Synopsis

Keywords: Multiple Sclerosis, Diffusion/other diffusion imaging techniques, personalised-NODDI

Neurite Orientation Dispersion and Density Imaging (NODDI) is a multi-compartmental model for microstructure characterization utilising diffusion-weighted MRI. Here, a recently proposed personalized-NODDI (pNODDI) modelling technique was used to assess microstructural alterations in a cohort of Multiple Sclerosis (MS) patients. Our findings showed that pNODDI metrics were sensitive to MS pathology and could discriminate between healthy controls and MS. Furthermore, pNODDI metrics of key brain regions increased the discriminative power between the two groups. Overall, pNODDI provides a personalized and clinically feasible model for microstructure, increasing sensitivity to pathological alterations.

INTRODUCTION

Neurite Orientation Dispersion and Density Imaging (NODDI)¹ is one of the most utilised multi-compartmental diffusion MRI models, and relies on two assumptions: firstly, that parallel intra-axonal diffusivity (D_a) is equal to 1.7x10^-3 mm²/s; and secondly, that extra-cellular and intra-cellular compartments have the same intrinsic diffusivity equal to D_a. Recently, different implementations of NODDI have been proposed, in which the value of D_a is defined, for example, as the mean axial diffusivity (AD) within the body of the corpus callosum (BodyCC), calculated from the diffusion tensor (DT) fitting, in a group of healthy controls (HC)².
However, this approach does not account for the fact that D_a in the body of the corpus callosum may exhibit biological inter-subject variability, implying that residual biases may exist due to the fixed D_a value (NODDI metrics are known to change as D_a changes³). To account for this, we developed a personalized-NODDI (pNODDI) approach by substituting the fixed D_a with a personalized value obtained from a subject-specific pipeline⁴. Here, D_a was assumed to be equal to the mean AD of the BodyCC of each subject, thus allowing pNODDI to be subject-specific, as already proposed. In the present work the accuracy and sensitivity of pNODDI were tested in a cohort of subjects affected by progressive multiple sclerosis (MS). Orientation Dispersion Index and Neurite Density Index obtained with pNODDI (pODI and pNDI) of several brain regions were used to test the discriminative power of the method.

METHODS

Subjects & Acquisition
Data were collected as part of a project investigating quantitative MRI in progressive MS⁵.
Subjects were divided into two groups: 17 Healthy Controls (HC) (12 females; 57±10y) and 14 progressive MS patients (7 females; 57±8y).
The following MRI data were acquired using a 3T Philips Ingenia CX scanner:
- Brain 3D FLAIR (TR/TE = 4800/258 ms, TI = 1650 ms, flip angle 40°, 1x1x1 mm³) and 3D T1-weighted scans (TR/TE = 6.9/3.1 ms, TI = 810 ms, flip angle 8°, 1x1x1 mm³)
- Diffusion-weighted (DW) images were acquired with a spin-echo EPI sequence (TR/TE = 6287/96 ms, 2x2x2 mm³, 20/20/36 DW-directions/shell, b=1000/2000/2800 s/mm², 3 b=0 (b₀) images).
Pre-processing
Corrections for Gibb’s artefacts, background noise, susceptibility-induced and eddy currents distortions were performed on the DW images using a combination of MRtrix3⁶ and FSL⁷ commands.
T1w images were aligned to DW space and used to segment regions of interest (ROIs): white matter (WM), cortex grey matter (GM), genu, body, and splenium of the corpus callosum (GenuCC, BodyCC and SpleniumCC), the thalami (ThalLeft and ThalRight) and the hippocampi (HippoLeft, HippoRight). All ROI masks were thresholded at 0.99 to eliminate partial volume effects due to contamination from other tissues. A fully automated segmentation of WM lesions was performed on the 3D FLAIR images with the NicMS software, manually corrected by an experienced rater⁸. Normal-appearing white matter (NAWM) masks were obtained by subtracting lesions from the previously selected ROIs (NAWM, NAGenuCC, NABodyCC, NASpleniumCC).
pNODDI analysis
Mean AD was calculated in the NABodyCC for each subject (pD_a ) to replace default intra-axonal diffusivity in NODDI in the pNODDI pipeline.
pNDI and pODI metrics were calculated using the NODDI toolbox¹ (Matlab), after changing the input for the pNODDI’s pipeline as described above. Mean pODI and pNDI were evaluated in the defined ROIs, also including voxel’s value weighting for the presence of CSF⁹ (Figure 1).
Statistics
Statistical analyses were performed using SPSS¹⁰. Data normality was evaluated via a Shapiro-Wilk test. Then, the ability of pNODDI metrics to detect pathological alterations was assessed through independent-sample t-tests which compared HC and MS subjects. The discriminative power of pNODDI metrics was assessed by performing a discriminant analysis using the group as a dependent variable and considering, as independent variables, (i) pNDI (or pODI) of all brain regions together; (ii) pNDI (or pODI) of each specific brain region separately.

