Figure 1. A 27-year-old female (EDSS score = 3, top row) with MS and a 52-year-old female (EDSS score = 4, bottom row) with NMOSD. (a,f) T2-weighted image. (b,g) A 60-Hz MRE magnitude image. (c,h) Stiffness. (d,i) Storage modulus. (e,j) Loss modulus.

Figure 3. Correlation of MRE parameters and clinical data (EDSS and disease course) in MS and NMOSDs patients.

Regression plots for the normal appearing white matter region showing the relationship between logarithmically transformed serum neurofilament light (NfL) chain and measures for myelin content (myelin water fraction (A) r=-0.54, p=0.016; radial diffusivity (C) r=0.58, p=0.0004) or axon integrity (axial diffusivity (B) r=0.54, p=0.029; fractional anisotropy (D) r=-0.56, p=0.004). 95% confidence interval is shaded.

Regression plots for the relationship between logarithmically transformed serum neurofilament light chain (NfL) and volumetric imaging measures (normalized brain volume (A) r=-0.56, p=0.008; cube root lesion volume (B) r=0.59, p=0.0002; deep grey matter volume (C) r=-0.57, p=0.002; thalamus volume (D) r=-0.56, p=0.006; cortical thickness (E) r=-0.56, p=0.004). 95% confidence interval is shaded.

Figure 2: Abnormal metabolic images of mI, NAA, and mI/NAA together with clinical MRI in patients with MS. Small subcortical/juxtacortical lesions (A), which appear inconspicuous on high-resolution MRI (indicated with blue arrows), are well depicted on mI/NAA maps. In (B), red arrows indicate regions in the NAWM with increased mI only, and yellow arrows indicate regions with increased mI and decreased NAA, where no, or only diffuse changes are visible on clinical MRI. Green arrows indicate white matter lesions, where elevated mI appears beyond T2-visible pathology.

Figure 3: Examples of metabolic ratio images and clinical MRI from healthy controls and MS patients with various levels of clinical disability. mI/tCr and mI/NAA tended to be increased in the NAWM of all patient subgroups, while NAA/tCr was decreased only in subgroups of patients with mild and more severe clinical disability.

Figure 2. (A):Brain tissue volume, normalised for subject head size, was estimated with SIENAX. Final SIENAX segmentation results of whole brain (top row) and peripheral cortex masked segmentation (bottom row). (B):Partial volume segmentation of the MS lesions using T2-FLAIR and filled T1-MPRAGE structural image was segmented using FSLFAST.

Figure 3. (A)PCG neurometabolite/tCr ratio (tNAA and Glx) in HCs and RRMS fingolimod and injectables (INJ: GA+interferon) disease modifying therapies. (B) PFC neurometabolite/tCr ratio (NAA and m-Ins) in HCs and RRMS fingolimod and injectable (INJ: GA+interferon) DMTs.

Figure 3: Maps of change in CVR with immunomodulation in MS patients: (a) map in absolute units; (b) Statistical Parametric (t-score) map; (c) thresholded (p<0.05) map of null-hypothesis probability, corrected for multiple comparison using the threshold free cluster enhancement (TFCE) approach. Coordinates (in mm) refers to the MNI152 template.

Figure 2: (a) Global CVRs in GM; Left: Average and associated standard error of the mean; Right: Violin plots. (b) Change in CVR with treatment as a function of the pre-treatment CVR in GM. (** p<0.01)

Figure 2: CNN architecture: An ensemble of 10 parallel networks was used for classifying lesion patches as either CVS+, CVS−, or CVSe. With regard to the previous CVSNet implementation, a third output class for CVSe was added, and different combinations of input channels were tested to investigate their impact on the classification performance.

Figure 4: Lesion-wise performance comparison of the different models in the pure testing set for each class. Overall performance was rather similar across the models but increased slightly with the number of used input channels, except for model E, which showed comparable performance to model D despite having three additional input channels. Accuracy levels are highlighted, as this is regarded the most relevant metric with respect to a clinical application. With accuracies of 75% to 80% in the best models, the network performance is approaching levels of human inter-rater agreement.

