Figure 4. (a) CBF map at rest (top) and two activation tasks (FT, middle; FB, bottom). (b) CBF increase evoked by the FT and FB tasks are highlighted in two insets. (c) Laminar profile of CBF at three tasks. (d) Laminar profile of CBF increase evoked by FT and FB. FT induced CBF increase shows a ‘double-peak’ pattern which corresponds to sensory input (superficial) and motor output (deep) layers respectively. FT induced CBF increase shows a weaker increase in superficial layers which corresponds to exteroception sensory input and minimal motor output.

Figure 5. Six axial slices of the control images (a), resting-state perfusion maps (PLD=1000ms) (b) and CBF maps at rest, two activation tasks and the difference between rest and activation (blue boxes) (c). The proposed technique has sufficient coverage for perfusion measurement in M1 and supplementary motor area (SMA), etc.

Figure 3: Reconstructed perfusion-weighted images with intentional motion and without intentional motion. The acquisition technique is indicated by 3D GRASE/3DGP and reconstruction without motion correction (NO), 3D GRASE three-dimensional rigid body motion correction (MOCO), standard PROPELLER motion correction (STD) and the proposed joint estimation of motion and distortion (JET) is applied. Note that in contrast to the original 3DGP method, a single motion estimate is calculated for each brick instead of the slice-by-slice approach to increase the robustness.

Figure 4: a.) Procedure of 3DGP-JET. All bricks (three out of twenty control bricks from the third subject shown here) are used to estimate the underlying distortion field and individual motion parameters (Eq. 3). The distoriton field is then used to compensate distortion in individual bricks by unwarping with interpolated inverse field maps. Arrows indicate areas with large off-resonances and visible distortion in uncorrected 3DGP images; and b.) Mean and standard deviation of calculated motion parameters for the datasets without intentional motion using 3DGP-JET and 3DGP-STD.

Figure 3. Experimental territory mapping using B0 local field control to selectively label internal carotid (IC) arteries. (a) The 1-cm labeling plane is shown with the baseline B0 field maps and voxel freq. histogram inside vertebral and IC artery masks. (b) Shimmed B0 maps and histograms showing spectral separation of the “desired” left IC from the “reject”. (c) Similar for right IC. Mean perfusion maps (difference between labeled and control case) at right show the applied shim field successfully shifts most of the “reject” arterial spins outside the passband of the labeling pulse.

Figure 1. Loop geometry and photo/rendering of experimental and simulated B0 shim array hardware considered for ASL applications. B0 field map coverage for each array is shown at left. Representative element field maps are shown for both designs at level of slice shown, in the ASL labeling region. (a) 32-ch “AC/DC” shim array used for proof-of-concept territory mapping experiments. (b) Simulated 12-ch shim array specialized for ASL shimming applications, now under development.

Fig 3. From top to bottom, cerebral blood flow, arterial transit time, water exchange rate, arterial blood volume, and water permeability maps were estimated for 8 slices.

Fig 2. (a) Sequence diagram of proposed 2D multi-slice multi-delay DWASL. (b) Bipolar diffusion weighting gradient was applied after slice excitation RF pulse which increased TE by 30ms. (c) Slice acquisition order was permuted through sets. 8 pairs of control and labeled image were obtained for each set. Slices of first half were imaged at odd PLDs, and second half were imaged at even PLDs.

Figure 1 - Interleaved-TE ASL data obtained simultaneously for relative cortical perfusion and relative BCSFB-ASL signal. Averaged time courses for relative cortical perfusion (column 1) and the individual animal averages (column 2). Relative BCSFB-ASL signal averaged time courses (column 2) are shown alongside their respective individual animal averages (column 3). Rows represent each drug or challenge: row 1 - CO₂ (a-d), row 2 - vasopressin (e-h), row 3 - caffeine (i-l). Further details of the drug challenges can be found in Table 1.

