Dynamic representation of the blood, GM and CSF fractions of the ASL signal as calculated with the previous parameter maps at arbitrary time points. The labeling duration is 3s and the PLD is varied from 0 to 5s.

Parametric maps obtained from voxel-wise fitting of the (smoothed) signal. The CBF (named f in the dynamic model) and the ATT (or δ) are shown in A and B. C shows the novel parameter Tbl->CSF, the exchange time between the blood and CSF compartments. In D a close up of the circle of Willis (top left) and a higher slice in the ATT map are given. The bottom row of D present two slices in the Tbl->CSF map which show low values in the choroid plexus (circled). Tbl->GM is not shown. The lowest slice of the imaging volume crossed with the labeling plane which explains the discrepancy in values.

Figure 1. Temporal curves of the ICAs, IJVs and CSF flow rates of a healthy volunteer in the paced (A, B) and deep (C, D) conditions. The whole duration (60 s) is shown in A and C, a sub-portion (seconds from 30 to 45) in B and D. Legend: ICA= internal carotid artery; IJV= internal jugular vein; CSF=cerebrospinal fluid.

Figure 2. Amplitude spectra of internal carotid arteries (ICAs), internal jugular veins (IJVs), cerebrospinal fluid (CSF) flow rates (in (ml/s)2/Hz), separately for paced (A) and deep (B) breathing. The lowest peak corresponds to the breathing frequency (BR) of the subject, and increased its amplitude with deep breathing. Conversely, the heart rate (HR) peak and its second (2*HR) and third (3*HR) harmonics decreased their amplitudes with deep breathing.

a) stenosis in the upper segment of the right IJV (white arrow). b) the presence of robust posterior and lateral condylar and deep cervical veins (double white arrows).

The curves represent overall neurofluid dynamics during cardiac cycle (CC). Cerebral arterial flow (red) (carotid and vertebral), IJV I(blue) was corrected (dotted line) to equal cerebral arterial flow. Intracranial blood volume change during CC (yellow), intracranial CSF volume change (green). Note how CSF responds to the intracranial blood pulsation. Nevertheless, CSF volume doesn’t fully compensate the blood volume and a net intracranial volume change is seen (pink).

Figure 3: CSF inflow amplitude depends on visual stimulus frequency. The amplitude of the cortical BOLD (left panel) and CSF flow (right panel) responses to the visual stimulus each depend on checkerboard flicker frequency, with the largest responses to 12 Hz and 40 Hz. Shading indicates standard error; arrow indicates the timing of the CSF inflow event.

Figure 1: CSF flow is locked to the visually-evoked BOLD cortical signal (N=6). A) Example positioning of the functional image. B) Example smoothed cortical (green) and CSF (purple) traces from a single subject. Black bars indicate when the stimulus was on. The signals show consistent anticorrelated responses, with upwards CSF flow when the BOLD signal declines. C) The mean neural, hemodynamic and CSF signals across subjects shows that neural activity precedes hemodynamic and CSF responses.

Figure 3. (A) CSF flow as stroke volume exchanged during one cardiac phase (systole or diastole) in upright and supine postures. The positive values indicate cranio-caudal direction while negative values indicate caudo-cranial direction. (B)CSF peak velocity measured during diastole and systole in supine and upright. Notice the difference in CSF peak velocity observed between supine and upright in diastole (****=p<0.0001) but are not significant in systole (ns=not significant).

Figure 2. CSF flow rate at each time point over one cardiac cycle of a volunteer in the upright and supine posture at mid-C2. Note that the positive systolic peak (second half of the cardiac cycle) of the upright posture is only slightly reduced relative to the supine posture. Conversely, that of the negative upright diastolic peak (first half of the cardiac cycle) is reduced much more compared to the supine diastolic peak.

Figure 4. Typical results from human scans on 3T from regions around the dural sinuses (DS) where lymphatic vessels were identified in previous studies. (a) FLAIR overlaid with ΔS/S from the ROI; (b) The proposed 3D TSE image overlaid with ΔS/S; (c) Average time course from 3D TSE from the ROI, and four parametric maps extracted from the time course: onset time (T_onset), time to peak (T_peak), |ΔS/S|, and [Gd]. The vertical dashed line indicates the time when Gd is injected.

