Fig. 4: 3D DEPSI scans of the liver, acquired respectively two hours (A and C) and 3 hours (B and D) after intake of deuterated glucose (glucose). In the MRI-DMI overlays (A and B), the liver contours are highlighted. Selected voxels in the liver (C and D) show both the glucose peak (blue marker) and the water peak (orange marker). It can be seen that the glucose peak decreases over time. Note that the glucose signal in the stomach (A, blue marker) can also be detected and that it is much higher 2 hours after intake than 3 hours after intake (B).

Fig. 3: 3D DEPSI datasets of the human liver at natural abundance, acquired with Hamming weighting (A) and without Hamming weighting (B). The left panels show overlays of the ¹H MRI images (Dixon) and the ²H spectra. The right panel highlights 6 voxels located in the liver. The average SNR of the deuterated water signal in the ROI (white box) is 2.4 times higher for the Hamming weighted DEPSI acquisition.

DMI data collected at the indicated stages following the intravenous administration of ²H_6,6’-glucose to a preeclamptic pregnant mouse. Metabolic maps of ²H_6,6’-glucose (A) and its metabolic products ²H_3,3’-lactate (B) and ²H-water (C) are here shown for illustration purposes. The anatomical ¹H image on top of which all ²H data are shown is depicted in (D). Time traces for all metabolites (E) and a zoom into lactate traces (F) are shown for the entire time series.

²H MRS and MRSI spectra arising after intravenous administration of ²H_6,6’-glucose to pregnant preeclamptic and control mice. The columns A and C present non-localized spectra from the ²H MRS and organ-specific ²H spectra localized from the MRSI data (B, D) at similar post-injection time-points. Signals for ²H_6,6’-glucose and its metabolic products water and lactate are indicated by letters G, W and L respectively.

Figure 1. The 4D dynamics of hyperpolarized [2-¹³C]pyruvate, [5-¹³C]glutamate, [2-¹³C]lactate (downfield peak), and [2-¹³C]lactate (upfield peak) in the healthy human brain. The displayed images were obtained after denoising using a patch-based denoising method. The first 10 timeframes, acquired with a temporal resolution of 3 s, are displayed. The upper window level was adjusted to 20% of the maximum intensity voxel for pyruvate and to 100% for lactate and glutamate.

Figure 3. Representative AUC images of [2-¹³C]pyruvate, [5-¹³C]glutamate, and Σ[2-¹³C]lactate signals summed over 20 timeframes with ¹H anatomical images from a healthy brain volunteer.

Figure 1. Imaging mechanism: hyperpolarized ¹³C pyruvate and urea are intravenously injected in a single bolus, then leave the vasculature and enter exocellular space. Pyruvate can enter the cell and be metabolized into lactate, alanine, or CO₂/bicarbonate, evaluating multiple metabolic fluxes, whereas urea is predominately extracellular and metabolic inactive, serving as a perfusion contrast agent. (K^trans, vascular transfer constant; MCT, monocarboxylate transporter; LDH, lactate dehydrogenase; ALT, alanine transaminase; PDH, pyruvate dehydrogenase).

Figure 5. A balanced steady-state free precession (bSSFP) sequence for hyperpolarized (HP) ¹³C urea (A, highlighted area indicates readout duration) produced superior image qualities (SNR, resolution, sharpness) than a single shot spiral gradient echo (GRE) sequence (B). (C): Representative images of sum HP ¹³C signal of a healthy rat overlaid on T₂ weight images: flip angels of lactate, alanine, pyruvate, and urea are 30°, 30°, 8°, ± 25°; 2.5 × 2.5 × 21mm spatial resolution; 4.2s temporal resolution. (D) HP signal dynamics in the aorta and kidneys of a healthy rat.

Figure 2 Representative in vivo human brain ³¹P CSI results from 3D dataset before (A) and after (B) using the SPICE denoising method. The data were acquired using Siemens CSI sequence with the following parameters: TR = 300 ms; TE = 0.35 ms; FOV = 220´220´200; Matrix = 24×24×12; Resolution (nominal) = 9.1×9.1×16 mm³; Average (weighted) = 10; bandwidth = 5000 Hz; vector size = 1024; scan time = 32 min. The red lines outline the brain edge and the left panels show the representative in vivo ³¹P spectra taken from 4 voxels within the black box as shown on the left panels.