Figure 2 - Adult (n = 14) vs aged (n = 14) response to vasopressin.

a) Relative BCSFB-ASL signal obtained using Interleaved-TE ASL, showing both adult and aged averaged time courses (vasopressin administration at 10 mins).

b) Adult (n = 14) vs aged (n = 14) response to vasopressin, individual animal averages (baseline and post-vasopressin challenge).

Figure 3. The DIR (GM) map, S₀ map, T_2v (ms) map and Y_v maps for the two subjects. The extrapolated S₀ values are proportional to the venous CBV and, as expected, show higher signal in the GM than in the WM. The fitted T₂ maps and the corresponding Y_v maps reveal a more uniform contrast between GM and WM, as expected for brain OEF. Several voxels (mainly in white matter region) with very low signal intensities were excluded during the data processing.

Figure 2. a) Idealized cartoon depicting the evolution of the longitudinal magnetization of water in different compartments. Relaxed and dephased spins are denoted by upright arrows and solid circles, respectively. b) Simulated magnetization evolution for the arterial and venous blood. The consecutive FT-VSI and NSI pulses invert spins flowing above the V_CUTOFF for arterial nulling and preserve the spins moving below the V_CUTOFF. These preserved spins will largely outflow into the venules during TI, thus yielding a high SNR.

Figure 3: The VESPA decoded VSASL and PCASL data, with the CBF and ATT maps calculated by voxelwise fitting the VESPA signal model to the decoded VESPA data. Five slices from 1 subject are shown. Longer ATTs can be seen in the white matter, posterior circulation, and at both internal and external border zones. Before fitting, the data were voxelwise calibrated by the M0 image and corrected for slice-dependent T1 decay.

Figure 1: Sequence diagrams, label/control encoding, and contrast decoding strategies for PCASL, VSASL and VESPA. In VESPA, VSASL and PCASL are combined and used to label different pools of inflowing arterial blood. VSASL labeling (VS1) uses a spatially-selective VSASL module surrounding the imaging slab. The PCASL labeling plane is immediately inferior to the VSASL slab. By alternating label and control conditions in different orders, VSASL and PCASL contrasts can be linearly encoded into the acquired images and decoded by simple image arithmetic

Fig. 2: High-resolution (2.5x2.5x2.5mm³) regional cerebral blood flow (rCBF) maps acquired with 3D multi-shot, stack-of-spirals pCASL, from three representative infants aged 2months (A), 8months (B) and 17months (C).

Fig. 3: Main effect of age on cerebral blood flow (CBF) during infancy. (A) Global CBF increases during infancy. Data points represent global CBF measured with phase-contrast MRI of each subject. (B) rCBF increases throughout the cortex, but more prominently in the heteromodal association cortex. Images thresholded at t(df=21) > 2.83 with p<0.01. (C) heterogeneous developmental rate of rCBF (ml/100g/min/month) across cortex during infancy. Slower and faster rCBF growth rates are shown in cool and warm colors, respectively.

Figure 1. Regional CBF (mean ± standard error) between baseline and follow-up in major GM structures of Study 2.

Table 3. Linear mixed models for assessing the changes of global CBF and normalized volume in Study 1 (N=127) and Study 2 (N=115). Age, time interval, gender, GM probability and WM probability were added as the fixed effects in Model 1, and age-by-follow-up time interaction was additionally added in Model 2.

Figure 3: (A) MS-FAIR perfusion and CTA images in a patient with extensive PE at diagnosis. Red arrows show CTA pulmonary artery filling defects, which correspond well with ASL perfusion defects. Yellow arrows indicate vessels with normal filling on CT that correspond well with regions of higher ASL perfusion. (B) The same patient 6 months after tPA treatment, showing significantly improved perfusion. PVO before and after treatment were 73% and 8% measured by CT, and 63% and 39% by ASL. A residual defect was detected by ASL (B, red arrows), while the vascular occlusion by CT had resolved.