Figure 1. Simulation results for the proposed 3D TSE sequences: (a-c) 3T, T1-dominant; (d-f) 3T, T2-dominant; (g-i) 7T, T1-dominant; (j-l) 7T, T2-dominant. The fractional MR signals (Mz/M0) are displayed as functions of shot number (SN), TE and TR. Signals of gray matter (GM), white matter (WM) and blood are most suppressed (lines close to zero). The contrast is defined as (“CSF with Gd” – CSF).

Figure 3. Scatterplots of PI measures in the LSA and ePVS score in the BG (A) and CSO (B). Least square linear regression lines are added for visualization. ePVS = enlarged perivascular space, LSA = lenticulostriate arteries, BG = Basal Ganglia, CSO = Centrum Semiovale. ePVS count is scored as follows: 0=<10 ePVS, 1=10-25 ePVS, 2=>25-40 ePVS, 3=>40 ePVS.

Table 2. Spearman’s rho correlations between between ePVS count of the Basal Ganglia and Centrum Semiovale and PC-MRI measures. Significant associations are depicted in bold, ( * p < 0.05, ** p < 0.01). PC-MRI = phase contrast magnetic resonance imaging, ePVS = enlarged perivascular space, LSA = lenticulostriate arteries, MCA = middle cerebral artery, r_s = Spearman’s rho, v = blood flow velocity, q = volumetric flow rate.

Figure 2. 3D dynamic contrast-enhanced MRI at the level of the eye and the optic nerve (ON) before (A) and at peak intensity (B) after Gd infusion into the subarachnoid space (SAS) of the lumbar spine. (C) Maximum intensity projection (MIP) after image segmentation of the eye and ON in (B). Arrows indicate the regions of interest for quantitative analysis. (ONSAS: optic nerve subarachnoid space; ONP: optic nerve parenchyma; P: posterior; A: anterior; MT: muscle tissue; OB: olfactory bulb)

Figure 3: (Left column) Temporal evolution of signal enhancement in the posterior optic nerve subarachnoid space (ONSAS-P), anterior ONSAS (ONSAS-A), posterior ON parenchyma (ONP-P), anterior ON parenchyma (ONP-A), muscle tissue (MT), and olfactory bulb (OB) before and after Gd contrast infusion into SAS of the lumbar spine at timepoint zero. (Right column) Relative contrast uptake in different regions of interest using area under curve (AUC). **p<0.01, ***p<0.001, ****p<0.0001.

Figure 1. Tract-based statistical images (coloured regions) showing: A) Increased MD in WM regions linked to cognitive function in CAD patients at baseline compared to age-matched HC subjects; Widespread increases in B) MD and C) RD throughout WM in older HC subjects relative to younger healthy adults; and D) Widespread improvements in WM integrity given by increased FA, decreased MD, and decreased RD in CAD patients following CR. Statistical images are thresholded by p < 0.05 (FWE-corrected).

Figure 2. Group mean MD values in coloured clusters shown in Fig. 1A. Compared to younger controls, older controls have a 14.9% increase in MD. Increased MD is observed in CAD patients at baseline compared to controls with an 8.9% recovery in MD following CR. *p < 0.05 (FWE-corrected)

Table 3. Clinical group differences in D_par, D_int and f_int as calculated from M, P, S and trace images. CON = controls, MCI = mild cognitive impairment, AD = Alzheimer’s disease, NAWM = normal appearing white matter, CC = Corpus Callosum, D_int = interstitial fluid diffusion, D_par = parenchymal diffusion, f_int = interstitial fluid fraction. β represents standardized beta coefficients. Only significant correlations (i.e., p<0.05) are reported. Analyses were adjusted for age, sex, absolute and relative sum of squares.

