Time-series cerebral blood flow rCBF images are shown for the right hippocampus (A) and right middle temporal lobe (B) of Patient 1 from Group 1. With the progression of time, the rCBF values show an increased signal for structures in the limbic system and the cortical area for all patients in Group 1.

T-value maps generated using the SPM12 software results are shown for the MCA region of the brain comparing all the patient groups at each time-point. A single patient rCBF map is shown with intensity overlays, which show the regional differences for each group comparison. Higher T-values indicate greater differences compared to the lower T-values. Groups 1 and 3 show the greatest difference between each other compared to the other group comparisons particularly at the 1-month and 2-month time-points. By 4 months, groups 1 and 3 showed more differences compared to the others.

Fig1: Cortical variation of the correlation between the complement of water content (macromolecular content) and R1, described in a linear model with coefficients beta0 and beta1, and by Pearson’s correlation coefficient. We mention that R1 and 1/H2O show a nearly identical correlation to R1 and (1-H2O), although the physical models behind the two dependencies are very different [14] (data not shown). The sensorimotor areas, known to have high myelination, show strong correlation with (1-H2O) (large beta1 and r), accounting for practically all the relaxation rate (small beta0).

Fig3: Age dependence of mean values as well as of coefficients describing the linear dependence between selected pairs of quantitative parameters in ROIs from the left hemisphere. Only correlations with p<0.05 are shown.

Figure 2. The four states identified by k-means clustering analysis and the corresponding cluster centroids. The total number and percentage of reoccurrence times of each state are listed above each cluster (A), together with 5% strongest connections of each state (B). BG: basal ganglia; AUD: auditory; VIS: visual; SMN: sensorimotor; CEN: cognitive executive; DMN: default mode; CB: cerebellar.

Figure 1. Independent components (n = 33) identified by group independent component analysis. (A) Independent component spatial maps divided on the seven functional networks. (B) Group averaged static functional connectivity matrix between pairs of independent components.

Figure 3. a) Scatter plots showing NLI and CT’s correlation with age in left caudal anterior cingulate. Partial plots of NLI against age controlling for CT are shown on the far right; b) The slopes of the linear fitting lines describing the decreases of standardized NLI and CT with age for all atlas regions. * indicates statistical significance; c) The slopes of linear fitting lines of the NLI as a function of age, controlling for CT. Color indicates slope values from significant partial correlations. Darker blue indicates faster NLI decline with age controlling for CT.

Figure 4. a) Sample T₁w/T₂w profiles of the youngest (blue) and oldest 25% participants (red) from the left caudal anterior cingulate region; b) Partial plot showing T₁w/T₂w value at white matter surface as a function of age after controlling for CT; c) Regions with significant age-related T₁w/T₂w decreases controlling for CT at pial and white matter surface. Color indicates the p-values of partial correlation after Bonferroni correction.

Fig.1: Analytic pipeline. Step1: The time-course signal of 53 ICNs have been identified using group-ICA in the Neuromark template. Step2: A taper sliding window was used to segment the time-course signals and then calculated the functional network connectivity (FNC). Each subject has 139 FNCs with a size of 53×53. Step3: A k-means clustering with correlation as distance metrics was used to group FNCs to three distinct clusters. Then, based on the state vector, we calculated the occupancy rate (OCR) for each subject. Step4: Finally, we compared the dFNC states across two sessions.

Fig. 5: Three identified states by k-means clustering methods for mild dementia (MD) patients. The top panel shows the identified states in sessions1, and the bottom panel shows the identified states in session2. The color bar shows the strength of connectivity. The hot colors show positive and cold colors show negative connectivity. SCN: Subcortical network, ADN: auditory network, SMN: sensorimotor network, VSN: visual network, CCN: cognitive control network, DMN: default-mode network, and CBN: cerebellar network.

Figure 3. Maps of regression coefficients on the subcortical brain structures for the association of the shape measure (Jacobian determinant) with Alzheimer’s pathology, atherosclerosis, and TDP-43 pathology (global FDR correction). Negative coefficients mean inward deformation. The results are shown on the left hemisphere since only one hemisphere was imaged per participant and hence the right hemisphere segmentations were mirrored to left and combined with the rest of the left side segmentations in this analysis.