Figure 1. Calculation of the different CNS compartments in a study subject. A) Global brain volumetry (FSL SIENAx); B) Cortical thickness (Freesurfer) and related cortical parcellation using Desikan atlas; C) Deep grey matter volume (FSL FIRST); D) Cerebellar volumes (SPM12 SUIT); and E) Cervical cord segmentation (Jim software, active surface method). Each colour represents a different compartment for each subgroup of images. Abbreviations: CTh=cortical thickness; GM=grey matter; AS=active surface; A=anterior; P=posterior; S=superior; I=inferior; L=left; R=right.

Figure 2. Bar charts showing relative importance of MRI predictors of EDSS score, and of reaching different EDSS milestone (3.0, 4.0, 6.0), in MS patients selected with random forest analyses (p<0.05). Colors reflect the magnitude of Spearman’s correlation of each predictor with EDSS score. Standardized beta coefficient from univariate logistic regression models is represented for classification models. NGMV=normalized grey matter volume; NBV=normalized brain volume; CTh=cortical thickness; CSAn=normalized cross-cross sectional area; DGM=deep grey matter

Figure 1. Immunohistological staining for myelin basic protein (MBP) and for CR3/43 (left) and parametric maps (MTsat, MWF, qT1, QSM) showing the characteristics of different histopathological MS lesions subtypes in (right): A) inactive lesion; (B) active lesion; (C) mixed active/inactive lesions and (D) remyelinated part of a lesion.

Figure 2. Quantitative MRI measures across different lesion categories and NAWM. *** p < 0,0001; ** p< 0,001, * p< 0,05.

Figure 1: Flowchart for the construction of structural connectivity networks based on tractometry. Combining the T1 parcellation, the tractogram and 3 different diffusion models, we built the connectomes weighted by different microstructural maps and we compared their topological properties.

TABLE 1: Group comparison performed with the linear robust model, where Gender, Age and Density are covariates. A Holm Post-Hoc correction was applied (i) for each network metrics of each microstructural map (Adj p-value metrics) and (ii) for each network metrics extracted from all microstructural maps of every diffusion model (Adj p-value model) to account for multiple comparison. Statistically significant results are highlighted in bold.

FIGURE 1. An exemplary illustration of generating rNOE weighted (rNOEw) images (C) using control images (A) and label images (B).

FIGURE 2. Representative rNOEw images of NC (A), NMO (B) and MS (C). (D) comparison of rNOEw signal among three cohorts of subjects (NC = 20, NMO = 15, MS = 20). Significance levels: *, p<0.05; **, p<0.01.

Figure 1. Lesion-NAWM asymmetries compared between men and women. Diffusion tensor imaging (top), high angular resolution compartment (middle), and diffusion orientation analyses indicate differences in lesion-NAWM asymmetry between women and men. *p<0.05, **p<0.01

Figure 2. Diffusion metrics reflecting significant differences in microstructure pathology between men and women. Calculation of diffusion measures took advantage of the A) average diffusion b0. Measures showing significant sex-differences in microstructure include B) radial diffusivity, C) mean diffusivity, and D) fractional anisotropy diffusion tensor measures as well as E) axonal density, F) axonal diameter, G) intracellular volume fraction and H) orientation distribution function energy.

Fig. 3. Combined structural and functional connectivity for (a) motor, (b) cognitive pathways and (c) pathway-combined SFCI evolution over 24 months of fingolimod therapy.

Fig. 1. Equations for calculating motor- and cognitive-pathway specific combined connectivity measures (Eq. 1 and 2) and SFCI (Eq. 3). SMC: motor pathway, FP: cognitive pathway, f_c: functional connectivity, TD: transverse diffusivity.