Figure 1. A graphic display of the primary imaging directions, M (measurement direction: left-right, in red), P (phase-encoding direction: anterior-posterior, in purple) and S (selection direction: superior-inferior, in blue).

Figure 5. Schemes of waste clearance. A: Schematic of waste clearance in body tissue outside the brain. B: Schematic of CWC illustrating the participation of both the venous system and CSF system; The main difference between CWC and body waste clearance is the extra layer of CSF (red dashed line in B). C: Illustration of the one-way transfer of cerebral waster from the brain parenchyma to capillaries or venules as well as much smaller diameter of the para-venous pathway comparing to the corresponding venous pathway, which makes it less effective in waste clearance anatomically.

Figure 4. Participation of the parenchymal venous pathway in CWC. SWI signal intensity changes post-SPIO CSF tracer infusion in normal rats. A-B: Quantitative MRI signal intensity changes pre-/post-low dose 75μg (A) and high dose 270 μg (B) FeREX (100nm) in azp and azicv; demonstrating CSF tracer presence in azicv, but not azp, at low dose. C-D: Quantitative MRI signal intensity changes pre-/post-low dose 75μg (C) and high dose 240 μg (D) Ferumoxytol (21nm) in azp and azicv; demonstrating CSF tracer presence in azicv, but not azp, at low dose.

Fig.2. Region of interest (ROI) placement for the manual-based ALPS index

A 5-mm-diameter spherical ROI was placed in the projection and association areas.

Fig.3. The Box plot of the ALPS index

The ALPS index in the HT group was significantly lower than that in the control group (p value < 0.05).

Figure 2: (A) CBF is, overall, inversely associated with age in both gray matter and white matter. (B) ATT is directly associated with age in both gray matter and white matter. However, these relationships may be different when comparing early to late middle-age adults to older adults.

Figure 3: (A) The ratio of gray-to-white matter CBF generally decreases with age until the fourth quartile in which there is more variability. (B) The ratio of gray-to-white matter ATT also increases with age and also exhibits greater variability in the fourth quartile.

Figure 3. MPRAGE and perfusion map from one healthy subject (a) and NPH patient (b). CP was indicated by red arrows. Enlarged ventricle as well as choroid plexus hyperplasia can be observed in the NPH patient. (c) and (d) show scatter plot of average perfusion signal and fitted curves in GM and CP averaged across four healthy subjects and in the NPH patient, respectively. A delayed peak with comparable signal intensity was observed in CP as compared to GM in healthy subjects, while overall CP flow signal was relatively higher than GM in the NPH patient.

Figure 2. A center slice of perfusion image acquired at fifteen PLDs through direct subtraction (a) and KWIA filtering (b). CP was indicated by red arrows.

Figure 1. (a) ΔADC and (b) ADCmean in the positive group before and after the lumbar tap, and each of representative images.

Figure 3. (a) ΔADC and (b) ADCmean in the positive group before and after the shunt surgery, and each of representative images.

Figure 3. Axial T1W images (top row) and segmented brain using freesurfer (bottom row) used for ventricular volume evaluation in iNPH patient

Figure 2. Correlation analysis between a reduction in ventricular volume and changes in iNPHGS scores in cognitive domain showing a moderate degree of a significant negative correlation (r = -0.408, p = 0.023).

Figure 1. Study overview. From left to right, 817 ex-vivo participants were preprocessed and their corresponding EPVS masks were generated using a deep learning segmentation model. EPVS were quantified in the whole brain and multiple ROIs to investigate the associations of EPVS with neuropathologies, cognition, demographics and risk factors. This study found EPVS to be associated with gross and microscopic infarcts, cerebral amyloid angiopathy, cognitive decline in visuospatial ability, and with diabetes.

Figure 5. Neuropathologic, demographic, and risk factor associations of quantified EPVS in each of the brain ROIs evaluated in this study. Abbreviations and format shown at the bottom of the figure.