Figure 1. Demographics and clinical characteristics of the participants

Tract-based spatial statistics (TBSS) analysis showing regions on the white matter skeleton where the null hypothesis is rejected (yellow with red filling; p≤0.05) and accepted (green; p>0.05). The null hypothesis proposes that there exists no positive correlation between neurite density (NDI) and total cerebral blood flow (tCBF) in cognitively normal (CN) subjects. The mean fractional anisotropy image is used as the underlay.

Tract-based spatial statistics (TBSS) analysis showing regions on the white matter skeleton (at 3 MNI coordinates) where the null hypothesis is rejected (yellow with red filling; p≤0.05) and accepted (green; p>0.05). The null hypothesis proposes that neurite density (NDI) is not greater in cognitively normal (CN) subjects compared to Alzheimer’s disease (AD) subjects. The mean fractional anisotropy image is used as the underlay. Note some areas were erroneously masked after TBSS registration.

One example set of images per group in the study. The top row shows the age- and sex-corrected median stiffness map. The bottom row shows the effect size map.

Example quantile maps at one slice location for a single NPH participant (top row) and the CU group (bottom row).

Fig. 3. (A) Sagittal slices (arranged from medial to lateral) and (B) axial slices (arranged from superior to inferior) showing regions of significant R₂ shortening associated with LATE-NC. Corrected p-value maps obtained from voxel-wise analysis are shown in blue (p<0.05).

Figure 1. Axial views of representative Tau-PET (left) and QSM images in an APOE ε4 positive subject.

Figure 3. Correlations between Tau PET SUVR and MRI measures in the temporal lobe of APOE-ε4 positive subjects. Significant correlations were seen between susceptibility, FA, and MD.

Figure 1: Representative region plots for structural parameters MWF, R₁, and R₂, and DTI indices. MWF, R₁, R₂, MD, RD, and AxD conform to a nonlinear trajectory along the lifespan of the adult in the human brainstem. For each ROI, the coefficient of determination, R^₂, and the significance of the linear regression model, p, are reported.

Figure 2: MWF, relaxation times, and DTI indices standardized and plotted as a function of age for six structures to show similarities and differences in the brainstem across the adult lifespan. Three WM regions and three GM nuclei were chosen specifically since they demonstrated significant quadratic associations with age across all parameters. Diffusivity indices were inverted for easier comparisons.

Effect-size from ANCOVA for main effects of group and sex. The results are shown at a large-effect size level for the main effect of group and medium effect-size level for the main effect of sex. *FW index did not show any significant cluster at large effect-size level. The violin plots show the mean values of each metrics inside the significant clusters. No significant difference across groups was found for the interaction term.

Effect-size from post-hoc comparison between females and males inside AD and HC groups. We found significant sexual dimorphism at large effect-size level inside the AD group by all metrics. In HC group, only VBM detected differences between females and males at the large effect-size level. *FA and FW index, for HC group, detected differences only at medium effect-size level. Males showed higher FA, higher FW index, and lower GM volumes than females in AD group.

Figure 1: Sagittal, coronal, and axial slices of the MIITRA T1w and DTI templates, gray matter labels, and confidence maps of the gray matter labels (average confidence 0.892). All resources were constructed in 0.5mm voxel-size using a super-resolution-based approach.

Figure 2: Box plots of the (A) Dice coefficient and (B) Jaccard coefficient for the overlap between the MIITRA gray matter labels warped to an individual’s space and the manually edited reference labels in that individual’s space.

Fig.1 A: The difference of ALFF between the presbycusis (PC) and normal hearing controls (NH) groups in slow-4. B: The difference of ALFF between the PC and NH groups in slow-5. Hot and cold colors indicate significantly higher and lower ALFF in the PC group than in the NH group, respectively. Results obtained by a two-sample t-test. FDR corrected p < 0.01. Abbreviations: L, left; R, right; ALFF, amplitude of low-frequency fluctuation.