Figure 4: Mixed model results

Figure 1: Example of MS lesion with greater initial change in susceptibility and less myelin recovery at 12 months

Figure 1. Representative hippocampal segmentation, showing sagittal (left) and coronal (right) slices. The top panel highlights the segmentation, while the bottom panel shows the underlying anatomy. (light orange - DG; yellow - CA1; green - SUB; dark orange - CA2 and CA3)

Figure 2. Relationship of TP1 right SUB volume and TP1 verbal fluency. The fit line represents the full sample. (FULL= full sample; RR = relapse remitting sample; SP = secondary progressive sample)

Figure 4: Boxplots divided based on participant subtype (clinically isolated syndrome (CIS), relapsing-remitting multiple sclerosis (RR) and secondary progressive multiple sclerosis (SP)) with (+) and without (–) diffusely abnormal white matter (DAWM). Neurofilament values were normalised to matched healthy control values. Significance between DAWM– and DAWM+ are indicated with *p<0.05, **p<0.005.

Figure 3: Neurofilament (NfL) and MRI measurements (mean and (range)) for the different MS subtypes with (+) and without (–) diffusely abnormal white matter (DAWM) (CIS: clinically isolated syndrome; RRMS: relapsing-remitting multiple sclerosis; SPMS: secondary progressive multiple). Bolded values indicate a significant difference between DAWM+ and DAWM–.

Figure 4. Periventricular gradients of (A) abnormal WM volume (ml) and (B) normalized abnormal WM volume to the total band volume, which were identified as the NAWM voxel exceeding a z-score threshold of 2 (i.e. exceeding the prediction interval at 95% level of confidence of the T₁ normative linear model). Periventricular gradients in NAWM of (C) absolute z-scores and (D) absolute z-scores > 2. Error bars indicate two standard errors. Reported values in the first band should be carefully considered as they may be affected by partial-volume effects between CSF and WM tissues.

Figure 2. Representative example z-score maps overlaid onto the MP2RAGE contrast in early MS patients with low Expanded Disability Status Scale (EDSS) scores and progressive MS patients with higher EDSS.

Figure 1. Representative distributions of T₁ z-scores in example WM tracts of two patients with different Expanded Disability Status Scale (EDSS) score overlayed on the MPRAGE contrast. A more severe alteration of WM tissues was found in the patient with higher EDSS. WM tracts abbreviations: CST: cortico spinal tract; CC: corpus callosum; CT: corticothalamic pathway; ILF: inferior longitudinal fasciculus; ML: medial lemniscus; MCP: middle cerebellar peduncle.

Figure 2. Spearman correlations of WM tract-specific metrics with EDSS scores. (A) Table of WM tract names and corresponding abbreviations arranged into the five main brain pathways categories (namely association, projection, brainstem, cerebellum and commissural). (B) Tract-specific lesion loads compared to Total Lesion Volume (TLV) and Total Lesion Count (TLC). (C) Tract-specific T₁ abnormalities within lesions. (D) Tract-specific T₁ abnormalities in NAWM. (E) Tract-specific T₁ abnormalities in the whole WM (NAWM + lesions).

Relative ventricle volumes for individual patients (conversion to RRMS depicted by red). A. A substantial fraction of the CIS patients (23%) showed contractions of ventricle volumes over time, beyond the range of variation in healthy subjects (±6%). B. The majority of CIS patients did not show contractions in ventricle volume beyond the range of variation of healthy subjects. Many of these patients, both those who converted to RRMS and those who did not, showed increased ventricle volume over time, consistent with brain atrophy and neurodegeneration.

A. There was no significant difference in median EDSS between CIS patients with and without ventricle contractions >6%. The EDSS of the MS patients with contractions was similar to that of the CIS patients; EDSS of MS patients with contractions was significantly lower than that of MS patients without contractions (p=0.0063, Mann-Whitney test). B. CIS patients with ventricle contractions >6% were significantly younger than those without (p=0.0447, Mann-Whitney test). In the MS cohort, patients with ventricle contractions trended lower in age, though this is not significant.