Figure 1: (A) Sagittal MS-FAIR quantified perfusion images in a patient with extensive filling defects across the right lung noted by CTA 6 days prior to ASL imaging. The edges of the right lung are outlined in red. Average PVO for this patient was 55% measured by CTA, and 44% measured by ASL. (B) Overlaid PVO values on the perfusion images, with higher degrees of obstruction corresponding well with the perfusion defects in Fig 1A. Overlaid PVO values under 50% were made transparent for visualization purposes

Table 2. Requirements. This table provides the basic requirements on the operating system, software, and data input types. It further lists the compatible vendors (note that other vendors might also work on each pipeline even though they were not tested for it). All pipelines were developed and tested to work with human brain scans. The last column lists other organs and animals that were tested.

Table 1. Contact details. This table contains the basic details about the pipeline such as the pipeline author, affiliation, and a link to a scientific publication with more details about the pipeline. Furthermore, it is indicated if the pipeline is available on request or by direct download (see TF 1.1 website¹⁵ for the links to the pipelines and author contacts). Last columns indicate if the pipeline is provided with a manual or guide, open source code, and a graphical user interface (GUI).

Figure 1: Projected timeline for the OSIPI ASL Challenge. The challenge will be open for a period of six months, with winning teams presenting at the ISMRM Perfusion Study Group meeting at ISMRM 2022.

Figure 2: Example perfusion-weighted images (arbitrary units) for A) population-based and B) synthetic datasets.

Figure 2. a) Three ATT distribution profiles used for PLD optimization b) and generated optimal PLDs for each ATT distribution along with evenly spaced PLDs .

Figure 3. The estimated CBF and ATT maps and corresponding error maps with the four ASL protocols at SNR of 15dB from a subject. Ground truth was CBF map and ATT map calculated from the in vivo ASL scan, respectively.

Figure 1. a) An example of the shine-through effect in a healthy volunteer seen in the perfusion block of a fully acquired time-encoded pCASL (te-pCASL) dataset noticeable as negative signal in the large arteries and/or positive signal in higher slices. b) The temporal standard deviation (tStdv) of the same dataset shows that the effect (high tStdv in vessels) is present in all post-label-delays (PLDs) and that tStdv is an insightful measure of the effect.

Figure 4. ASL (subtracted) signal is shown for a single slice low in the brain, for the seven individual experiments in which RF (labeling) was enabled for single blocks only (columns). The separation of the individual blocks in each experiment can be seen in the black squares which have positive ASL signal. We observed that the shine-through artefact (red arrows) was visible in the disabled blocks, (negative or positive signal), and surprisingly, also in the experiment without any labeling at all (last column, experiment #8). See Figure 1 for explanation of the experiment numbers.

Averaged ASL images of each PLD with different eTEs and T₂ maps in the MNI-space. The left and right panels show the acquired images without and with DANTE, respectively.

Averaged CBF and transit time (TT) maps in the MNI-space.

Figure 2. CBF sensitivity maps to macrovascular contamination by eight protocols (a to h) and average (i). Each protocol has CBF maps estimated by GKM (left), GKMmvc (middle), and their difference (GKMmvc - GKM, right). Units: error% per 1% aBV.

Figure 3. CBF sensitivity as a function of tissue ATT across protocols under three assumptions. (a) Percentage; (b) Fixed; (c) Fixed interval. Each assumption has CBF maps estimated by GKM (left), GKMmvc (middle), and their difference (GKMmvc - GKM, right). Units: error% per 1% aBV.

Figure 4: Estimates for Perfusion, BAT and tissue T1 obtained from a healthy human subject using our combined VSI and MRF-ASL framework.

Figure 1: Composition of a single repetition time (TR) and details of the pulse sequence used in our work. The TR is varied across acquisitions by varying the Post Labeling Delays.

Phantom error maps (% error) reflecting residual eddy current effects before and after VS gradient optimization, for DHRS and BIR8 VS labeling schemes. The bar graph reflects the mean absolute error across the phantom volume. Significant improvements are seen for the x, y, and z gradient axes for DHRS. The BIR-8 errors are small for each gradient axis, and little or no improvement was seen.

