Figure 4: Representative T1, T1_w, PDFF, and R2* maps from a 64-year-old patient with NAFLD. The mean T1, PDFF, and R2* of reference were 779 ms, 15.5%, 89 s^-1, respectively, while the mean T1, T1_w, PDFF, and R2* measured from MT-ME were 785 ms, 643 ms, 15.5%, and 85 s^-1.

Figure 3: (A) Bland-Altman plots demonstrated good in vivo repeatability of the multiparametric mapping from MT-ME. Theoverall inter-scan differences for T1 and T1_w were less than 3%, while the differences for PDFF and R2* were less than10%. (B) The regression analysis of T1, PDFF, and R2* measured with MT-ME and references showed good agreement on the 14 volunteers. The Rs was 0.990, 0.976, and 0.953, respectively.

Figure 3. T₁ maps from a breath-hold study on a NAFLD patient are presented. The first row shows the T₁ maps from the composite reconstruction, the second and third rows show the T₁ maps from the proposed fat/water separated T₁ reconstruction.

Figure 4. Comparison between the clinically available PDFF mapping sequence (IDEAL) and the proposed technique on a NAFLD patient. The first row shows the 3 slices generated from the proposed technique and the second row shows slices from an IDEAL scan at similar anatomical locations.

Figure 5: 2D MOLLI is a breath-held T1 mapping method, that is often corrupted by the motion artifacts. Shown is the comparison of T1 maps generated with 2D breath-hold MOLLI and 3D free breathing IR-SoS in the liver. T1 measurements in liver parenchyma from two methods show good agreement. Proposed method shows estimation inaccuracies in longer T1 species due to lack of sequence optimization.

Figure 4: Shown is the comparison of T1 maps from different slices generated with 2D MOLLI and 3D IR-SoS in the knee. ROI measurements show good agreement between the proposed method and the reference T1 maps.

Figure 2: Demonstrating the training (a) and testing (b) pipelines for deep learning based T1 mapping. In the figure, the pipelines are shown for a single slice. During training, TI images are obtained with and without truncation of the acquired k-space data along the T1 recovery curve (T1RC). The images from the truncated T1RC k-space are used as input and the T1 map from the full T1RC is used as target.

Figure 5: T1 estimation performance for a subject with a liver lesion for both the proposed DL framework and NLLS fitting with $$$\hat{N}=8,12,16,20,24$$$ and $$$28$$$ and TIs as input. In (a) results for three 9x9 pixel ROIs are given. The height of each bar represents the average T1 value within a given ROI, with error bars representing the standard deviation. In (b), visual examples are provided for qualitative evaluation of the two techniques over the liver lesion with varying TIs.

Figure 3: Abdominal images (3 out of 32 TEs) and corresponding T2 maps obtained from data acquired with the VFA and constant FA RADTSE pulse sequences for subjects with (top) hepatic hemangioma, (middle) liver metastases, and (bottom) hepatocellular carcinoma. The lower SAR in VFA RADTSE increases slice efficiency by 60% (11 vs 7 slices per breath hold). Note the reduced noise in the VFA T2 map in the central part of the liver (arrows) compared to the constant flip angle allowing for better lesion conspicuity with the VFA scheme.

Figure 5: (A)T2 distributions show excellent separation between malignant (22 metastases, 2 HCC) and benign (8 cysts, 1 hemangioma) lesions for constant and variable FA methods. Mean T2: 89.2±16.2ms (malignancies) and 242.9±67.7ms (benign) for constant FA and 89.3±17.1ms (malignancies) and 219.0±58.6ms (benign) for VFA. A t-test showed that the T2 distributions are similar between the constant and variable FA for malignant(p=0.998) and benign lesions(p=0.435). (B)Relative contrast between malignant lesions and adjacent liver is superior for the VFA method(p=0.013).

Fig. 4. T₁ and T₂ maps with mSASHA, SASHA, MOLLI, and linear T₂p-bSSFP in a healthy volunteer at 3T.

Fig. 1. Sequence diagram for multi-parametric SASHA. A series of images are acquired without magnetization preparation, with saturation-recovery preparation, and with both saturation-recovery and T₂-preparation.

Fig. 5 a) Volumetric SA quantification on the reference (top) and proposed (bottom) T₂ maps of Subject #2 obtained from the proposed approach. b) Quantification, for the same subject, inside an ROI manually drawn within the region encompassing the abnormal T₁/T₂ values (marked with a white arrow in Figure 4). Good correlation for both T₁ and T₂ between standard and proposed quantification methods is shown.

Fig 4. Standard T₁ MOLLI and 2D bSSFP (top row) and proposed joint T₁/T₂ (bottom row) maps obtained two patients with clinical findings. Subject #2, diagnosed with acute myocarditis, presented elevated T₁ and T₂ values (white arrows). Subject #3 presented with a myocardial infarction in the mid septal region (yellow arrows).

The in vivo signal phase difference is shown for three slices. Simulations were performed for the sequence parameters to calculate from the T₂ time estimates from the phase-difference. The reference T₂ map and the reconstructed phase based T₂ map for three slices are depicted, showing a homogeneous myocardium for two of the three slices. One slice suffered from signal loss in the septal region likely due to B₁⁺ inhomogeneities.