Figure 1 Representative phantom ³¹P CSI results from 3D dataset before (A) and after (B) using the SPICE denoising method and Siemens CSI sequence and parameters: TR = 200 ms; TE = 1.6 ms; FOV = 240×240×80; Matrix = 12×12×4; Resolution = 20×20×20 mm³; Average = 4; bandwidth = 5000 Hz; vector size = 800; scan time = 8 min.

Figure 4: In-vivo cellular sensitivity and resolution of DCs. 25k and 250k cells injected subcutaneously 2cm (A) and 5cm apart (F). Signal from 25k cells is hidden within the signal from 250k cells, with indistinct peaks from the corresponding signal intensity profile (B). Signals are resolved using 3D imaging (35 projections) after window leveling to the lower signal (C). MRI confirms signal voids from 25k (D) and 250k cells (E). Signal from 250k, liver/gut, and 25k is seen with 2D (G) and 3D (H) images.

Figure 2: Cellular detection limits of DCs labeled with Vivotrax and Synomag-D with and without TAs. Images are of a single sample scanned at a time. No TAs limits detection of Synomag-D to 25k cells. 12k and 6k cell samples were not detectable (black X). With TAs, 6000 cells are visible for both Vivotrax and Synomag-D. 500k samples using TAs for labelling were not prepared (gray squares). Extracellular iron contributes to overestimation of signal with Vivotrax. More signal is detected using Vivotrax with TAs than with without TAs.

Figure 3. (A) Anatomical image of the lower-leg muscle showing the voxels used for linescan and ¹H-MRS (B) LASER spectrum (TE/TR=25/2000ms, 3*1.8*1.5mm³) and the carnosine signal fitted using LCModel (C) Low resolution Z-spectrum (D) Low (blue) and high (red) resolution Z-spectra showing Cr / PCr peaks (E) Zoom showing the Car and APT peaks from two different rats. All peaks were detected with a significance p-value < 0.001, except for the APT peak for rat 2 (p = 0.017). Full circles: rat 1, Empty circles: rat 2.

Figure 1. The “peak finder” is an algorithm that automatically detects peaks and assesses their significance. The original data (A) are flipped for peak detection (B). A baseline is estimated for each peak using Gaussian Processes (GP) and a chi-square value χ²_data is calculated (C). The code randomly generates pseudo-experiments assuming a Gaussian distribution around the baseline with a variance in accordance with the confidence interval returned by the GP (D). The fraction of simulated curves with χ²_pseudo < χ²_data is taken as p-value of the detected peak. The output is shown in (E).

Fig. 3: Quantitative ²³Na concentration in the human kidney at 7T in all directions for not respiratory sorted data, inhaled and exhaled state. Slices in each direction were chosen to have the kidneys visible (transversal: 38, coronal: 53, saggital: 33, for a 80×80×80 matrix).

Fig. 2: Phantom (a) and in-vivo (b) data in all directions. First column shows the not respiratory sorted data corrected with the unsorted B₁⁺ map, second column shows the inhaled state data corrected with the inhale sorted B₁⁺ map, third and fourth column show the exhaled state data corrected with the exhale sorted B₁⁺ map and the unsorted B₁⁺ map respectively. Slices in each direction were chosen to have the kidneys visible (transversal: 38, coronal: 53, saggital: 33, for a 80×80×80 matrix).

Figure 2: Example of multimodal image segmentation and registration. A) Top: During the ¹H session, apparent diffusion coefficient (ADC) and fractional anisotropy (FA) maps were obtained. Bottom: During the ²³Na session, total sodium concentration maps (TSC) and low-resolution MPRAGE images for registration were acquired. B) ROIs were obtained from the high-resolution MPRAGE, then registered to the TSC using the low-resolution MPRAGE and to the diffusion maps using the diffusion-b₀ image.

Figure 3: Boxplots of TSC, ADC, and FA distributions in mTBI and control (CTL) groups. Boxplots exclude the three oldest mTBI patients who lacked age-matched controls. FWM FA and caudate TSC were lower in mTBI patients compared to controls (MW, ♦: p<0.05). Note that nine out of 12 ROIs revealed lower median TSC values in mTBI patients compared to controls. gWM: global WM; CC: corpus callosum; BCC: body of CC; GCC: genu of CC; SCC: splenium of CC; CR: corona radiata; FWM: frontal WM; PWM: posterior WM; gcGM: global cortical GM; CA: caudate; PA: pallidus; PU: putamen; TH: thalamus.

Figure 4: In-vivo results. Sodium images are shown in the first row, which has 3 mm³ isotropic resolution. Proton T₁,T₂, B₁⁺ maps from the three directions are shown beneath the sodium images.