Human perfusion-weighted images acquired before and after VS gradient optimization for both DHRS and BIR-8 VS modules in x and y gradient directions. z direction was not acquired due to subject intolerance. Marked EC effects are seen for the standard DHRS approach, which are qualitatively improved after optimization. BIR-8 errors are small for each gradient axis, and little or no improvement was seen.

Figure 4: Bland-Altman plots and within-subject coefficient of variation (wsCV) obtained from the global myocardial perfusion measurements in the in vivo datasets

Figure 3: Boxplots of MBF measurements in vivo averaged across the two intrasession acquisitions for the different breathing strategies (BH=breathhold, SB=synchronized breathing and FB=free breathing) and post-processing steps (O=Original dataset, P= after pairwise registration, G= after groupwise registration, DD = after detecting and discarding outliers). *Indicates statistically significant differences between datasets (p=0.05).

Single slice segmentation example. Segmentations are displayed in blue contours. (A) M0-image with whole kidney contours as a result from U-Ne2. T1-map and perfusion map with cortical contours as a result from U-Net3, corrected with U-Net2 output. (B) Reference cortical contours manually drawn by observer1. (C) Cortical contours manually drawn by observer2. Good agreement between the three different cortical contours can be seen; the bright cortical perfusion signal is captured by all contours, assuring accurate mean cortical RBF calculation.

Graphic representation of our segmentation cascade for renal cortical voxel extraction. U-net1 roughly localizes both kidneys on full FOV M0-images. This information is used for image cropping, to remove background which permits faster image registration. U-net2 segments fine whole kidney masks on the cropped and registered M0-images. U-net3 is predicting cortical voxels from cropped and registered T1-maps. In a last step, the predicted cortical and whole kidney masks from U-Net2 and U-Net3 are combined to remove eventual erroneous cortical predictions outside of the kidney.

Figure1. Network structure of the denoising model. The network is consisted of a local pathway and a global pathway. The local pathway is combined with five layers of 3D convolutional layers and one 2D ConvLSTM layer and a 3D convolutional layer at the end. Inputs to the network are several perfusion image series and the outputs are the denoised perfusion image.

Figure 3 Model Outputs with different input time steps. The predictions are produced with 20, 15, 10 and 5 time steps as input using the corresponding model.

Figure 3: Simulation results: a) labeling efficiencies of the different slice selective gradients and B₁⁺ amplitudes, as well as the mean B₁⁺ amplitude of the 5 subjects after shimming and the required B₁⁺ amplitude. b) normalized labeling slice profiles of the different slice selective gradients: with lower gradient amplitude the labeling slice gets thicker and vice versa.

Figure 5: Example of one subject with 7mT/m and the individual adjusted shim.

Fig. 3. Representative percent signal change map of 3D EPI-pCASL and spatial SNR of perfusion map at 7T. (a) CBF map by 3D EPI-pCASL with 5mm slice thickness. (b) Table of spatial SNR of perfusion map by thickness. Despite limitation of coil coverage and B0, B1 inhomogeneity, the perfusion map from 7T MRI have consistent quality with that of 3T. Also, spatial SNR is similar with that of 3T. Because lowered labeling efficiency by hardware limitation could not be calculated, percent signal change map is represented instead of CBF map.

Fig. 4. Functional activation map by (a)perfusion and (b)BOLD activation. For the both perfusion and BOLD, activations were shown simultaneously at visual (occipital lobe) and motor cortex (white arrows). Although the sensitivity of perfusion activation is lower than that of BOLD, spatial localization is much specified by perfusion activation map.

Figure 2: PCASL labeling efficiency as a function of B_1mean measured in ICAs with G_mean 0.25mT/m. For 4 of the arteries, the labeling efficiency maximized ~0.6 at a B_1mean of ~0.4uT. Note that the maximum achievable B_1mean was limited in RICA of subject 3 compared to others due to increased B₁⁺ inhomogeneities in the labeling plane.