Figure 4. Illustration showing that all MRI-visible vasculature were surrounded by CSF. A) sT2W image with bright CSF and dark blood; B) T1W with bright blood signal but dark CSF signal; C) CSF-only images, acquired by T2-TSE sequence with TE = 345ms at 3T; D) a 3D rendering of the CSF-only images. A) and B) were results from the high-resolution STAGE scans. The enlarged image in C) shows several MRI-visible vasculature (dark curves) surrounded by bright CSF. Images were generated from the second experiment as described in the method section.

Figure 1. CSF-filled peri-neural spaces surrounding the first five cranial nerves (CN I-CN V). A) Olfactory nerves (CN I); B) Optic nerves (CN II); C) Oculomotor nerves (CN III); D) Right trochlear nerve (CN IV); E) Left CN IV; and F) Trigeminal nerves (CN V). Space within the yellow lines: respective nerve (hypointense signal). Space between the red and yellow lines: peri-neural space (hyperintense signal). Space between the blue lines: CN V (hypointense signal). All images were generated from the first experiment as described in the method section.

Fig. 4. Relationship between PVS volume fraction and PSQI score in mTBI. PVS volume fraction positively correlated with PSQI score in TBIPCE subgroup (tbiANDpce in blue, p=0.0026, r2= 0.13, blue), but not in TBIonly (tbionly in red) subgroup.

Fig. 3. Group comparisons of PVS volume fraction. TBIPCE subgroup had significantly higher PVS volume fraction than controls (F=6.0, 0.296% vs 0.273%, *corrected p=0.03, effect size: ω2 = 0.08 and η2 =0.1, calculated by setting the confidence coefficient equal to 10%, which is appropriate if the alpha for the F test is .05))

Figure 2. Plots of CSF water fraction (CSFF) change with age in the brain cortex and white matter of healthy volunteers (n=20). Statistical significance was found in the frontal (p=0.0011) and temporal (p=0.0093) cortex after correction for multiple comparisons.

Figure 1. Example of quantitative maps of myelin water fraction (MWF), intra/extracellular water fraction (IEWF), and CSF water fraction (CSFF) obtained by FAST-T2 sequence in a 50-year-old female subject.

Figure 1: (A): White matter hyperintense (WMH) regions are automatically segmented on FLAIR images from baseline and follow-up and named as WMHmask1 and WMHmask2. Blue arrow shows the normal appearing white matter (NAWM) that progressed to WMH in 4-year follow up. (B): The intersection of WMHMask1 & WMHMask2 is referred as “Old WMH” at baseline and the subtraction of WMHMask1 from WMHMask2 [(WMHMask2-WMHMask1)>0] is referred as “Newly Grown WMH” at 4-year follow up (which was NAWM at baseline).

Figure 4: The relative values of diffusion, perfusion, and T1 signal intensities among the three tissue types is displayed. The bars are displayed such that the “Healthy WM” is on the left-end, the “Old WMH” is displayed on the right-end, with the “Newly Grown WMH” displayed proportionally along the bar. It can be seen that the characteristics of the “Newly Grown WMH” are closer to the healthy tissue in all structural/anatomical parameters, but their CBF resembles more to the “Old WMH” tissues.

Figure 1 (A) Extracted ventricles with automatically placed oblique axial and coronal reference planes; (B) using the oblique slice from (A), the ventricle walls (black) are sliced, candidate points for the callosal wall are identified (yellow). Refiled points (red) indicate those included in the final calculation. Lines are then fit through the points on each side of the vertex and the angle is calculated; (C) reports the correlation between manual and automated measurements.

Figure 2 Histogram of 4,980 automatic callosal angle measurements. The data is skewed towards lower angles. Thresholds for suspected normal pressure hydrocephalus are indicated for several publications.

Figure 2. Summary density plots of all participants (n=19) and clusters (n=331) showing distribution differences between layers 1-4 and WMHs (top) and cortical tissue (bottom). (A) Distribution of CBF values (B) Distribution of FW values

Figure 4. Results of linear mixed model comparing the mean CBF values of (A) WMHs and layers 1-4 to normal-appearing white matter, indicated in red and (B) layers 1-4 to WMHs, indicated in red.