Figure 1: Sequence diagram of the 3D simultaneous technique. The bars on top show the sodium excitations with constant flip angle and the proton MRF train with various flip angles. The details of the sequence for different segments are shown in the corresponding cards on the bottom.

Figure 1. Natural abundance DMRSI data in the right lower leg positioned parallel with B₀. (A) Coronal T1w image. The white rectangle indicates the slice shown in panels B and C. Transversal T1w image with (B) segmentation of Tibialis Anterior (TA), Soleus (SOL) and Gastrocnemius Medialis (GM) muscles, and (C) overlaid with DMRSI data. Spectra from selected voxels in TA (D), SOL (E) and GM (F), as indicated in panel C, showing the signal from HDO at 4.7 ppm (x-axis range: 0-10 ppm = 457 Hz).

Figure 3. Natural abundance DMRSI data in the right lower leg positioned at an angle of approximately 45° with respect to B₀. (A) Sagittal T1w image. The white rectangle indicates the slice shown in panels B and C (cropped in AP direction to size of open part of rectangle). Transversal T1w image with (B) segmentation of TA, SOL and GM muscles, and (C) overlaid with DMRSI data. Spectra from selected voxels in TA (D), SOL (E) and GM (F), as indicated in panel C, showing the signal from HDO at 4.7 ppm (x-axis range: 0-10 ppm = 457 Hz).

Fig1. 7T 31P MR spectrum acquired from the occipital-parietal region of a 30 years old female using a partial volume coil. Note that the intracellular Pi signal is split into two separate peaks with 0.1 ppm difference in chemical shift.

Fig 3. 7T 31P MR spectrum averaged over eight healthy subjects. The data were acquired from the posterior brain using a partial volume coil. Note that the nearly symmetric line-broadening at Pi in reference to PCr (blue trace).

Figure 1. 3D DMI measurements at baseline and 0-130 min after oral administration of [6,6’-²H₂]glucose (scan 1). (A) Overlay of last DMI data set (130 min) on transversal Dixon water image, with equal scaling of all spectra. (B-D) DMI spectra at baseline and up to 130 min after intake of deuterated glucose from voxels in the anterior (B) and posterior (C) side of the liver and the stomach (D).

Figure 3. 3D DMI measurements 68-205 min after oral administration of [6,6’-²H₂]glucose (scan 2). (A) Overlay of a DMI data set (160 min after intake) on transversal Dixon water image, with equal scaling of all spectra. (B-D) DMI spectra recorded from 68-205 min after intake of deuterated glucose from voxels in the anterior (B) and posterior (C) side of the liver and the stomach (D).

Figure 3. Interscan test-retest repeatability of PCr, α-ATP, β-ATP, and γ-ATP peaks from quantitative, non-localized ³¹P MRS of 7 human subjects. The corresponding CVs estimated by Monte Carlo Simulation are shown in hatched bars. The dashed horizon line represents a 5% threshold.

Figure 2. Comparison of Trueness (RMSE), linearity (R²), and relative uncertainty (CV) of α-ATP, β-ATP, γ-ATP, and PCr peaks quantified by HSVD, AMARES, conventional peak integration, lineshape fitting, and simple summation. The RMSE, R², and CV values are normalized to (0, 1) range so that the reference point (center) represents the worst performance (0) and the circle represents the best performance (1). The normalized RMSE, R², and CV values of α-ATP, β-ATP, γ-ATP, and PCr peaks are distributed evenly along the angular axes (circle).

Figure 3. The spectra acquired in vivo from a pulse acquire and a spin echo acquisition with the dual band pulse aimed to hit the PDE and PME spin groups from the entire head, where B₀ shimming is optimized over the yellow box. All MR spectra were displayed using a 10‐Hz Lorentzian line‐broadening filter. Note the well refocused signals of PME and PDE.

Figure 2. Spectra acquired from a phantom with a spin echo (A) and a pulse acquire (B) acquisition. The dual band pulse at the frequency offset 3 resulted in a full refocusing of the three peaks of PE, PC and GPC. All MR spectra were displayed using a 10‐Hz Lorentzian line‐broadening filter. The yellow box is the shim box.

Figure 2. Modelling vs phantom validation. A) SNR was computed on magnitude images using an identically positioned ROI (red block) on a signal and pure noise image according to the NEMA standard method. Parameters: TR=6.5 ms, TE=3.0 ms, NSA=217, recon matrix=32x32x7, recon voxel size=21.88x21.88x25 mm³, acquisition time=5 min. The Philips bSFFP sequence incorporated an elliptical shutter. B) Slight variation in the profile of the bSSFP signal amplitude versus flip angle was observed when comparing the theoretical modelling to the phantom tests.