Figure 2. Hub disruption over 2 years in patients (yellow) and HC (purple). Header of the each plot indicates the visit month. Each imaging metric is shown from up to down as DTI, rs-fMRI and MEG respectively. The mean connectivity of each node of each metric in the group of controls at baseline (x axis, ⟨Healthy at Baseline⟩) is plotted versus the difference between groups in mean connectivity of each node of each metric in each group at each visit and mean connectivity of each node of each metric in the group of controls at baseline (y axis, ⟨Groups at each visit – Healthy at Baseline⟩).

Figure 3. The correlation between individual hub disruption of each imaging metric and PASAT performance over 2 years in MS. Header of the each plotshows the visit month (0: baseline, 12: 1-year, 24: 2-year). Orange, green and blue indicate the structural, rs-fMRI and MEG hub disruption respectively. Slope of each imaging metric (x axis) is the ratio of the mean connectivity of each node in the group of controls at baseline to the the difference in connectivity of each node of each subject at each visit and mean connectivity of each node in the group of controls at baseline.

Change in grey matter BBB permeability from baseline to six months for subjects with and without (NEDA) disease activity within two years. Mean difference –0.038 ml/100g/min, 95% confidence interval -0.073;-0.004, p = 0.03.

Change in grey matter BBB permeability from baseline after treatment as measured by Ki for subjects with and without (NEDA) disease activity within two years. Dashed lines = individual trajectories, solid line = group means, error bars = means ± standard error of the mean.

Fig. 1. DWIs of b=9380 s/mm² of (col I) shiverer and (col II) WT mice spinal cords, and signal-b curves of (c) anterior-dorsal and (f) posterior-dorsal column, and (i) lateral white-matter tracts up to b = 42,890 s/mm² for UHb-rDWI and 2210 s/mm² for UHb-aDWI. Signals were averaged over two nearby slices of two shiverer (red) and two adjacent slices of six WT (black) mouse spinal cords.

Fig. 2. EM images of dorsal (sensory) column of (a, c) shiverer and (b, d) WT mice spinal cords in two different magnifications (a, b) x1,100 and (c, d) x11,000. In this EM picture, the myelin sheaths of the shiverer mouse is less numbered and looser than that of the wild-type mouse.

Figure 2. Healthy subject results: Upper: Bar and error plots, and Under: Table of U-fiber labels.

Figure 1. Processing steps realized to obtain the U-fiber diffusivity maps (FA, MD, RD, and AD). (i) Segmentation of Diffusivity maps, using DSI Studio software. (ii) Coregistration between diffusivity maps and T1w-3D image. (iii) Normalization of T1w-3D image to the MNI space. (iv) The deformation matrix of the Normalized T1w-3D image was applied to the U-fiber masks in order to adapt to the diffusivity maps (Inverse normalization). (v) Application of the LNAO-SWM79 Atlas mask to each patient’s diffusivity map.

Figure 1: Representative views of MTR and ihMTR template maps from the healthy control population.

Figure 3: FWE-corrected (1-p)-value maps of the voxel-based analysis in the MNI space between MS and HC subjects for ihMTR (yellow) and MTR (blue). (1-p)-value maps are shown in a range spanning from 0.95 to 1.00.

Figure 1. A) Prescription of the 3D SHINKEI and MTR sequences for imaging the lumbar plexus in the coronal plane: B) Example image obtained using the 3D SHINKEI at the level of the lumbar plexus (L2-L4 segments shown); C) Manual segmentation of the lumbar segments with separate binary masks created for the preganglionic, ganglionic and postganglionic regions shown in blue, red, yellow, respectively.

Figure 2. A) Prescription of the high-resolution fat-suppressed T2-weighted and MTR sequences for imaging the sciatic nerve in the axial plane; B) Example image obtained using the high-resolution fat-suppressed T2-weighted sequence; C) Manual segmentation of the sciatic nerve (binary mask shown in yellow).

