a standard non-negative least squares (NNLS) algorithm (Lawson and Hanson, 1974). Points with nonzero weights are stored and merged with a new randomly generated set of 200 (R 2 , D || , D ⊥ , θ, ϕ) points, and the weights of the merged set of points are found through a NNLS fit (Lawson and Hanson, 1974).…”

“…(2) via a standard non-negative least squares (NNLS) algorithm (Lawson and Hanson, 1974). Points with nonzero weights are stored and merged with a new randomly generated set of 200 (R 2 , D || , D ⊥ , θ, ϕ) points, and the weights of the merged set of points are found through a NNLS fit (Lawson and Hanson, 1974). The process of selecting points with nonzero weights, subsequently merging them with a random (R 2 , D || , D ⊥ , θ , ϕ) configuration, and finally fitting the merged set is repeated a total of 20 times in order to find a P (R 2 , D || , D ⊥ , θ, ϕ) distribution yielding a low residual sum of squares.…”

“…Following 20 rounds, the resulting (R 2 , D || , D ⊥ , θ , ϕ) configuration is selected, split, and subjected to a small random mutation. The original and mutated configurations are merged and a new P (R 2 , D || , D ⊥ , θ , ϕ) distribution is determined by fitting the merged set to the data using the NNLS algorithm (Lawson and Hanson, 1974). The mu- Experimental (gray circles) and fitted (black points) S(τ E , b) signals from three representative voxels containing white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF).…”

Evaluation of principal component analysis image denoising on multi‐exponential MRI relaxometry

“…Multi‐exponential MRI relaxometry is a powerful tool for characterizing tissue at the sub‐voxel level, the most well known example of which is the use of multi‐exponential () relaxometry for myelin water imaging in brain and nerve . Similar myelin imaging has been implemented based on multi‐exponential relaxometry, and other example applications of multi‐exponential MRI relaxometry include characterization of muscle, 6–9 cartilage, and tumors . Similarly, inversion‐recovery based quantitative magnetization transfer is effectively a bi‐exponential signal analysis .…”

Transferring principles of solid-state and Laplace NMR to the field of <i>in vivo</i> brain MRI

“…The signals from heterogeneous materials are often approximated as integral transformations of nonparametric distributions of relaxation rates or diffusivities (Istratov and Vyvenko, 1999), which may be estimated by Laplace inversion of data acquired as a function of the relevant experimental variable (Whittall and MacKay, 1989). Within the context of human brain MRI, the components of the distributions have been assigned to water populations residing in specific tissue microenvironments such as myelin (Mackay et al, 1994) and tumors (Laule et al, 2017). The power to resolve and individually characterize the different components can be boosted by combining multiple relaxation-and diffusion-encoding blocks and analyzing the data as joint probability distributions of the relevant observables (English et al, 1991).…”