Figure 4: Whole-brain CBF maps acquired with a 2D-EPI PCASL sequence at 7T. The PCASL sequence parameters were G_max=3.5mT/m, G_mean=0.25mT/m, Hanning RF pulse duration/separation=800/1800s, nominal FA=35°, labeling duration/PLD=1.5s, matrix=64x64, resolution=3x3x3mm³, 23 slices, TE/TR=12/4500ms, GRAPPA factor=2 and TA=6min. With a labeling efficiency of 0.6, the average CBF in the cortex was about 66mL/100g/min.

Figure 5 - Top: CBF map slices from the experiment, calculated from 10 ASL repeats with the following motion circumstance: a) No deliberate motion. b) Nodding motion. c) Nodding motion with ESPIRiT based motion correction. Bottom: d) Correlation between the partial volume corrected voxel GM measured when the test subject was deliberately nodding v.s. when they were remaining still. e) Equivalent correlation after ESPIRiT based motion correction was applied.

Figure 2 - 1) The raw k-space data from all four shot acquisitions from all control image repeats are used to calculate coil sensitivity maps. 2) These are used to reconstruct full images from each of the individual shots. 3) These are then registered by maximizing mutual information. 4) The corrected tag and control image are registered by minimising the standard deviation of the subtraction, found to be the best metric for this problem. 5) The motion corrected data is used to calculate CBF maps.

Figure 1: SLIDER acquisition scheme. (a) SMS-EPI pCASL acquisition sequence for each set of low-resolution images, with constrained slice dependent background suppression. (b)Acquisition scheme for SLIDER2, SLIDER3 and SLIDER4, different number of scans with slice shift are needed, 2 for SLIDER2, 3 for SLIDER3 and 4 for SLIDER4. Then high-resolution images can be reconstructed.

Figure2: SLIDER super-resolution reconstructed ASL images and high-resolution ASL images in the axial direction. The reconstructed images well follow the high-resolution images, with an improved SNR. The green and red boxes and corresponding zoomed images show two typical slices with SNR improvement.

Figure 3. (A) Perfusion weighted signal ΔM [% of M0] from Subject 3 from TI = 250-1650 ms; (B) T1 map (ms) (C) CBF, aBAT tBAT and RMSE maps using the single-TI model, the Buxton’s model and FPOCK model.

Figure 1. Pulse sequence diagram for the SALL-FAIR sequence (A) and the illustration of relative spatial positions of RF pulses (B).

Figure 1 – Anatomical and perfusion templates included in BATS

Figure 5 – Simulated ASL signal at different PLDs

a. Tissue segmentation volume from the input ground truth obtained from the ICBM 2009a Nonlinear Symmetric atlas¹². Values in the table in b. are assigned to each of the regions for each quantity volume, relaxation times are for 3 Tesla. c. Example DRO output control, label and corresponding difference images. d. Four transversal MRI slices output from ExploreASL for two DROs with same perfusion but with and without motion. tSNR = temporal signal-to-noise ratio.

Histograms (200 bins) of the randomly sampled input parameters (a-e). Labelling efficiency values were limited to a maximum of 1, resulting in an excess of samples with value 1.0. Histograms (200 bins) and fitted gaussians of the output mean CBF for grey matter (f) and white matter (g). Input-output graph showing the values of the mean CBF in each ROI for both the resampled ground truth, and the results form the Whitepaper equation calculation (h), error bars show one standard deviation.

Figure 3. Signal intensities (open circle) and their regression analysis results (solid line) over 20 PLDs are shown for 3 runs of 2D pCASL in one session (a), and for all 5 sessions across 5 weeks (b). Curves were shifted horizontally slightly so that they all peak at LD+PLD = 5000 ms.

Figure 2. perfusion difference images of 20 PLDs ranging from 0.2 s to 7.8 s from 3 runs (a, b and c) of 2D pCASL sequences in one session, compared against images from one run (d) of another session showing similar signal intensities.