Fig. 2: FLAIR images and DL segmentation from an MS patient in the CombiRx study over 60 months. Good delineation is obtained for brain tissue, lesions (red), and DAWM (blue). Part of DAWM can be clearly seen to evolve into focal lesions (red arrowhead), while other parts of DAWM (blue arrowhead) persisted for 60 months.

Fig. 1. DAWM segmentation using CNN. A U-net is trained for brain segmentation (GM, WM, CSF, T2L) from multimodal MR images. The output tissue scores are used in histogram and connectivity analysis to obtain DAWM segmentation.

Table 1. Intra-thalamic magnetic susceptibility (ppm) in patients and controls.

Table 2. Intra-thalamic iron content (mg) in patients and controls.

Comparison of compartment Axial Diffusivity (cAD) (A), compartment Radial Diffusivity (cRD) (B), compartment Fractional Anisotropy (cFA) (C), ompartment Mean Diffusivity (cMD) (D), heterogeneity index (HEI) (E), and free water fraction (F) between MS lesions and Contralateral NAWM in patients. Significance levels for Mann-Whitney U tests are displayed as *P<0.05, **P≤0.01, ***P≤0.001.

Lesion and contralateral normal appearing white matter (NAWM) segmentation in T2 weighted FLAIR acquisitions. Hyperintense white matter lesions (A) were manually segmented (B). A linear transform was used to mirror lesions in the contralateral side of the brain, and a white matter mask was used to refine segmentations (C,D).

Fluid attenuated IR images of a MS patient. Left/right: Sagittally acquired vs coronally reformatted images. Top panels: FLAIR. Bottom panels: True-T2 DIR of similar image planes. Multiple focal lesions are seen in the images, situated in deep WM as well as GM-associated.

Simulated True-T2 DIR sequence with a T2 preparation phase. Figure elements as in fig. 1; in addition an information box showing the simulator settings is included in the left panel. In the right panel is shown an information box showing signal and contrast strength and efficiency, that is, relative SNR/CNR per unit of acquisition time. The percentages shown refer to the corresponding ones in fig. 1.

T2-weighted imaging (A, E), FLAIR (B, F), T1radiab (C, G) and TRAFF2 (D, H) imaging findings (white arrow) in a 47-year-old female with disease progression, EDSS score at baseline of 2.5 and 1-year follow up of 3.0, (A, B, C, D) and 46-year-old female with no change in her baseline EDSS score of 2.5 at 1-year follow up (E, F, G, H).

Potential of adiabatic T1rho and Relaxation Along a Fictitious Field imaging (TRAFF2) for prediction of Expanded Disability Status Scale (EDSS) and Multiple Sclerosis Severity Score (MSSS), correcting for age of patients. The values are R2, with statistically significant correlations after Bonferroni correction over 12 evaluations marked as (raw p-value threshold): * 0.05 (p<4.17x10-3), ** 0.01 (p<8.33 x10-4) and *** 0.001 (p<8.33 x10-5). AUC = Area Under Receiver Operating Characteristic Curve; 95% CI = 95% Confidence Interval.

Figure 3. Comparison of FLAIR-based lesion maps and ICVF-based relevance maps for three correctly classified MS patients (in rows). Red shows the brain areas with lesions in the lesion maps and positive relevant voxels in relevance maps.

Figure 1. ICVF images and corresponding relevance maps overlaid on the corresponding ICVF images for two correctly classified HC and two correctly classified MS subjects. Red shows the positive relevance of the voxel information for correct classification and blue indicates voxels possessing a negative relevance.

Figure 1: Post mortem images showing an exemplary high-inflammatory PRL (a.) and a low-inflammatory PRL (b). (top) Double immunohistochemistry of myelin basic-protein (MBP) (brown) with CR3/43 (MHC II, (blue) in an exemplary highly inflamed PRL (box, a.) and low-inflammatory PRL (box, b.) On the bottom, 3D EPI, QSM derived from 3D EPI and qT1 derived from MP2RAGE.