Figure 3. Pilot in vivo acquisitions. T1w images displayed for anatomical reference (A-C), ⁷Li-MRI images showing the ⁷Li distribution in the brain (D-F). The ⁷Li-MRI in vivo images acquired from three different test patients showed signal variations across multiple sites. The Philips data was interpolated to 21.88x21.88x25 mm³.

Fig. 3. (a)-(b): The ¹H/¹⁹F/²³Na/³¹P MR measured images obtained with the specific phantom. (a) only scanned by using the quadruple-nuclear RF coil. (b) Combined the quadruple-nuclear coil and triple tuned local ¹⁹F/²³Na/³¹P receive coil. (c)-(d): SNR maps for the ¹H/¹⁹F/²³Na/³¹P imaging. (c) Without the local triple-tuned RF coils. (d) Combined with the local triple-tuned ¹⁹F/²³Na/³¹P receive RF coils. The position of the local receive RF coil was indicated by the white dotted-line, and the SNR of the region indicated by the red arrow is significantly improved.

Fig. 4. (a)-(b): The ¹H/¹⁹F/²³Na/³¹P MR measured images obtained with the independent nuclide test phantom. (a) only scanned by using the quadruple-nuclear RF coil. (b) Combined the quadruple-nuclear coil and triple tuned local ¹⁹F/²³Na/³¹P receive coil. (c)-(d): SNR maps for the ¹H/¹⁹F/²³Na/³¹P imaging. (c) Without the local triple-tuned RF coils. (d) Combined with the local triple-tuned ¹⁹F/²³Na/³¹P receive RF coils. The position of the local receive RF coil was indicated by the white dotted-line, and the SNR of the region indicated by the red arrow is significantly improved.

Figure 2. Sodium T_2short and T_2long obtained from bi-exponential fitting of CPMG (a) for samples collected from the foot/leg and abdomen from control patients and (b) for all samples from control and diabetic patients (samples from foot/leg and abdomen from control patients are presented together in one control group). ***P-value<0.001.

Figure 1. Sodium CPMG echo integrals of the foot/leg skin samples from control (green) and diabetic (red) patients as a function of echo time (TE). Squares represent experimental data; solid lines correspond to the calculated fitting function.

Figure 2: One representative image slice of one test case

Top: Quantified FI, RI (S = 5) and the CNN output image with L₂ and L_GDL. Bottom: Corresponding TSC error maps within the whole brain.

Figure 1: Representative image slice of one patient in its different versions. Full image (FI) and reduced images (RI) with varying factors S (2, 4, 5 10), and the corresponding CNN out images for L₂ and L_GDL.

Figure 2: Boxplots of TSC in mTBI and control (CTL) from linear regression analysis. Boxplots show the 1^st, 2^nd (median), and 3^rd quartiles (box), ±95% (whiskers), and means (*) of the TSC distributions of the controls and mTBI patients for global grey and white matter (GM, WM, respectively) using linear regression. The boxplots excluded the three elder mTBI patients who lacked an age-matched control. Note that TSC in the GM and WM was decreased in mTBI when compared to controls (age-adjusted analysis of covariance, ♦: p<0.05).

Table 2: Correlations between clinical and TSC measures. Direct and partial Spearman correlations (r) and p-values are presented for the association of TSC with the scores from the Brief Test of Adult Cognition by Telephone (BTACT) and Rivermead Post-Concussion Symptoms Questionnaire (RPQ) without (direct) and with (partial) adjustment for the elapsed time from injury to imaging, respectively. Only statistically significant correlations are shown (p<0.05).

Figure 2: One representative ²³Na MRi slice from one patient with the ROI segmentation for the different quantification methods A: vials, B: CSF, and C: VH.

Figure 3: Top row: TSC map from ²³Na MRI of one patient and one exemplary slice with quantification based on A: reference vials, B: cerebral spinal fluid, C: vitreous humor. Bottom row: corresponding error maps of D: absolute TSC difference between A and B and E: absolute TSC difference between A and C

Fig.4: a) Dependence of the longitudinal and the transverse relaxation times on the BSA concentration. In the samples with a BSA concentration of 0% and 1% w/v, no sodium TQ signal was observed. Consequently, the TQTPPI FID was only fitted by a mono-exponential SQ signal decay. B) Dependence of the sodium TQ signal on the BSA concentration. The sodium TQ signal increased almost linearly with the BSA concentration. c) Zoomed TQTPPI spectra for different BSA concentrations.

