Comparison of Neurite Orientation Dispersion and Density Imaging and Two-Compartment Spherical Mean Technique Parameter Maps in Multiple Sclerosis

Johnson, Daniel A.; Ricciardi, Antonio; Brownlee, Wallace; Kanber, Baris; Prados, Ferrán; Collorone, Sara; Kaden, Enrico; Toosy, Ahmed; Alexander, Daniel C.; Wheeler-Kingshott, C; Ciccarelli, Olga; Grussu, Francesco

doi:10.3389/fneur.2021.662855

Cited by 17 publications

(24 citation statements)

References 45 publications

(72 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…39 Abnormality 4 also has lower v cyl and λ cyl than average WM and exemplifies typical white matter lesions in multiple sclerosis reported in Ref. 40. We highlight that these abnormalities are not exact representations of specific pathologies, which vary widely, but serve to demonstrate estimation performance in abnormal tissue configurations likely to arise in practice.…”

Section: Simulated Abnormal Parameter Combinationssupporting

confidence: 58%

Training data distribution significantly impacts the estimation of tissue microstructure with machine learning

Gyori

Palombo

Clark

et al. 2021

Magnetic Resonance in Med

Self Cite

View full text Add to dashboard Cite

Purpose Supervised machine learning (ML) provides a compelling alternative to traditional model fitting for parameter mapping in quantitative MRI. The aim of this work is to demonstrate and quantify the effect of different training data distributions on the accuracy and precision of parameter estimates when supervised ML is used for fitting. Methods We fit a two‐ and three‐compartment biophysical model to diffusion measurements from in‐vivo human brain, as well as simulated diffusion data, using both traditional model fitting and supervised ML. For supervised ML, we train several artificial neural networks, as well as random forest regressors, on different distributions of ground truth parameters. We compare the accuracy and precision of parameter estimates obtained from the different estimation approaches using synthetic test data. Results When the distribution of parameter combinations in the training set matches those observed in healthy human data sets, we observe high precision, but inaccurate estimates for atypical parameter combinations. In contrast, when training data is sampled uniformly from the entire plausible parameter space, estimates tend to be more accurate for atypical parameter combinations but may have lower precision for typical parameter combinations. Conclusion This work highlights that estimation of model parameters using supervised ML depends strongly on the training‐set distribution. We show that high precision obtained using ML may mask strong bias, and visual assessment of the parameter maps is not sufficient for evaluating the quality of the estimates.

show abstract

Section: Simulated Abnormal Parameter Combinationssupporting

confidence: 58%

Training data distribution significantly impacts the estimation of tissue microstructure with machine learning

Gyori

Palombo

Clark

et al. 2021

Magnetic Resonance in Med

Self Cite

View full text Add to dashboard Cite

show abstract

“…Somewhat surprisingly, given a weak negative association reported elsewhere, 60 NODDI ICVF and ISOVF were weakly positively related at low ISOVF values. A larger ‘hindered’ diffusion volume fraction (ECVF), assumed to be due to glial cells infiltration and cell bodies, could lead to both a reduction in ISOVF and ICVF.…”

Section: Discussionmentioning

confidence: 53%

“…In our cohort of patients with recently diagnosed RRMS, we observed an unexpected increase in NODDI ICVF within NAWM over one year. NODDI ICVF represents the 'restricted' diffusion signal fraction within the non-isotropic diffusion signal; ICVF is typically lower in RRMS compared to healthy controls 28,60,67 and longitudinal decreases in NAWM ICVF have previously been noted. 68 The biological mechanism driving the increase in NODDI ICVF, with no concomitant change in ISOVF, is unclear, but could partly be attributed to axonal swelling, axonal bundling following demyelination, or axonal repair.…”

Section: Normal-appearing White Mattermentioning

confidence: 99%

Longitudinal microstructural MRI markers of demyelination and neurodegeneration in early relapsing-remitting multiple sclerosis: magnetisation transfer, water diffusion and g-ratio