Figure 2. Joint plot showing the relationship between qT1 and diffusion parameters estimated from multi-shell diffusion MRI. Left: Voxel-wise joint plot with kernel density estimate (KDE) Right: Lesion-wise plot showing the relationship between qT1 and diffusion parameters using a scatter plot. First row: stick fraction, representing the intra-cellular component. Second row: ball fraction, representing the extra-cellular unrestricted component. Third row: sphere fraction, representing the extra-cellular restricted component.

Figure 1. Example Illustration of the proposed anatomical coordinates with one slice from the axial direction.

Figure 2. The computed anatomical coordinates are concatenated with the incoming feature tensor through channel dimension, followed by standard convolution, batch normalization, and ReLU activation for feature fusion. c is the number of channels in the feature tensor.

Figure 1. Postmortem T2-MRI and calculated maps. The shown are T2 (A), T2-Contrast (B), T2-Dissimilarity (C), T2-Entropy (D): all from average GLCM, and T2-entropy filter (E). Different colors demonstrate different lesion types: red: WML, green: remyelinated WML, blue: type IV cortical lesion, yellow: type III cortical lesion.

Figure 4) In-vivo lesion severity. 25th percentile of contrast and both contrast and dissimilarity were used to detect the less severe (potentially remyelinated) lesions. And 75th percentile of the same parameters were applied to have more severe lesions (highly demyelinated and axonal damaged). We found that most of the patients have both types of lesions and the percentage of the less severe lesions is in the range of reported remyelination percentage in MS.

Fig. 1. Longitudinal MRI of brain lesions at baseline, 4-week cuprizone, and 10-day withdrawal (n=12). Lesions are observed in the splenium, genu, and cerebellar nuclei in T₂-weighted and myelin water fraction images. Example histology of the same animals are shown. Inserts show lesion areas at 5× magnification. GFAP~astrocytes, Iba1~microglia, dMBP~denatured myelin basic protein, and solochrome~myelin.

Fig. 2. Bar plots for longitudinal MRI. Error bars show standard error about the mean. Sample size: n=12 at baseline, 4-week cuprizone, and 10-day withdrawal. MWF=myelin water fraction, ADC=apparent diffusion coefficient, and FA=fractional anisotropy. One-way ANOVA and post-hoc t-tests were performed between baseline control, 4-week cuprizone, and 10-day withdrawal groups (*p<0.05, **p<0.01).

Figure 1: Representative 3D FLAIR* and Phase images from one PMS patient showing the same PRL a) at baseline and b) at the 13 months post-DMT follow-up.

Figure 3: Summary of the lesion-wise results for each patient with paramagnetic rim lesions (n=11 of 13 total patients studied, 85%). a) Number of rims detected in session 1 (baseline) and session 2 (follow-up) for each patient, overlaid on the total number of rims adjudicated in the consensus manual reading. b) Number of matching predictions between baseline and follow-up. Abbreviations: Consistency_p, probability-based consistency; Consistency_b, binary-based consistency.

Figure 1. The flowchart of the FnLRT algorithm. The artifacts in the features are substantially reduced.

Figure 3. The MWF maps of 5 representative subjects obtained from the L+S, LRT and FnLRT reconstructions (R = 6).

The myelin water fraction (MWF) a) mean, b) standard deviation, and c) coefficient of variation (or myelin heterogeneity index, MHI) of the normal appearing white matter in healthy controls (HC), relapsing remitting MS (RRMS), primary progressive MS (PPMS), and secondary progressive MS (SPMS). A one-way ANOVA test with Tukey correction showed that the MHI of SPMS was higher than HC (p=0.01) and trended towards higher than RRMS (p=0.065).

The myelin heterogeneity index (MHI = MWF standard deviation/MWF mean) of all MS patients was correlated with Expanded Disability Status Scale (EDSS) (p = 0.046, R = 0.3) using a Pearson correlation. Shaded area is 95% confidence interval.

Figure 1: Sketch of the sub-segmentation processing of the lesion into core and edge regions based on the T₁w contrast and from the manually drawn mask on the FLAIR image (a), and illustrative depiction of the evolution of MR contrasts over time in an active lesion (b).