***p<0.001, *p<0.01

Figure 1. Representation WMHs lesions on T2-FLAIR images in mild, moderate and severe lesion groups and its corresponding lesion segmentation result (delineated in red).

Figure 2. A) Representation of CVR maps in WMHs patients. B) Correlation analysis between lesion volume and white matter CVR. C) Group-wise comparison of white matter CVR between mild, moderate and severe lesion groups.

Fig. 3. The gBOLD-CSF coupling is correlated with cortical Aβ and cognitive decline. The age-&gender-adjusted gBOLD-CSF coupling strength decreased with increasing cortical Aβ SUVRs at baseline (A) but not their changes in the following two years (B). This gBOLD-CSF coupling is significantly correlated with the 2-year MMSE score changes (C) but not with its baseline (D). The linear regression lines were estimated based on the linear least-squares fitting. Each dot represents a single session.

Fig. 2. The dependency of the gBOLD-CSF coupling on AD risk factors and disease conditions. The strength of the gBOLD-CSF coupling (global BOLD-CSF correlation at +3 sec lag) decreased with age (A; Spearman’s r = 0.24, the linear mixed model with Satterthwaite's method¹⁷) and in female participants (B). This gBOLD-CSF coupling (age-&gender-adjusted) also decreased with AD development (C) and with the APOE 𝛆4 allele number carrying (D). Error bar: SEM; session number showed in each bar.

Figure 1: Mean CBF (ml/100g/min) for females across decades: A- 20 to 29 yo; B- 50 to 59 yo; C- 60 to 69 yo; D – 70 to 79 yo. The graph demonstrates the relationship between age and CBF.

Figure 2: Mean CVR (ml/100g/min/ΔmmHg CO2) for females across decades: A - 20 to 29 yo; B- 50 to 59 yo; C- 60 to 69 yo; D – 70 to 79 yo. The graph demonstrates the relationship between age and CVR.

Figure 4 Significantly smaller choroid plexus volume in healthy volunteers compared to the MCI and AD patients

Figure 2 Choroid plexus volume changes with age in healthy volunteers

Fig.1. The obtained fronto-pontine tract are overlaid on a FLAIR image with corresponding group-averaged left (b) and right (c) tract-profiles. (a) Yellow binary masks: WMH segmentations. The tracts are passing through WMH. (b,c) The AFD tract profiles (upper) of higher WMH (blue) and MCI (red) groups are statistically significant compared to those of lower WMH group (dark green) at the location of WMH in WMH tract profiles (lower, light green: lower WMH, light blue: higher WMH, magenta: MCI groups). Error bars: S.E.M. Asterisk marks: p < 0.05 (blue: higher WMH, pink: MCI groups).

Fig. 3. The comparisons of the amount of WMH (a), mean AFD (b), and connectivity (c) on tractometry between lower WMH, higher WMH and MCI groups. Differences between groups was statistically significant for each comparison across all groups. (p < 0.0001). The central marks of box plots indicate the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to the most extreme data points not considered outliers, and the outliers are plotted individually using the '+' symbol

CSF flows in the aqueduct and spinal spaces.

CSF flows continously acquired by EPI PC at the spinal level.

Base on the respiratory signal, EPI PC signal can be post processed. CSF flow dynamic is reconstructed along cardiac cyle during inspiration (red) and expiration (blue) periods. CSF flows oscillate around the zero line during each crdiac cycle.

Figure 1. The architecture of the proposed unsupervised learning segmentation method UPSS. UPSS is composed of mainly two parts: a Frangi filter as convolution neural network (CNN) with fixed gaussian kernels and a simple convolutional neural network like Unet. The results from these two parts are used as inputs of a conditional random filed (CRF) as the recurrent neural network (RNN) to perform segmentation post-processing. The three parallel backpropagations are conducted during each training step to effectively train all the weights and parameters of UPSS.

Figure2. Qualitative assessment of segmentation results of Unet and UPSS. The input image a., Frangi filter result b., UPSS result c., Unet result d., comparison between Frangi filter result and UPSS result e., comparison between Frangi filter result and Unet result f.