Fig.2: TQTPPI FID fit results of the samples containing amino acids or α-LA. The TQ SNR of each sample was less than three and thus the TQ term in the TQTPPI FID fit function was omitted. Hence, the TQTPPI FID fit function consisted only of a mono-exponential SQ signal decay. Exemplary TQTPPI spectra of these samples are shown in Fig.3.

Figure 1. Co-registered masks of ROIs including thalamus (a), dorsal striatum (b), caudate nucleus (c) and insula (d) superimposed on the averaged ¹⁷O MR images.

Figure 2. Dynamic H₂¹⁷O concentration (C_H2¹⁷O) in all examined ROIs. Blue dashed lines indicate the ¹⁷O₂ inhalation time period.

Figure 2. Representative T₂ and T₂*-weighted images of the rat brain after MCAO surgery in the day 28 after injection of viral vectors or PBS. Immunofluorescent staining was done for ferritin (FerrH) and activated microglia/macrophages (CD68). eGFP signal was read at the fluorescent microscope with a wavelength of 488 nm. Areas of ischemic lesion were delineated on T₂-weighted and propagated to T₂*-weighed and to the histological images. Green arrows point to areas of the transgene expression near SVZ outside of ischemic lesion.

Figure 3. Prussian Blue histological staining for detection of iron accumulation in the zone near lateral ventricle and SVZ (top row) and in the ischemic lesion zone (bottom row). Green arrows point to the cells with iron accumulation in the zone near the SVZ; red arrows point to the cells with iron accumulation in the ischemic lesion. Scale bar corresponds to 50 µm.

(a) Experimental setup consisting of three AML sample tubes lying on top of the ²H coil (green circle), with overlaid ²H spectra. (b) Single voxel ²H spectra for the untreated (U) and cisplatin-treated samples (T24 and T48) two hours after injection of [6,6’-²H₂]glucose. (c) Time course of ²H-labeled species 10 minutes after injection of [6,6’-²H₂]glucose. Error bars are the standard deviation across 6 samples. Stars indicate a statistically significance difference between T48 and U, and T48 and T24 (p<0.05)

Representative TUNEL and H&E stained sections (10x and 200x magnification) from untreated and cisplatin-treated cell pellets, which were fixed in formalin for at least 96 hours after scanning. Formalin fixation may have caused pellet size shrinkage.

Figure 1. Two different models tested (Top row) accounting for different circulation characteristic. Result of fitting (bottom row) using the two different models

Figure 3 representative k21 and k22 parameter maps from each mice group. all maps from respective parameter shares the same colorbar range.

Figure 3. In vivo results. A. TTC staining of the post-stroke mouse brain. B. CMRO₂ maps overlayed on T₂-weighted image. C. Histogram of CMRO₂ values in the ipsilateral and contralateral hemisphere.

Figure 1. Intermediate results in the simulation study at high SNR. A. The entire cost function (blue line) and the error from data consistency term (red line) in each iteration. B. Intermediate CMRO₂ mapping results in each iteration.

Figure2. Box plots show of the liver stiffness and ADC values in the area receiving high dose and low dose of stereotactic body radiotherapy (SBRT). Liver stiffness and ADC values were significantly increased in both areas after SBRT. Exclude cases where ROI cannot be obtained sufficiently.

Figure3. Case1: 79-year-old man with HCC at S6. The dose of SBRT was 50Gy/25Fr (PTV-95). The liver stiffnesses after SBRT were higher than those before SBRT in both area receiving >30Gy and area receiving <15Gy.

Fig. 5: ²H images for different times of one volunteer for water, glucose and Glx with and without referencing to a baseline water measurement (A). Zero-filling to twice the original resolution was applied. In (B), signals for different voxels are shown for water, glucose and Glx for the same volunteer. The colors of the frame label the corresponding metabolic images in (A). (C) shows data from four different volunteers (different colors) for water, glucose, Glx and lipid/lactate for summed voxels over different regions of the head.

Fig. 2: The results for three volunteers of a whole brain FID measurement after the oral administration of deuterated glucose for four resonances (A). The temporal resolution of the measurement is 2 minutes. In (B), several spectra from volunteer 3 are shown at different time points. The time points are labeled with different colors.

Figure 1. (left) Photograph of the solenoid coil which can be switched between 1H and 19F using a mechanical switch (arrow). (centre) Photograph of the phantom placed in the centre of the coil. (right) Schematic of the phantom with vials filled with water (blue), fomblin (orange) and a single tube of PFOB (red).