Fig.4 ALFF and related FC differences between the presbycusis (PC) and normal hearing controls (NH) groups in slow-4. Between-group differences in FC analyses were only found related to the seed of left DLPFC. Hot color indicates significantly higher FC in the PC group than in the NH group, respectively. Results obtained by a two-sample t-test. FDR corrected p < 0.05. Abbreviations: L, left; R, right; ALFF, amplitude of low-frequency fluctuation; FC, functional connectivity; DLPFC, dorsolateral prefrontal cortex.

Table 1. Summary of the statistical results of the demographic data.

Figure 1. Results of voxel-based comparisons of myelin water fraction (MWF), gray matter volume (GMV), and white matter volume (WMV) among the three participant groups.

Fig.1 ROI selection of gray matter nucleus.(A)The HCN、PUT、GP、VT are shown front and back. (B)Showing the RN and SN. (C) Showing the DN. (D) Showing the FG.

Fig.2 Comparison of gray matter nucleus MSV between AD and CON group. ★P < 0.05.

Table 1. List of parameters used in the simulation with the condition that only the vascular structure exists in the voxel

Figure 1. Simulation modeling

This figure shows the simulation models with 3D image of the modeled vascular structure (1a) and the corresponding cross-section taken along the z-axis (1b), 3D image of the modeled Aβ or microbleeds (1c) and the corresponding cross-section taken along the z-axis (1d), and 3D image of the modeled vascular structure with the amyloid-beta plaque or microbleed structure (1e) and the corresponding cross-section taken along the z-axis (1f).

Table 1. Summary of the demographic data, results of the neuropsychological test, the three blood-based biomarkers, and results of statistical analyses.

Figure 1. Results of the voxel-based multiple regression analyses between brain tissue volumes and the levels of the three blood biomarkers

GMV, gray matter volume; WMV, white matter volume; PPIA, peptidylprolyl isomerase A (PPIA, known as cyclophilin A, CyPA); HO-1, heme oxygenase-1; IRE1, and inositol-requiring enzyme 1

Brain clusters showed AxD differences between AD group and HC group after FWE correction.

Correlations between diffusion indexes and cognitive scores in Alzheimer’s patients.

Figure 1. MRI and ultrasound in aged normal (WKY) and hypertensive rats (SHR). A) The brains of aged (14-19 mo) WKY and SHR were characterized by multiple MRI contrast mechanisms to detect lesions. SHR had enlarged ventricles and more CMBs (yellow arrow heads) when compared to WKY. B) Quantification of ultrasound parameters in the same aged animals was performed including vessel wall structure, pulse velocity, flow ratio (calculated), and tissue velocity. The * indicates a P value less than 0.05 and the ** indicates a P value less than 0.01.

Figure 2. rs-fMRI-EEG power ratio versus age: epoch-based EEG. The fMRI-to-beta power ratio (SDratio) exhibits the greatest age effects (young > old) (f, h). The effects are most pronounced in the 0.1-0.3 Hz range, and the affected regions span nearly the entire anterior cortex, also including subcortical regions such as the putamen and pallidum. The fMRI-to-theta ratio is lower in the young subjects in only the paracentral cortex.

Figure 1. Models for the mediation analyses

Figure 2. Structural MRI and MRE stiffness images of A) a healthy older adult, and B) a patient with mild cognitive impairment (MCI). While smaller volumes of the hippocampus (arrows) are predictive of MCI, viscoelasticity measures from MRE presented a greater risk factor for a clinical diagnosis.

Figure 1. Age and sex adjusted association (risk ratios and 95% CIs) between hippocampal neuroimaging measures and the risk of MCI.

Figure 3. Worsening of atlas quality with increased simulated brain atrophy indexed by the age of the target elderly brain. Distance is measured with respect to the gold atlas (see Figure 1).

Figure 4. Sensitivity to simulated brain atrophy is resolved when using brain rejuvenation (Figure 2) before FreeSurfer.

Figure 5. Significance of each coefficient incorporated into the linear regression analysis of NDI or ODI derived from NODDI or modified NODDI in each region-of-interest (ROI). * indicates p < 0.05, ** indicates p < 0.01, + indicates p < 0.1, and - indicates non-significant effects. All p-values presented are obtained after FDR correction.

Figure 4. Examples of regional ODI trends as a function of age. A) ODI results derived from the original NODDI approach, and B) ODI results derived using our modified NODDI approach. Several regions investigated show U-shaped trends of ODI with age. Results were very similar for both NODDI approaches (see Fig. 5).