Figure 1: ¹⁸F-FPIA PET/MRI acquisition. Axial images of a representative patient of the standardised uptake value corrected by the tracer uptake in the whole blood (SUVc), the contrast agent plasma/interstitium transfer rate constant K^trans and the extracellular extravascular volume fraction v_e resulted from the post-processing of DCE-MRI data, the apparent diffusion coefficient (ADC) map from DWI-MRI acquisition, the relative contrast agent mean transit time (rMTT) derived from DSC-MRI and the cerebral blood flow (CBF) derived from the sequence of ASL.

Figure 2: The least absolute shrinkage and selection operator (LASSO) was applied to extract a composite vector - the most significant tumour grade predictor - from 25 parameters determined by analysis of simultaneously acquired PET and MRI data. Five-fold cross-validation was performed to select lambda minimum to give the minimum cross-validated error (A). The weighted sum of 3 selected features (B) gave the Grade predictive Vector GpV, which showed an AUC and accuracy of 1 in the discrimination between low and high-grade gliomas (C).

Figure 1. A 32-year-old woman (Patient #2) with IDH1-mutant GBM. ROIs of tumoral (red) and mirrored (green) regions were manually drawn on T2-FLAIR image referring to contrast-enhancing regions in post-contrast T1 image, and successively overlaid onto multi-parametric perfusion maps (lower row) to measure the maximum and mean values within ROIs. As indicated by red arrows, the CBF map shows obviously high values in regions of large vessels (confirmed by low ATT values) within tumor found in T2-FLAIR image, whereas the aCBV map doesn’t display elevated perfusion.

Figure 2. A 57-year-old man (Patient #4) with IDH1-wild type GBM. ROIs of tumoral (red) and mirrored (green) regions were manually delineated on T2-FLAIR image and overlaid onto multi-parametric perfusion maps (lower row). Ring-enhancing lesions in post-contrast T1 image are coincident with ring-like regions with hyper-perfusion in CBF and aCBV maps. Prolonged ATT values especially in the occipital region are observed.

Figure 1. Two runs of 3D pCASL with TSE-CASPR and three dynamics of 3D pCASL with GraSE in a 26-year old healthy volunteers (a) and a 71-year old GBM patient (b).

Table 1. Intraclass correlation coefficient (ICC) and its 95% confidence interval (CI) for 2 runs of 3D pCASL with TSE-CASPR and 3 dynamics of 3D pCASL with GraSE.

Figure 2. Representative images of PA (A-F), WT (J-L) and squamous cell carcinoma (M-R). The first column was the fat-suppressed T2-weighted image (A, J, M). The second (B, H, N), third (C, I, O), fourth (D, J, P), fifth (E, K, Q) and sixth (F, L, R) columns were TBF maps at different PLDs (PLD=1025ms, 1525ms, 2025ms, 2525ms, 3025ms), respectively. WT (H-L) showed relatively highest TBF, followed by squamous cell carcinoma (N-R) and PA (B-F). The unit of TBF is ml/100g/min. TBF= tumor blood flow; PLDs= post-labeling delay times; PAs= pleomorphic adenomas; WTs= warthin’s tumors.

Figure 1. Line graphs showing the changes of TBF at different PLDs. TBF= tumor blood flow; PLDs= post-labeling delay times; PAs= pleomorphic adenomas; WTs= warthin’s tumors; MTs= malignant tumors.

Figure 5: Perfusion weighted images using background suppression with two inversion pulses and the proposed multi-breathhold acquisition scheme and a free-breathing acquisition respectively. Arrows indicate areas with subtraction errors under free-breathing which are likely misinterpreted as perfusion signal.

Figure 3: Three exemplary slices of perfusion-weighted images using two inversion pulses and different inflow times TI.