Figure 2. (left) S11 plots of the 1H and 19F switched-mode RF coil. (center) Spin-echo 1H image from the phantom, (right) Corrsponding spin-echo 19F image.

Figure 2: Sequence diagrams of (A) conventional enhanced SISTINA with DISCOBALL UTE and MGRE MQC readout, and (B) optimized enhanced SISTINA with FLORET UTE and MQC readouts. Three hard RF pulses with flip angle α₁=α₂=α₃=90°, preparation time τ=10ms, evolution time δ=50us. 12-step phase cycling scheme is applied to separate SQ- and TQ-weighted signal. Rewinder gradients are applied immediately after every readout to avoid the effect of residual magnetization on high-order coherences. Spoilers are used to dephase the residual transversal magnetization after the last MQC readout.

Figure 4: The first-echo UTE, SQ, and TQ images after B₀ and B₁ correction obtained from three measurements depicted in Table 1. (Left) UTE, (middle) SQ, and (right) TQ images of (A) conventional enhanced SISTINA, (B) enhanced SISTINA with FLORET UTE readouts and MGRE MQC readouts, and (C) optimized enhanced SISTINA with FLORET UTE and FLORET MQC readouts. (D) the image differences of UTE, SQ, and TQ images: (left) the UTE image difference between the second and third measurements; (middle) the SQ and (right) TQ image differences between the first and second measurements.

Fig. 1 Add-on ²³Na MRI platform equipped with the 1.5T ¹H system. (a) Schematic overview of the system. (b) 1.5T magnet. (c) ¹H transmit/receive coil without the RF shield. (d) ²³Na transmit/received coil without the RF shield. (e) ¹H pick-up coil. (f) Crossband convertor.

Fig. 5 Demonstration of ¹H/²³Na imaging. (a) Saturated saline and water phantom. (b) Live mouse. (c) SNR measured in phantom and in vivo studies. (d,e) ¹H-MRI and ²³Na-MRI for (d) phantom and (e) live mouse. The red and blue squares indicate the noise and signal areas evaluated for the SNR measurement, respectively.

Figure 1. Illustration of the proposed reconstruction method. Tissue-based structural information from high-quality proton images is incorporated into the reconstruction of normal image features using a motion-compensated GS model. Sodium-dependent novel features (e.g., lesions) are reconstructed from the residues using a sparse model.

Figure 3. Phantom and healthy subject results. (A) Phantom ¹H reference image, ²³Na FFT and proposed reconstruction, with corresponding calibration curves. Note that both curves are linear, but the FFT reconstruction had much larger standard deviation (5.3%) than the proposed method (2.8%). (B) Images from two healthy subjects. Note the marked improvement in image quality by the proposed method. (C) Regional TSC values obtained using the proposed method, which were consistent across different brain regions and subjects, demonstrating the reproducibility of the proposed method.

Fig.3: Phantom measurements of the transverse relaxation acquired with MRF (top) and the reference method (bottom). Here, both a mono- and a biexponential FID were fitted to multi-echo GRE images (FA=90°;TR=200ms;TE: 32 samples in [0.35,54.6]ms;T_acq=1h47min). The red masks in the T_2l^* and T_2s^* maps indicate pixels that were determined to relax monoexponentially (biexponentially in T₂^* maps). For the reference measurement, this distinction was determined by the goodness of the fit R², whereas MRF determined this differentiation intrinsically in the dictionary matching step.

Fig.2: Measurements of T₁ and ΔB₀ in a phantom with ²³Na-MRF (top) and the references methods (bottom). The measurement time of ²³Na-MRF was 1h4min. Furthermore, the 1^st frame of the MRF data after compression into the new basis (²³Na image) and a spin density (SD) image are shown. The T₁ reference was acquired via fitting a monoexponential model to the data of an IR sequence with varying inversion time (FA=90°,TR=290ms;TI = [4,20,40,70,100,130,170,200,250]ms,T_acq=5h48min). ΔB₀ was determined via a phase difference measurement (FA=45°;TR=25ms;TE=[0.55,5.55]ms,T_acq=7min).

Figure 1 Temperature dependence of dielectric constant (A) and loss (B) measured in the BST-BT uHDC disks at three measurement frequencies.

Figure 3 Representative B₁+ maps in axial orientation (left panels) and B₁+ profiles along the vertical dashed lines acquired under (A) control without the uHDC disks, (B) with one uHDC disk and (C) two stacked disks. The ratio images between (D) one uHDC disk versus control and (E) two stacked disks versus control. The yellow rings represent the 15-cm ¹⁷O RF coil.