York

Meijboom

Thrippleton

et al. 2022

Preprint

View full text Add to dashboard Cite

IntroductionRelapsing-remitting multiple sclerosis (RRMS) is a chronic, immune-mediated neurodegenerative disease with a highly heterogeneous disease course. Conventional MRI is widely used for diagnosis and disease monitoring but provides limited specific information on MS-related damage. Improved in vivo biomarkers, which are sensitive to subtle changes in RRMS pathology, are needed for early patient stratification and evaluation of neuroprotective and disease modifying therapies (DMTs).Quantitative MRI methods, such as myelin-sensitive magnetisation transfer ratio (MTR) or saturation (MTsat) imaging, axon-sensitive diffusion Neurite Orientation Dispersion and Density Imaging (NODDI) and derived aggregate measures (e.g. MR g-ratio) may provide more specific indices of microstructural damage in RRMS.The aim of this study was to determine the sensitivity of microstructural MRI markers (magnetisation transfer ratio [MTR], MTsat, NODDI isotropic/intraneurite volume fractions, and g-ratio) to myelin and axonal damage in white matter lesions (WML) and normal appearing white matter (NAWM) and their longitudinal change in early RRMS.MethodsSeventy-seven people were recruited to an extended MRI sub-study of Future-MS, a longitudinal study of people with recently diagnosed RRMS. MR imaging and clinical assessment were performed at baseline, prior to initiation of any DMTs, and one year later. Twelve healthy volunteers additionally received the same MR protocol, repeated within two weeks, to determine test-retest agreement. Approval was granted from the local Research Ethics committee and participants provided written informed consent.A 3T MR imaging protocol included structural 3D T1-weighted and T2-weighted FLAIR sequences, magnetisation transfer acquisition (comprising two proton density images with and without a saturation pulse, plus a T1-weighted image), and multishell diffusion-weighted 2D echo-planar imaging.MTR and MTsat were calculated from magnetisation transfer data, and NODDI isotropic and intraneurite volume fractions (ISOVF and ICVF, respectively) from diffusion-weighted data. The g-ratio was calculated from MTsat and NODDI data. WML and NAWM tissue masks were segmented from structural images.Tissue contrast for MTsat and MTR was compared with paired t-tests. Voxelwise distributions of microstructural metrics were examined graphically.Longitudinal change in MR-derived measures in NAWM and WML was assessed with paired t-tests (α=0.05) with follow-up linear mixed models, where significant, to control for confounding factors, with False Discovery Rate (FDR) correction for multiple comparisons.ResultsComplete longitudinal data for MTsat and MTR was available for sixty-two patients, and sixty patients for NODDI and g-ratio.Contrast between NAWM and WML was greater for MTsat than MTR (mean difference = 16.9 [95% CI 15.96 to 17.89], paired t-test: t(74)=34.97, p<0.001). Voxelwise histograms showed a negative linear dependence on the approximation of T1 recovery (T1app) for MTsat, but not MTR, in NAWM and a non-linear association in WML.Simulations and voxelwise histograms demonstrated that g-ratio is inversely related to MTsat and NODDI ISOVF but positively related to NODDI ICVF. MTsat and NODDI ICVF were not associated in control white matter or NAWM but broadly co- correlated in WML.In NAWM, g-ratio and NODDI ICVF increased significantly over one year (mean difference = 0.004 [95% CI 0.001 to 0.007] and 0.002 [0.0002 to 0.004], paired t-test: t(59) = 2.60 and 2.29, p = 0.012 and 0.025 respectively) and MTsat decreased (mean difference = -0.032 [95% CI -0.061 to -0.003], t(61) = -2.19, p = 0.033). Neither MTR nor NODDI ISOVF changed significantly over one year (p=0.94 and p=0.67).After accounting for covariates and adjustment for multiple comparisons, the longitudinal increase in g-ratio and in NODDI ICVF remained significant (linear mixed model: adjusted mean difference = 0.007 and 0.004, t(75.9) = 3.08 and t(90.6) = 3.51, FDR-corrected p = 0.029 and 0.007). The decrease in MTsat did not survive correction for multiple comparisons (FDR-corrected p = 0.11).In WML, paired t-tests revealed increases in MTsat, MTR, NODDI ISOVF and ICVF over one year, but no change in g-ratio. Follow-up linear mixed models, however, showed that only changes in MTsat and NODDI metrics were significant after correction for confounding factors and multiple comparisons.Longitudinal change did not, however, generally exceed healthy control test-retest limits of agreement except for NODDI ISOVF and ICVF in WML.DiscussionHigher tissue contrast for MTsat compared with MTR suggests MTsat is more sensitive to demyelination than widely used MTR. Comparisons of microstructural MRI data distributions indicate that MTsat is better suited to measurement of subtle myelin loss in NAWM compared with MTR.NAWM data distributions suggest that MTsat and NODDI ICVF provide independent measures of myelin and axonal integrity. Simulations show that g-ratio increases as MTsat decreases, but changes in NODDI parameters may complicate interpretation of longitudinal change in g-ratio.The observed increase in g-ratio and MTsat in NAWM is indicative of subtle myelin loss in patients with early RRMS. The unexpected increase observed in NODDI ICVF suggests axonal swelling and/or axonal repair, and may confound g-ratio.Increases in MTsat and NODDI ICVF in WML are also suggestive of tissue repair. WML g-ratio did not change over time, likely due to competing effects of longitudinal MTsat and NODDI changes.Despite significant groupwise effects, detection of longitudinal change at an individual level may be limited by technique test-retest agreement in early RRMS.ConclusionMTsat and g-ratio are promising in vivo biomarkers of myelin integrity in early RRMS, and appear more sensitive than MTR for detecting subtle demyelination. Our findings suggest that MTsat and g-ratio measures may be most sensitive for detecting subtle change in NAWM, as compared with more severely damaged WML where T1 effects and neuroaxonal damage predominate. Complex dependence of g- ratio on NODDI parameters illustrates the limitation of interpreting such aggregate measures in isolation, and independent consideration of myelin-sensitive and axonal neuroimaging markers may ultimately be more informative for longitudinal tracking of neuropathology in RRMS. Technique reproducibility currently limits evaluation in individual patients, and future research is warranted to validate diffusion and MT models across heterogeneous microstructural damage found in MS, for translation into clinical practice and trial platforms.