Table 1 Demographic data Note: Mean±SD

Figure 1. Manual regions of interest(ROIs)were placed on the CBF map by using the aligned anatomical image as guidance. ROI was about 20-100mm². (A)bilateral superior frontal gyrus and posterior central gyrus .(B)bilateral superior temporal gyrus.(C)bilateral occipital lobe(D)basal ganglia (bilateral thalamus, globus pallidus, putamen, caudate nucleus )

Figure 1 Results of two‐sample t‐test between dementia patients and HC subjects. The above: PDD minus HC. The below: AD minus HC. Red color represents the increased perfusion, while the blue color represents the decreased perfusion (P_FWE < 0.001).

Figure 2 The normalized CBF differences between the PDD patients and the AD patients. Compared with AD, the PDD patients showed decreased CBF in the bilateral putamen and right SMA, as well as increased CBF in the right MTG, and right precuneus.

Figure 3. Scatter plots of median GM CBF vs. clinical and cognitive scores. R and p-values are obtained via Pearson’s correlation.

Figure 2. Scatter plots of median grey matter CBF vs. white matter lesion volumes, as assessed by T1-weighted (left panel) or T2-weighted (central panel) images. The right panel shows the correlation between CBF and the T1/T2 volume ratio. R and p-values are obtained via Pearson’s correlation.

Figure 1

Figure 1: (A) Group averaged CBF obtained from 436 subjects (age=50.4±3.5 years, 54% female) from the NHLBI Coronary Artery Risk Development in Young Adults (CARDIA) study; (B) A periventricular white matter (PVWM) region of interest (ROI) obtained by thresholding the group averaged map in (A) to CBF<12.5ml/100g/min.

Figure 2: (A)-(D): Scatter plots comparing the (A) whole brain (WB), (B) gray matter (GM), (C) white matter (WM) and (D) periventricular WM (PVWM) CBF obtained from the two sessions using single-PLD (shown in blue) and multi-PLD (shown in red) ASL. The black lines correspond to the unity line and the titles show the correlation coefficients. (E)-(H) and (I)-(L) show Bland-Altman Plots corresponding to (E & I) WB CBF, (F & J) GM CBF, (G & K) WM CBF and (H & L) PVWM CBF using the two sessions of single and multi-PLD data respectively. The wsCV values are in the titles of the plots.

Figure 1: Brain regions with reduced CBF in PAH patients over control subjects. The sites with reduced CBF in PAH patients included the bilateral frontal white matter (a, b), left anterior (c), mid (d), and posterior (e), and right anterior (i) cingulate, left insula (f), bilateral corona radiata (g, h), and bilateral prefrontal cortices (j, k). All images are in neurological convention (L = left; R = right). Color bar indicates t-statistic values.

Figure 2: Negative correlations emerged between CBF and mood symptoms and positive associations between cognition and CBF in PAH patients. Correlations appeared between mood scores and CBF values of the right insula (a, c), and left basal forebrain (b, d), and between MoCA scores and CBF values of the bilateral basal forebrain (e, f) and right insula (g). Figure conventions are same as in Figure 1.

Figure 4 Comparison of CVR estimates. Box plots shows the median cerebrovascular reactivity (CVR) for each of the three pCASL (pseudocontinuous Arterial Spin Labelling) Parameter sets. CVR is the percentage change in cerebral blood flow per mmHg partial pressure of end-tidal carbon dioxide (PetCO₂).

Figure 3 ROI Selection and Blood Velocity Results. A: Phase contrast velocity region of interest (ROI) selection of left/right (L/R) internal carotid arteries (ICA) and vertebral arteries (VA). B: Average maximum velocity of feeding arteries (N=9, F=4, mean±s.d.) at three carbon dioxide (CO₂) levels: hypocapnia (hypo, -10mmHg below baseline), normocapia (normo) and hypercapnia (hypo, +10mmHg above baseline).

Figure 2: Structural and flow maps of an example patient. Occlusion are seen in right MCA and PCA territories in MRA of CoW and hyperintense focal spots in T2 FLAIR (marked by yellow arrows). In all PET and ASL modalities, CBF and CVR in the right MCA and PCA territories were lower than the right side.