Fig. 1: Schematic and layout of the dedicated ¹⁹F MRI RF coil system. Printed circuit boards of the Tx and Rx coil elements are shown on the left hand side. S_ij curves, circuit diagrams and the component values can be found in https://github.com/alibaz/F19Coils

Fig. 4: S matrix of the 6 channel anterior part of the 19F MRI coil system measured with phantom and in vivo loading. In the phantom S_ii<-20dB and S_ij<-16 dB, and in the animal an S_ii<-14dB and S_ij<-18 dB was achieved.

Detection sensitivity and image quality comparison of balanced spiral spectroscopic imaging and 3D balanced UTE SSFP using a 0.1μl sample of PFOB. The spectroscopic imaging sequence is capable of detecting the sample with NSA=1, whereas the imaging sequence starts picking up the signal with higher NSA's. In terms of image quality, spectroscopic imaging shows significantly less spreading than its imaging counterpart, offering accurate localization of the target.

a) all phase-encoded spectroscopic imaging sequence with balanced gradients. Encoding can be done in 2D or 3D.

b) pseudo-spiral k-space sampling scheme in 2D. 3D encoding is done by stacking 2D trajectories.

a) The TQTPPI pulse sequence consisted of three 90° pulses and an additional 180° refocusing pulse. In every phase step both the evolution time and a phase are incremented. b) A fully sampled TQTPPI FID and the corresponding NUS TQTPPI FID are shown. The undersampling pattern was sinusoidal Poisson gap sampling c) A Fourier transformation leads to a sparse spectrum consisting of SQ and TQ signals. The Fourier transform of the NUS FID shows an increased noise level and reduced SQ and TQ signals. The spectrum and FID can be recovered using CS.

Variation of undersampling factor in the range of 2 to 16 using simulated data. The parameters were: A_TQ/A_SQ=10%, SNR=70, T_2S=39ms and T_2F=10ms. a), b) and c) show A_TQ/A_SQ, T_2S and T_2F, respectively. The deviation to ground truth was less than 5% for undersampling factors up to 5. For T_2S there is a small offset for NUSF.

Figure 2: Interleaving sequence working scheme and in-house built circuit for safety testing

Figure 5: In vivo B0 map image of the thigh of a healthy volunteer at[CTR1] 7 T, 1H spectrum and interleaved 31P spectrum.

Figure 2. Trajectories of the proposed RSD scheme. (A) The disc, filled with the interleaved TPI trajectories shown in Fig. 1A, is rotated along the x-axis to form a sphere. (B) A half-filled sphere with TPI discs.

Figure 3. Sodium phantom images on 3 orthogonal slices. The red numbers on the larger bottles count the bright rings of the annular pattern in each bottle. The dotted yellow rectangle includes the smaller tubes with the center dark spot.

Figure 3. A) Metabolite amplitudes of the first ¹³C MRSI dataset: top row - acquired data; lower row - TRI de-noised data (core rank: 7x6x7). A T₂-weighted image shows the positions of the brain, tumour and a [¹³C]urea phantom. Relative amplitudes are indicated by the colour bar, except for [1-¹³C]pyruvate, which is decreased ten-fold. B) As for (A), but for the second dataset (core rank 3x3x3). C) The first three fibers in the spectral dimension of the tensor decomposition in (B), shown as spectra.

Figure 1. A) Carbon-13 MRSI spectra from the head of a rat with an orthotopically implanted patient-derived glioblastoma xenograft. Rows 7 to 10, columns 13 to 22 and chemical shifts 209 ppm to 141 ppm are shown. Spectra were baseline corrected and line broadened to 25 Hz. B) The same data but reconstructed with a core tensor size of 7x6x7 and processed similarly. C) The residual of the acquired data minus the TRI data with no line broadening or baseline correction.

Figure 3. In vivo cell tracking experiments (a) reconstructed with fully sampled k-space. (b) was reconstructed with 21% k-space data, retrospectively selected from (a). (c) was the time frame following (b) reconstructed with the same undersampling ratio. The red circle highlights a location where the cell is not visible in (a), but visible in (b), and had exited the slice in (c). A few other cells that are visible in (b) and (c), but not in (a), are highlighted in green.

The cartesian sampling scheme illustration. A fully sampled k-space (a) can be retrospectively undersampled at different ratios, such as (b) and (c).