Figure 1. Mean myelin water fraction (MWF) values for each APOE genotype (N = 92), with values relative to mean APOE ε33 MWF values. Results are shown for eighteen WM and GM brain structures/ROIs. The APOE ε2/- group exhibits higher mean MWF values as compared to the APOE ε4/- or reference groups (APOE ε33) in most ROIs while the APOE ε4/- group exhibits lower mean MWF values as compared to the reference group in several ROIs. * indicated p < 0.05 obtained from an ANCOVA analysis.

Table 1. Slope, β, and significance, ρ, of the regression terms incorporated in the multiple linear regression given by MWF ~ β₀ + β_age × age + β_sex × sex + β_APOE × APOE + β_race × race + β_age² × age². Sex and race results are not shown as they exhibited overall non-significant associations with MWF in most ROIs.

Figure 1: Visualization of study paradigm and spectroscopy data quality. A. MRS scans and Barnes Maze testing were performed in TgF344-AD and WT littermates at 4, 10, and 16 months. Rats were randomly separated into three groups treated with Naproxen: early long treatment, early short treatment, and late treatment. B. Localized 1H-MRS spectra from the hippocampus of a 4-month-old TgF344-AD rat acquired using the PRESS pulse sequence at 7T. Voxel placement is shown in the top right.

Figure 3: Early but not late treatment mitigates some disease-related metabolic changes. Three treatment paradigms were tested, with timing of Naproxen administration denoted in the figure legend. Linear mixed effects modelling was applied to examine age*genotype*treatment effects. Each rat is depicted by an individual data point. The linear model used to fit the data is represented by a line of best fit and 95% prediction interval (shaded). FDR correction applied at 5%.

Fig 2. Sample from motor cortex in FTLD-tau with four-repeat GGT tauopathy and clinical naPPA. Note the dark band in the middle layers of cortex, progressing along the entirety of the gyrus (green arrows) in the T2*w image (left). This pattern correlates with a similar band of iron (middle). We note that this band overlaps, but is not well-correlated with, myelinated radial fibers on LFB, which show minimal degeneration despite the high degree of gliosis (right).

Fig 1. Sample from motor cortex in FTLD-tau with four-repeat tauopathy and clinical PSP. Note the dark band in the middle layers of cortex, broadening as it progresses down the right side of the gyrus (green arrows) in the T2*w image (left). This pattern correlates with a similar band of iron (middle), but not with a particular change in tissue structure or myelination on LFB (right).

Fig. 1: Schematic of the template construction technique.

Fig. 3: A) Example axial, sagittal and coronal slices of the new template, MCALT_0.5mm, ICBM2009b_0.5mm, Colin27_0.5mm and HCP_0.5mm templates. B) Atypical anatomical features can be observed in the cortex of MCALT_0.5mm (red circles). C) Normalized power spectra of all templates.

Figure 1 - Axial views of MAP-MRI diffusion parameter maps in a healthy 60-year-old male. From left to right: RTOP (mm^-3), RTAP (mm^-2), RTPP (mm^-1), MSD (mm²), QIV (mm⁵), and NG (dimensionless).

Figure 2– Average MAP Age trajectories extracted from the cingulum for RTOP, RTAP, RTPP, MSD, NG, and QIV, with linear fits for the control group (blue), MCI group (orange), and AD group (red).

Figure 1. Comparison results of different DKI-derived parameters between HC and AD, aMCI and AD, and HC and aMCI, respectively. Green represents the FA skeleton overlaid on the FMRIB58_FA template. Yellow indicates the brain structures with significantly decreased parameter values in the aMCI and AD groups (p<0.01, TFCE-corrected). Red indicates the brain structures with significantly increased parameter values in the aMCI and AD groups (p<0.01, TFCE-corrected)

Figure 2. The correlation analyses with the maximal correlation coefficients between DKI-derived parameters and neuropsychological testing scores (MMSE scores, MoCA scores)

Fig 4. Group differences of functional connectivity in the subiculum, CA1-3 and CA4-DG networks between healthy older adults with negative CSF biomarker status (HO-) and positive CSF biomarker status (HO+). Significant at ^*P < 0.05.