show abstract

“…Several studies have reported that DTI is superior to conventional MRI in the evaluation of NAWM in MS patients; however, advanced dMRI techniques, as described in this review, are more sensitive and have the potential to show specific pathological changes in the microstructure in vivo . Clinically, advanced dMRI has been reported to be more sensitive in detecting microstructural changes in the brain that correlate with patient disability 118 and in tracking disease progression in NAWM over time, even before clinical symptoms occur. 119 In addition, several reports have shown the usefulness of advanced dMRI in NAWM and demyelinating foci.…”

Section: Discussionmentioning

confidence: 99%

Advanced Diffusion MR Imaging for Multiple Sclerosis in the Brain and Spinal Cord

Hori

Maekawa

Kamiya

et al. 2022

MRMS

View full text Add to dashboard Cite

Diffusion tensor imaging (DTI) has been established its usefulness in evaluating normal-appearing white matter (NAWM) and other lesions that are difficult to evaluate with routine clinical MRI in the evaluation of the brain and spinal cord lesions in multiple sclerosis (MS), a demyelinating disease. With the recent advances in the software and hardware of MRI systems, increasingly complex and sophisticated MRI and analysis methods, such as q-space imaging, diffusional kurtosis imaging, neurite orientation dispersion and density imaging, white matter tract integrity, and multiple diffusion encoding, referred to as advanced diffusion MRI, have been proposed. These are capable of capturing in vivo microstructural changes in the brain and spinal cord in normal and pathological states in greater detail than DTI.This paper reviews the current status of recent advanced diffusion MRI for assessing MS in vivo as part of an issue celebrating two decades of magnetic resonance in medical sciences (MRMS), an official journal of the Japanese Society of Magnetic Resonance in Medicine.

show abstract

Comparison of Neurite Orientation Dispersion and Density Imaging and Two-Compartment Spherical Mean Technique Parameter Maps in Multiple Sclerosis

Cited by 17 publications

References 45 publications

Training data distribution significantly impacts the estimation of tissue microstructure with machine learning

Training data distribution significantly impacts the estimation of tissue microstructure with machine learning

Longitudinal microstructural MRI markers of demyelination and neurodegeneration in early relapsing-remitting multiple sclerosis: magnetisation transfer, water diffusion and g-ratio

Advanced Diffusion MR Imaging for Multiple Sclerosis in the Brain and Spinal Cord

Contact Info

Product

Resources

About