Figure 5. (a) ¹H axial images acquired with 2D GEMS sequence. (b) Left: the ²H signal of Control scan and that of interleaved ²H- ¹⁷O scans (nt=20). Right: the ¹⁷O signal of the Control scan and that of interleaved ²H- ¹⁷O scans (nt=100). (c) Top: 20 deuterium spectra from interleaved ²H- ¹⁷O scans (nt=1×20); Bottom: 20 oxygen-17 spectra from interleaved ²H- ¹⁷O scans (nt=5×20).

Figure 1. Schematic circuit diagram of the triple-tuned coil design. For ¹H (no DC), modulate C2 and C5 to tune and match to 698MHz. For ²H (no DC) tune and match C2 and C5 to 107MHz. For ¹⁷O, make sure setup for ²H is established, apply DC and modulate C3 and C4 to tune and match to 94MHz.

Fig.3. PaVARS CEST images and signal evolution for healthy human brain at 3T. A, B and C are MTRasym images at 3.5ppm, Amide signal and NOE signal using LD quantification, respectively. Images from the 5 modules of PaVARS and the first principal component of the five modules are shown. D: T2w FLARE image showing the drawn ROI of putamen, caudate and globus pallidus. E shows the five PaVARS Zspectra and the corresponding Lorentzian fitted lines and the zoomed residual (LD) spectra. F plots the LD signal changes as sat. module #, with the error bar indicating standard deviation across subjects.

Fig.2. Phantom experiments of PaVARS CEST profile for three different metabolites, from left to right: Glu., pCr. and Cr. The upper row was obtained with constant B₁ =2uT, while the bottom row was obtained with an alternating B₁ of 0.6uT and 2.8uT. As seen, PaVARS allows a unique profile for Glu., pCr. and Cr., despite that pCr. and Cr. both contain guanidinium amine protons and the Glutamate amine resonances at similar offset frequency.

Figure 4: Group-average normalized metabolite maps overlaid on Montreal Neurological Institute-152

Figure 1: Single-subject normalized metabolite maps overlaid on the Montreal Neurological Institute-152 template

a. MRI images of mice injected with BHM;

b. TTC and HE staining;

c. CNR curves of BHM and Omniscan.

a.The particle size of BHM was about 67 nm.

b.The particles were regular and uniform spheres under transmission electron microscope

Figure 2. Zoom-in of experimental Z-spectra around the gas-phase peak. Z-spectra for (A, 340 fM) small and (B, 40 fM) medium MBs at different B₁ strengths. Z spectra obtained with a constant B₁ (5 μT) at different gas-volume concentrations with (C) small, (D) medium, and (E) large MBs. (F) Z-spectra of MBs with different sizes at a constant concentration of 400 fM. Signal loss increases with B₁ as well as with gas-volume concentration. HyperCEST efficiency is maximized for MBs with a 2 μm diameter.

Figure 1. Simulations of MBs hyperCEST contrast. MB samples are characterized by the gas-volume concentration (μL/L) and average diameter (μm). These characteristics are directly related to the exchange rate and participation rate of ¹²⁹Xe. Using the FHC solution, signal loss at different B₁ strengths is shown. Signal loss generally increases with B₁ and gas-volume concentration. The MB size must be optimized, and MBs lose efficiency as hyperCEST agents at sizes in excess of 2 μm.

Figure 2 A. MR water-selective cartilage scan (3D_WATSc) for neocartilage assessment at 3,6, 9, and 12 weeks after surgery (Left: No-treated group, Right: Treated group). B. Quantitative R1 relaxometry rate of the different hydrogels measured by T1 mapping scan in vivo. C. Macroscopic and histological observations of cartilage harvested during weeks 6 and 12 (white box: degenerated cartilage).

Figure 1 A. Schematic illustration of the experimental protocol. B. Structural observation and MRI characterization of hydrogels. C. T1-weigted MRI images and quantitative R1 relaxometry rate of SMNP-KGN hydrogel at various concentrations.

(A) Time lapse CSI images of a normal mouse brain after the injection of 50 mg unlabeled glucose. FOV 20 mm x 20 mm x 6 mm, Tr=2 s, 512 spectral FID points. Data is shown in magnitude mode without averaging. (B) Kinetics of the peak at 5.2 ppm (C) Contour map of the 5.2 ppm peak

Dose dependence of (A) glycogen and (B) glucose kinetics for 0, 25, and 50 mg doses by following the peaks indicated in (C)

MR maging of the epileptogenic region to evaluate the targeting capability of BMC NPs.

In vivo assessments of BMC NPs on hypoxia attenuation, cell apoptosis, and neuron protection.