Fig 5. Correlations of imaging measures with neuropsychological performance (A) and CSF biomarker measurements (B). The fitting lines are also shown to indicate trends.

Connectometry analysis detected differences in structural connectivity between HC and MCI in two distinct tracts: one between the left-lateraloccipital and the left-insula, which includes part of the left IFOF and the left ILF, and one between the left-precuneus and the left-supramarginal, which includes part of the left superior-longitudinal fasciculus. The yellow circles in the HC-MCI p-values connectometry matrix show the significant edges with p<0.05.

Voxel-based analysis of FA, IC, ISO, and EC maps. Lower values of FA in the MCI group, compared with HCs, were found in the fornix. Lower IC and higher ISO values were found in MCI in several white matter regions and tracts. No significant differences between groups, at p<0.05, were found for the EC metric.

Table 3 Pearson correlation between MoCA with white matter lesion volume, MoCA with gray matter thickness and white matter lesion volume with gray matter thickness

Table 2 Difference of MoCA, gray matter thickness and normalized gray matter volume in each age group by one-way ANOVA.

Figure 1. Relative volume distribution of hippocampal volume with respect to age. The lines indicate the 10th, 90th (dashed) and 50th (solid) percentiles for healthy controls [2]. The shapes of the scatter plot represent the clinical diagnosis (AD,MCI, normal) and the color represent the ATN classification.

Figure 2. ROC curves of the univariate approach using hippocampus and temporal GM relative volumes (in orange) with respect to Aβ42 (green), pTau (red), and Tau (blue). The logistic regression best hyperparameters were: L2 regularization with a coefficient C=0.001, without weighting the classes.

Figure 2. The three specific sub-networks identified by NBS. These sub-networks showed reduced connectivity with increased BMI. Figures were created using BrainNet Viewer⁹

Figure 1. Illustration of networks identified by different thresholds in NBS. Ultimately a threshold of 5 was chosen to be sufficiently specific.

Figure 1. Group maps of white matter tracts segmented with three methods. Colour scales show the voxel-wise frequency for each tract (total N=105).

Figure 2. Boxplots of FA and MD obtained in each segmented tract using each method.

Table 1. Summary of the statistical results of the demographic data and results of the neuropsychological tests obtained in the participants

Figure 1. Representative magnetic resonance images and metabolite spectra.

Figure 2. Neural network inversion calculated mean stiffness maps for two groups controlling for age and sex. Each column indicates an image from different slice locations, MNI coordinates are arranged from inferior to superior positions.

Figure 3. Neural network inversion calculated mean damping ratio maps for two groups after correction for age and sex. Each column indicates an image from different slice locations MNI coordinates are arranged from inferior to superior positions.

Table 1. Comparison of left and right hemispheres（genu and splenium of corpus callosum） among the relaxation times

Fig. 3 PC patients exhibited decreased internetwork connectivity between the DMN and auditory network (AN) and between DMN and anterior salience network (SN) and between SNI and SNII and between SN and visual network (VN), as well as increased internetwork connectivity between the DMN and dorsal attention network (DAN) and between anterior DMN and posterior DMN.

Fig. 2 Compared with controls, PC patients showed significantly decreased network connectivity within the default mode network (DMN), Left executive control network (lECN) and right executive control network (rECN).

Figure 1. Multi-atlas segmentation scheme for thalamic nuclei segmentation. The multi-atlas consists of 20 manually segmented WMn MPRAGE data which are warped to subject space and label fused using a majority voting scheme. A WMn template is used as an intermediate step to improve robustness and cropping is performed to improve speed and accuracy.

Figure 2. Thalamic nuclei segmentation labels from the modified THOMAS method overlaid on MPRAGE data in axial (top panel) and coronal (bottom panel) for a patient with AD (enlarged ventricles). Note visualization of small structures such as anteroventral (AV), centromedian (CM) and habenula (Hb).

Figure 2. The area under the ROC curve of train and validation set based on radiomics model

Table 1. Clinical statistics in the diagnosis.