RESULTS

The distributions of pDa showed a mean value of (1.77±0.10)·10^-3 mm²/s for the HC group and (1.71±0.16)·10^-3 mm²/s for the MS group (Figure 2).
pNDI was sensitive to MS pathology and identified differences between MS and HC in all the ROIs except for the left thalamus and the hippocampi, while pODI was sensitive to MS pathology only in the thalami (Figure 3).
Furthermore, pNDI and pODI can discriminate between the two groups with high accuracy in the same regions in which they are sensitive to pathology (Figure 4).

DISCUSSION & CONCLUSIONS

Here, the pNODDI was tested on a cohort of progressive MS patients and HC. pNODDI was obtained by personalizing D_a with subject-specific AD of the body of the corpus callosum. This method successfully discriminated between HC and MS in all the selected ROIs except for the hippocampi with a high sensitivity. In particular, pNDI can discriminate between the two groups when using metrics in the NAWM. Future studies are warranted to explore whether pNODDI is more sensitive than NODDI in MS lesions and whether it is sensitive to microstructural changes also in other neurological conditions.

Acknowledgements

Data were collected as part of a study funded by the MS Society grant #77 awarded to CWK and RS. EG receives funding from TDC Technology Dedicated to Care. RS has received research funding from the UK MS Society ( grant #77), CureDRPLA and Ataxia UK & we thank the UCL-UCLH Biomedical Research Centre for ongoing support. ED and FP receive funding from H2020 Research and Innovation Action Grants Human Brain Project (#785907, SGA2 and #945539, SGA3). ED receives funding from the MNL Project “Local Neuronal Microcircuits” of the Centro Fermi (Rome, Italy). CGWK receives funding from Horizon2020 (#634541), BRC (#BRC704/CAP/CGW), MRC (#MR/S026088/1). CGWK is a shareholder in Queen Square Analytics Ltd. CT is currently being funded by a Junior Leader La Caixa Fellowship (fellowship code is LCF/BQ/PI20/11760008), awarded by “la Caixa” Foundation (ID 100010434). She has also received the 2021 Merck’s Award for the Investigation in MS, awarded by Fundación Merck Salud (Spain) and a grant awarded by the Instituto de Salud Carlos III (ISCIII), Ministerio de Ciencia e Innovación de España (PI21/01860). In 2015, she received an ECTRIMS Post-doctoral Research Fellowship and has received funding from the UK MS Society. She is a member of the Editorial Board of Neurology. She has also received honoraria from Roche and Novartis and is a steering committee member of the O’HAND trial and of the Consensus group on Follow-on DMTs. FG receives the support of a fellowship from “la Caixa” Foundation (ID 100010434). The fellowship code is “LCF/BQ/PR22/11920010”. AC is supported by the ECTRIMS post-doc fellowship (2022), previously received a UK MS Society PhD studentship (2020), a Guarantors of Brain “Entry” clinical fellowship (2019), and an ECTRIMS-MAGNIMS fellowship (2018).