Figure 3: Example of an RVtoC curve adjustment over an ihMTR lesion in the core region (a), and boxplots of the average recovery rates per patients of ihMTR, MTR, RD and MPRAGE (b). Note that patient P01 was not included as the dynamic of its two lesions could not be resolved by the fitting model for both ihMTR and RD (ρ²<0.75).

Fig. 1. MS1: (a) DWIs of b = 1700 ~ 7350 s/mm² of the lesion slice and (b) ROIs at lesion and NAWM slice, 1.5 cm above the lesion toward the brain, and (c) signal-b curves of the recently-active lesion (red square) and NAWM (red circle), and averaged data (hollow square) at the corticospinal tract of seven healthy subjects, and (d) mean values of high-b diffusion coefficient D_H at the lesion, NAWM, and the healthy CSCs at the corticospinal tract. Green dotted line in (b) indicates the MCS data with 30 % demyelination in lesion.

Fig. 2. MS2: (a) T2WI, (b, c) DWIs of b = 1700 ~ 7350 s/mm² of the lesion slice and NAWM slice, 1.5 cm below the lesion away from the brain, (d) ROIs on lesion (top) and NAWM (bottom), (e) signal-b curves of the recently-active lesion, NAWM, and averaged data at the posterior column of 7 HS, and (f) mean values of high-b diffusion coefficient D_H at the lesion, NAWM, and the healthy CSCs at the corticospinal tract. White vertical arrows in (a) indicate lesion with rapidly decaying signal. Green dotted lines in (e) indicate the MCS data with 40 % and 15 % demyelination in lesion and NAWM, respectively.

Figure 3 Selective clinical MP-RAGE (first column), T2-FLAIR (second column) and STAIR-dUTE (last three columns) images of two representative patients with MS (first row: a 49-year-old female; second row: a 69-year-old female). MS lesions appeared hypointense on the MP-RAGE image and hyperintense on the T2-FLAIR image as indicated by the yellow arrows. These lesions also show signal loss on the magnitude images in the first echo images (third column), magnitude echo subtracted images (fourth column) and complex echo subtracted images (last column) using the STAIR-dUTE sequence.

Figure 2 A volunteer study showing the generation of ultrashort T2 signals with the methods used in this study. The first row shows magnitude images for the first echo (A) and the second echo (B), as well as the corresponding phase images for the first echo (C) and second echo (D). Panel E shows the ΔB0 field map used for the complex ES. Magnitude of the first echo (F), magnitude echo subtracted (G), and complex echo subtracted (H) images obtained with the STAIR-dUTE sequence are shown. The magnitude of the first echo image (A) is displayed again in (F) for closer comparison with (G) and (H).

Figure 4: Percentage change over 2 years in mean myelin water fraction (MWF) for each participant divided into participant subtype (healthy control (HC), relapsing remitting multiple sclerosis (RRMS), progressive multiple sclerosis (PMS)). The last bar (in red) is the mean over the participants.

Figure 2: Mean myelin water fraction (±standard deviation) at baseline (B) and follow-up (FU), absolute change in MWF over 2 years, percent MWF change over 2 years (%) and annual rate of MWF percent change divided into participant subtype (healthy control (HC), relapsing remitting multiple sclerosis (RRMS), progressive multiple sclerosis (PMS)). Volumes are taken at baseline. Significant differences between baseline and follow-up are denoted in pink with *p<0.01, ***p<0.0001.

Figure 1: Representative case showing the UNET’s prediction against the ANN’s prediction

Figure 2: RMSE for UNET and ANN Predictions

Figure 4: Example of T1 and 3DFLAIR images, F, NDI and T2* maps in a patient. F and NDI are low in the white matter lesions while T2* values are increased.