Figure 2: Regions in the right and left hemispheres (RH; LH) demonstrating significant differences in OBW versus OBM, where OBW demonstrated greater cortical thickness in these regions indicated in blue (p = 0.00010, corrected for age, education, study). The graphs demonstrate no influence of VO2peak on cortical fitness and the relationships for OBW and OBM separately in the RH (graph on the left; p > 0.05) and LH (graph on the right; p > 0.05).

Figure 1: Regions in the right and left hemispheres (RH; LH) demonstrating significant differences in females versus males, where females demonstrated greater cortical thickness in these regions in teal (regardless of BMI status). (p =0.00010; corrected for age, education, study)

Figure 2.  3xTg mice have smaller basal forebrain volume. (a-e) Automated image registration-based quantification of the basal forebrain volume of control and 3xTg mice. (f) Manually drawn ROI showed decreased MS volume in 3xTG mice. (g) The automated image registration-based quantification and manual ROI quantification show significant correlation. MS, medial septum; SI, substantia innominate; HDB, the horizontal limb of the diagonal band of Broca; VDB, vertical diagonal band nucleus; BNM, basal nucleus of Meynert. Data are shown in mean ± SEM, *P<0.05.

Figure 5. Correlation between MRI-detected volume and cholinergic cell proportion and density. (a-f) Simple linear regression of MRI volume of MS and VDB, as represented by the proportion of intracranial volume (% ICV), plotted against p75 positive cell proportion (n=12, p>0.05).

DGE-MRI results for brain parenchyma in APP/PS1 mice before and after IL-33 treatment. (A) Dynamic difference images in brain parenchyma. (B) Experimental DGE plots for brain parenchyma. (C-E) Comparison of the maximum signal S_max, clearance parameters S_max-△S_end and glucose uptake rate μ_in before and after treatment. Data were presented as mean ± SEM (n=4).

DGE-MRI results for CSF in APP/PS1 mice before and after IL-33 treatment. (A) Dynamic difference images in CSF. (B) Experimental DGE plots for CSF in APP/PS1 mice. (C-E) Comparison of the maximum signal S_max, clearance parameters S_max-△S_end and washout rate μ_out before and after treatment. Data were presented as mean ± SEM (n=4). *P<0.05, one-way ANOVA.

Figure 2. Complex wave images (real part) for control brain (top row) and brain of 5xFAD AD mouse model (bottom row), along with the delineated contour of zero displacement shown in white color.

Figure 1. Representative block diagram of the Magnetic Resonance Elastography (MRE) experimentation set-up.

Figure 2. Regional linear model of relative volume=ns(age,2)*sex+(1|subject) with columns A-E corresponding to different model terms. Each regional t-value for each model term is mapped to the corresponding atlas label and overlaid onto the anatomical average image as a regional t-value heatmap. Only regions with statistics that surpass a 5% FDR correction are plotted and the upper bound of the scale corresponds to the maximal t-value modeled by that term. t-value scale is shown for each term.

Figure 3. Data from selected grey matter regions is shown with individual data points shown and the overall model as with confidence interval as described in Methods overlaid. T-values for each region and model term are shown on bottom, with significant terms after 5% FDR correction given in bold.

List of brain regions and morphometry metrics that were found to be significantly associated (highlighted) with either education, education x disease status, or age x education x disease status.

Top: Sample hippocampal subfield segmentation results. Bottom: Distribution of Bayesian regression of hippocampal subfield volumes examining MCI effects and age x MCI interaction effects. Distribution tails crossing zero are non-significant. Presubiculum and subiculum volumes were significantly decreased in MCI. CA1, CA3, hippocampal tail, molecular layer, presubiculum, and subiculum volumes were negatively associated with age x MCI interaction.

Clinical and demographic data of participants

Fig. 1 The parcellation map with ROIs of the limbic-associated areas overlaid on DTI-derived maps.

Figure 1. Plots illustrating regional MWF, FA, MD, RD, or AxD values as a function of age for four representative cerebral white matter structures/ROIs. For each ROI, the coefficient of determination, R², is reported.

Figure 3. Plots illustrating examples of regional correlations of FA, MD, RD, or AxD and MWF. Results are shown for four representative cerebral white matter structures/ROIs. For each ROI, the coefficient of determination, R², is reported.