Figure 1: Structural and flow maps of an example patient. Severe stenosis and occlusion are seen in left ACA and MCA territories in MRA of CoW and hyperintense focal spots in T2 FLAIR (indicated by yellow arrows). In all PET and ASL modalities, CBF and CVR in the left ACA and MCA territories were lower than the right side.

(a) A Sample slice for T1 weighted MPRAGE, Label, Control, and M0 image. (b) A sample slice for 5-PLD ASL images. The PLD ranges from 500 to 2500 ms with an interval of 500 ms. (c) The perfusion map of the checkerboard phantom at PLD = 1000 that is embedded in a lower slice of the phantom. (d) The simulated perfusion signal of GM and WM using a 5-delay 3D pCASL protocol.

Resulted CBF (ml/100g/min) and ATT (ms) maps of the digital realistic phantom from 4 different ASL data processing software, including BASIL, Cereflow, MRI Cloud, and ASLtbx, under a normal SNR condition and a low SNR condition.

Fig. 1. Relationship between the P3 parameters and numbers of correct response. (a) rank correlation of between P3 amplitude and numbers of correct response. (b) rank correlation between P3 amplitude and numbers of correct response

Fig. 3. Relationship between P3 latency and global topology of brain networks. (a) significant rank correlation between P3 latency and global efficiency. (b) significant rank correlation between P3 latency and mean function connectivity. (c) significant rank correlation between P3 latency and characteristic path length.

(a) Bland Altman plot of the selected 10 brain regions of the 5 subjects. (b) The test-retest for kw of a subject.

(a) and (c) CBF and kw maps before caffeine uptake. (b) and (d) CBF and kw maps after caffeine uptake. Averaged in whole brain, a 26% decrease in CBF and 13% decrease in kw are observed.

Figure 2 – comparison between L1-wavelet CS and DCI-net reconstructions at various rates

Figure 1 – Illustration of the architecture of DCI-Net with 2D convolution.

Figure 1: Example of multiparametric MRI maps acquired during the two MRI exams: arterial spin labeling perfusion maps, diffusion coefficient (D) maps, pseudo-diffusion coefficient (D*) maps, perfusion fraction maps and longitudinal relaxation (T1) maps.

Figure 2: a) Bland-Almant plots for the RBF (ml/min/100g), D (×10−3 mm2/s), D* (×10−3 mm2/s), PF (%) and T1 (ms) calculated in the cortex and in the medulla. b) Correlation between cortical mean values of renal blood flow and perfusion fraction averaged for the two exams, with a correlation coefficient r = 0.47.

Figure 2: PCASL perfusion-weighted images of three healthy volunteers acquired under expiratory, inspiratory and free-breathing conditions. One pair and twelve pairs of label-control images were measured in breath-hold and free-breathing examinations, respectively.

Figure 1: Spatial arrangement of labeling and imaging planes on a sagittal image in PCASL measurements. The labeling plane (red) was positioned nearly perpendicular to the pulmonary trunk (yellow arrow) and images were acquired in coronal orientation (green). Time course of ECG-triggered PCASL sequence: labeling duration (τ) was limited to the systolic period and imaging was performed in diastole of successive cardiac cycle by adapting post-labeling delay (PLD).

Figure 2: Brain (a) and prostate (b) PWS and corresponding tSNR maps from one representative subject with four out of all ten slices. Both blood flow (CBF for brain and PBF for prostate) and blood volume (CBV for brain and PBV for prostate) maps were shown. M0 and DIR images were also shown above. Note that in the CBV PWS and tSNR images, the CSF in leteral ventricles display dark as it has negative signal due to a combination of long T2 values of CSF and the higher B1+ scale at the center of the brain.

Figure 4: Averaged blood flow PWS (a) and its tSNR (b), blood volume PWS (c) and its tSNR (d), in ROIs of GM, WM and prostate from four subjects.