Figure 5: Histogram of T2* distribution in the normal-appearing white matter and white matter lesions. Abbreviations: WM: white matter

Figure 3. The mean percent overlap of clusters with anatomical regions labeled in the N27 Atlas, by MS subtype. The percent overlap is defined as the number of voxels within a cluster that overlap with an anatomical region, divided by the total number of number of voxels in the cluster. Error bars represent the 95% confidence interval, calculated only for regions with more than one cluster.

Figure 2. Montages of clusters identified by thresholding the mean difference in SWI intensity between each MS cohort (BMS, RRMS, SPMS, PPMS) and the control group. The clusters are overlain on to the N27 atlas. The colourmap corresponds to the square root of the mean sum of squares due to treatment effect. A significance threshold of p=0.001 was used.

Figure 1. Example Illustration of how to compute lesion offsets and voxel weight based on ground-truth lesion labels. Individual lesions are marked by masks of different colors.

Figure 2. Qualitative comparison between baseline methods and our proposed MS-Voter.

Figure 3 CSFF with age in four cortical lobes. It clearly shows that the CSFF increases with age in all cortical lobes (Frontal: p=0.019, occipital: p<0.01, Parietal: p<0.01, temporal: p<0.01, all p-values are FDR adjusted).

Figure 2. WM CSFF and PVS score relation. PVS score was evaluated on the T2w image at the semiovale slice and the WM CSFF was measured at the WM region of the same slice. It shows CSFF increases as the PVS score increases.

Fig 2 Results for the Neurite Density Index NDI (i.e. the NODDI-derived intra-cellular fractional volume), using 6 different acquisition protocols. In a and b, NDI mean values and standard deviations obtained in different conditions of demyelination and axonal loss, respectively, considering a Rician noise with SNR=20 affects the synthetic DW signal. For reference, some specific combinations of the true intra-axonal volume fraction AVF and true intra-myelin volume fraction MVF are also shown.

Metrics AD, MD, RD, and FA are DTI-derived; AK, MK, RK, and KFA are DKI-derived, while neurite density index NDI, fractional volume of the isotropic compartment ν_iso, fractional volume ν_ic of the intra-cellular space, and orientation dispersion index ODI are NODDI-derived. Omitted results (-) refer to parameters showing non-significant changes between healthy and damaged condition (one-way ANOVA with p<0.01). NC means Not Calculated, since not all the sequences were used for all the DWI analyses.

Figure 3: Image series showing the correlation between MRI results, photographs and histology for all samples. a-d) High-resolution HYFI images, e-h) amplitude maps of the uT₂-S fit component, i-l) sample photographs, m-p) corresponding 10 µm cryosections with myelin staining using immunohistochemistry for MOG (brown) and haematoxylin counterstaining (purple). Note that sample 2 contains two tissue pieces (originally connected), but histology was only performed on the larger one.

Figure 1: Time evolution of signal intensity in SPI images of human MS brain tissue after D₂O exchange. Six of the fourteen images acquired for the SPI series of sample 1 are shown. The signal intensity drops as TE increases, and has all but vanished after 413 μs – a clear indication that the captured signal stems from tissues with T₂ times of around a hundred microseconds at most.

Fig.1 A schematic diagram of the QPM-myelin mapping (R₁·R₂^*) and MWF images procedures.

Fig. 5 Plots the relationship between MWF values and R₁·R₂^* product values derived from QPM across all four bilateral structures.

Figure 3. Representative quantitative maps of an MS patient. Plaques were defined as a white matter area of more than 5 mm in diameter, with abnormally high intensity on synthetic FLAIR images. Each plaque was manually segmented. All the ROIs placed on synthetic FLAIR images were copied and pasted onto the quantitative maps.

Figure 4. Simple linear regressions to assess the association between estimated time from the onset of plaques and quantitative MRI metrics. Plaque MVF and g-ratio in relation to the estimated time from the onset were significantly fitted to linear lines with negative and positive slopes, respectively. Plaque T1 and T2 in relation to the estimated time from the onset were significantly fitted to linear lines with positive slopes. NAWM T2 in relation to the estimated time from the onset was significantly fitted to a linear lineswith a negative slope.