Fig. 4 shows complementary contrast of a high resolution (0.4x0.4x0.3mm3) water content map and R2*-weighted image.

Fig. 2 Cortical profiles based on water content in different cortical areas.

Figure 1. Regions of interest, specifically CSF and vascular ROIs, for a representative subject, overlaid on the T1 anatomical. The lines indicate the fMRI slice orientation, and the arrow indicates the aqueduct. A = anterior, P = posterior, R = right.

Figure 2. The frequency spectrum of the fMR time series in CSF-related ROIs, contrasting spectra from other ROIs as well as the PPG-associated spectra. Spectra are averaged across subjects, and error bars represent standard error. CSF-related ROIs include the lateral ventricles (LV), the third ventricle (3rd V) and the cerebral aqueduct. All signals have been resampled (to the maximum frequency of the rs-fMRI data). Notice that for the PPG spectrum, the cardiac peak is substantially higher than the low-frequency peak.

Figure 1. Two Stroke STAGE cases with corrected FPs. First case (1st row) has a detected CMB (shown in red circle in A) on the SWI (A) that was located on the edge, and later using the extracted edge mask (B), it was removed from the SWI (C) as a false positive shown with cyan circle. Second case (2nd row) shows a detected CMB (shown in green circle) on the SWI (D) that was located on the vein, and later using the extracted vein mask (E), it was removed from the SWI (F) as a false positive shown with purple circle. Those are false positives that would have otherwise been included in the result.

Figure 2. Comparison of new and old pipelines on all datasets (STAGE and non-STAGE) (A) and only STAGE datasets (B) of the previously tested cohort. As seen in the figure, the new model works better on both datasets, and there is a performance improvement on STAGE datasets.

Figure 3 shows box plots of the flow standard deviation (stdev) and the average low-frequency power measured in the superior sagittal sinus (SSS) (n=10). The baseline condition is shown in white and the hypercapnic condition is shown in black. Hypercapnia was associated with a significant increase in flow standard deviation. There was also a trend for lower, low-frequency power in the hypercapnic condition compared with baseline.

Figure 4 shows box plots of the flow standard deviation (stdev) and the average low-frequency power measured in the superior sagittal sinus (SSS) at baseline. Young adults (n=5) are shown in white and middle-age/older adults (n=5) are shown in dashed blue lines. There was a trend for middle-age/older adults to demonstrate a smaller flow standard deviation, and lower, low-frequency power compared with young adults.

Figure 1. Average time course of mean BOLD signal in thalamus (blue), hippocampus (green), and the centrum semiovale (red), throughout a 12-minute scan at 1550 ms temporal resolution with two CO₂ challenges. Shaded areas represent the vasoconstriction (a; first challenge) and vasodilation (b; second challenge) segments to the BOLD-CVR analysis.

Figure 2. Axial slices of a parametric voxel-wise CVR map from representative participant (female, 65 year old), created using FSL FEAT (FSL, version 5.0, http://fsl.fmrib.ox.ac.uk) for visual inspection of global BOLD signal change. The colour bar denotes the range of parameter estimates obtained for each voxel.

Figure 3. Power maps of the cardiac- and respiratory-driven CSF motion components. The power of the cardiac-driven component looked like higher around the brainstem than that of the respiratory. Meanwhile, in the fourth ventricle, the contribution of the respiratory-driven component was high indicated by yellow arrow head.

Figure 2. Spectrograms of CSF velocity obtained from ROIs shown in Figure 1. Number at top-left on each image corresponds to the ROI number in Figure 1. Cardiac- and respiratory-driven velocities gradually changes from (1) to (4).

Figure 2. Representative DT_Ls at the entrance of fourth ventricle from the aqueduct of two volunteers: ROI-overlaid images (a) and ellipsoid-representation maps (b). Yellow boxes and green segments on the images (a) display the regions for ellipsoid-representation and for statistical analysis, respectively. DT_L is represented as an ellipsoid of which maximum ADC is 20×10^-9 m²/s for each voxel in the maps (b).

Figure 1. Representative multislice images of MD (a) and FA (b) of DT_L and DT_H. DT_L shows extremely high MD and FA in some subsegments of CSF.

Figure 5

Compared to artificial CSF simulating normal CSF, artificial CSF simulating bacterial meningitis and artificial CSF simulating obstruction of the subarachnoid space showed higher peaks and lower signal values at the side slope of the peak.

Figure 4

The peak heights/FWHMs of each albumin concentration were as follows: 2 mmol/L: 4.73×1011/0.016 ppm, 0.2 mmol/L: 2.08×1011/0.080 ppm, 0.02 mmol/L: 8.71×1011/0.016 ppm, 0.01 mmol/L: 9.75×1011/0.020 ppm, and 0.005 mmol/L: 9.45×1011/0.020 ppm.

Figure 1: Fast EPI acquisition (TR=100ms) showing signal changes due to CSF flow; axial slices with resolution of 1.53 x 1.53 x 3mm and sagittal slice with resolution of 1.5 x 1.5 x 4.36mm; a) bottom axial slice; b) top axial slice; c) sagittal slice.

Figure 4: Frequency spectrum with spatial localization of the signal from 4 different frequency bands. For each band, the example show a spatial coverage of the signal ranging from the bottom of the brain/lower cerebellum on the left to the center of the brain on the right. The bandwidth for each band is 0.24Hz. The center frequency is approximately a) 0.3Hz, b) 0.8Hz, c) 1.2Hz, d) 2.3Hz. The acquisition was done using an EPI sequence with TR=152ms with 15 slabs of 3 slices each for a total of 45 slices.

Contrast variation with temperature for T1 and T2WI.

Contrast ratio at different temperatures in T1WI and T2WI

Figure 2: Temperature dependence of T₁, T₂, T₂ using the forehead temperature measured in real time during the MRI measurement. Each parameter was differentiated for gray (gray triangles) and white (black circles) matter structures.

Table 1: Fitted linear model (y=a+bx) for T₁ and T₂ brain and forehead temperature, respectively, each differentiated for gray matter (GM) and white matter (WM). Further, the corresponding goodness of fit values, the correlation coefficient r, the root mean squared error (RMSE), and the p-value are listed.

Fig 5. Maps of the normalized area of intermediate exchanging amines for six CSF and one control phantoms (left). Note the variation between two labeled test tubes -- test tube "A" (CSF protein concentration= 487 mg/dL) and test tube "B" (177 mg/dL). "A" has a higher normalized area than "B". The corresponding z spectra (right) are also given.

FIg 2. Scatterplots showing the correlation between the normalized area for intermediate exchanging amines and CSF protein concentration (r=0.436, P=0.001). The straight and curved lines indicate the mean and 95% confidence interval.

Fig. 3: Example of MWF on 2 axial slices on 1 subject using data from (a) 7 flip angles and (b-d) top 3 performed subsets of using 3 flip angles. All sets of data produced MWF maps with similar contrast globally, and noise appearance followed the pattern suggested by the simulations. (c) Substantial differences to the full dataset result can be found as patches across the brain, depending on which flip angles were used, which is likely to be caused by the artefacts in the data and potential unbalanced MT effect (blue arrows).

Fig. 5: Compare to the results of using 7 flip angles (a), high-quality MWF and other quantitative parameters can still be obtained using data from only 3 flip angles on 1.4mm isotropic data (b), though the noise in these maps is more pronounced when compared to the 1.8mm isotropic MWF in Fig.3. (c) Bias caused by artefacts is also more noticeable in the high resolution maps because of the longer scan time. (d) Results with MP-PCA denoising only show negligible differences compared to those without denoising.

Fig. 2. The associations of gBOLD-CSF coupling to age, disease condition, MoCA, and UPDRS. (A) The strength of the gBOLD-CSF coupling (global BOLD-CSF correlation at +4-sec lag) decreased with age (Spearman’s ρ = 0.32, p = 0.0005). Age- and gender-adjusted gBOLD-CSF coupling is significant weaker in PD-MCI group (B) and significantly correlated with MoCA scores across all the subjects (C), within the PD group (D), but not across controls (E).

Fig. 1. Global BOLD is coupled with CSF changes. (A) left: global BOLD was averaged across gray matter; middle&right: CSF region at bottom fMRI slice. (B) Coupled changes of global BOLD and CSF signal from an example (indicated with arrows). (C) upper: averaged global BOLD-CSF cross-correlation across 118 subjects; lower: the one for the negative derivative of global BOLD and CSF signal. Shade: 95% confidence interval calculated with shuffled signals¹⁶. Error bar: standard error of the mean (SEM).

Figure 3 The resulting scalar maps obtained from the Bayesian regression and non-negative least squares approach. In Bayesian algorithm: v₁ is the myelin water fraction, v₂ and v₃ are the fractions of intra- and extra-axonal water and contamination by CSF, respectively. The relaxation times are presented by T_M, T_A, and T_CSF, respectively.

Figure 1 Algorithmic workflow of the optimised pipeline. The pipeline consists of four steps: noise correction of T₂ weighted images, Gibbs-ringing correction, normalisation and smoothing of all volumes, and finally estimation of myelin water fraction. Other possible correction steps are marked as “optional”.

Figure 1: The nine different sets of b-values used to estimate the intermediate fraction. Red: logarithmically scaled, green: linearly scaled, and blue: oversampled $$$b \cdot D_{int}$$$ range.

Figure 3: Precision and accuracy for the $$$f_{int}$$$ estimation in white matter hyper-intensities. The arrows show the effect of acquiring more b-values for logarithmically scaled (red) linearly scaled (green), and over-sampling $$$b \cdot D_{int}$$$ range (blue) b-value strategies.

Sequence of motion in a single slice across time. (Top) RL motion; (middle) AP motion; (bottom) SI motion, where red (positive) indicates left, posterior, and inferior motion, respectively.

Comparison of velocity profiles from a triggered “still frame” scan reconstructed with the full 30 q-space directions and 3 random subsets of directions. RGB color represents L/R, A/P, S/I magnitude (speed) respectively. Although there is a slight change in SNR between the images, the overall image is preserved.

Figure 4: Upper figure: Different slices of the whole brain H₂O maps (in percentage units) generated with the proposed method are displayed in the transverse, coronal and sagittal planes (from top to bottom). The water content is calibrated to that of the CSF which is assumed to have 100% water content. A CSF T1 value of 4300ms was used for calibration. Lower figure: Histogram plots of the SRR-H₂O and long-TR-H₂O values. The peak at around 70% is for the WM and that around 83% is for the GM.

Figure 3: Upper rows show water content maps (in percentage units) of an ROI in the phantom obtained with reference (long-TR-H₂O), proposed (SRR-H₂O) and the zero padded interpolation (Interpolation-H₂O) methods obtained with an equal scan time. Blurring can be noticed in the interpolation-H₂O method while the SRR-H₂O maps are comparable to the reference method. The lower figure plots the sigmoid fit to an edge selected which is shown as a red line in the top most figure. The steeper the fitted curve, the better the resolution.

Figure 1. Significant differences in T_W, T_G and T_C between the very preterm (VP) and full-term (FT) groups at age 13 years (p<0.05, FWE-corrected, adjusted for age and sex). Widespread decreases in T_W in the WM are accompanied by increases in T_G and/or T_C in similar areas. There are also some cortical decreases in T_G, which are accompanied by increases in T_C.

Figure 3. Significant correlations of neonatal brain abnormality score with T_W, T_G and T_C at age 13 years in the very preterm (VP) group (p<0.05, FWE-corrected, adjusted for age and sex). Widespread negative correlations between neonatal brain abnormality and T_W in the WM are accompanied by positive correlations with T_G and/or T_C in similar areas. T_C also showed positive correlations in some cortical areas.