A) Gradient echo acquisition with quadratically increasing phase increments and readout rewinder. B) Simulated phase for a GRE acquisition for different phase increments Φ C) The signal phase is depicted over time for varying T₁ and T₂ times. D) The signal phase and E) the phase difference for a phase increment of ±2° are shown as a function of T₁ and T₂. F) The sequence diagram is shown for multiple slices triggered to the end-diastolic phase for positive and negative phase increment. First all slices with positive phase increment are acquired followed by negative phase increment.

Figure 4. Typical volunteer T1 cine maps. Slice 1 and 2 are acquired and reconstructed simultaneously in MB multitasking, while separately in single-slice multitasking.

Figure 3. Typical volunteer images along cardiac-motion (a), inversion recovery (b) and respiratory-motion (c) dimensions. Slice 1 and 2 are acquired and reconstructed simultaneously in MB multitasking, while separately in single-slice multitasking.

Figure 1: Fast-spin echo imaging and T2-mapping results in a prostate cancer patient. (A) Clinical image, transverse T2w. (B,D) Results from k-t T2 mapping, (C,E) from T2-CSXD. (B,C) Show single echo images at TE=128ms, (D,E) show the resulting T2 maps (grey scale 0...250ms) from fitting all echoes images.

Figure 2: Optimization of T2-CSXD in terms of ΔT2[%] (relative deviation T2-CSXD versus k-t T2) and precision (SD: relative standard deviation). Data for ROIs within the prostate with low mean T2-values (A,B: 83ms) and mid-range T2-values (C,D: 145ms) are shown.

Figure 1: Pulse sequence timeline showing interleaved radial phase-contrast MRI preceding a background-suppressed T₂-prepared EPI readout to simultaneously measure blood flow velocity and T₂ (and SvO₂), respectively. 55 radial views are acquired in each of the five interleaves (each corresponding to a TE).

Figure 2: Pulse sequence workflow: phase-contrast consists of 55 radial views in each interleave. There are five interleaves for 275 total views in one velocity map to determine total cerebral blood flow (tCBF). Each interleave generates an image with a different T₂-weighting, from which T₂ is measured and then converted to SvO₂.

Figure 2. MR parameter maps and MIP of blood vessel images.

Figure 1. Flowchart of the proposed method.

Representative result mid-LV T2map image with LV segmentation lines (green and red) transferred after being drawn on a T2Prep TE_min=15 ms image. This T2map has good consistency and minimal “bad” mapping pixels.

T2Prep T2mapping Sequence. ECG-triggered with a T2prep module, followed by a delay, then a ramped flip angle readout module (bSSFP or FLASH) acquires four successive images at or soon after end-systole.

Figure 4. Phantom study results. A: Estimated T₁ maps from IR-FSE, MOLLI, and ECG-gated IR schemes of 8-8, 5-(3)-5-(3), and 10-(3)-10-(3) with spoiled GRE. B: Scatter plots of T₁ from MOLLI and ECG-gated IR schemes of 8-8, 5-(3)-5-(3), and 10-(3)-10-(3) with spoiled GRE. Solid line represents line of identity and dashed line represents line of regression. C: Bland-Altman plots from MOLLI and 8-8, 5-(3)-5-(3), and 10-(3)-10-(3) schemes with spoiled GRE. Solid line represents mean difference and dashed line represents the 95% confidence interval for limits of agreement.

Figure 5. In vivo study results. A: 3D T₁ maps from 5-(3)-5-(3) scheme with spoiled GRE and 2D T₁ maps from MOLLI. B: Slice view of the 3D T₁ map from 5-(3)-5-(3) scheme with spoiled GRE along the blue dotted line. The estimated T₁ in the apical, mid-cavity, and basal regions of the myocardium were 1165.9 ± 109.4, 1166.1 ± 85.5, and 1256.8 ± 162.6 ms from 5-(3)-5-(3) scheme with spoiled GRE and 1195.6 ± 118.7, 1185.8 ± 116.7, and 1115.0 ± 135.6 ms from MOLLI, respectively.

Figure 3: Scatter plot showing the correlation in lung T₁ measured using the VFA acquisition and Look-Locker acquisition, for both phantom and in vivo data. Regions of interest were drawn in the lung, liver and descending aorta (blood) for each participant. 4 participants were healthy volunteers, 7 were patients with IPF.

Figure 2: Example T₁ maps in a healthy volunteer and patient with IPF using Look-Locker and VFA acquisitions. A similar-slice UTE image is also shown for the patient, to visualise regions of increases lung density.

Figure 1: A 58-year-old woman with invasive ductal carcinoma of the right breast. An example of the synthetic images obtained before contrast agent injection. (a) T2(synthetic); (b) PD map; (c) T1 map; (d) T2map. The ROI showed the quantifications: T1-Pre = 1.71 10^-3ms, SD of T1-Pre = 0.19, T2-Pre = 80.00 ms, SD of T2-Pre = 4.00, PD-Pre = 59.00 pu and SD of PD-Pre = 2.50 in this ER-positive, PR-negative, HER2-negative and Ki-67 high proliferating tumor.

Note. Data are mean ± standard deviation. P values less than .05* were considered to indicate statistical significance.

Figure 1: A 51-year-old woman with invasive ductal carcinoma in the left breast. (a) diffusion weighted imaging (DWI); (b) T2 map; (c) PD map; (d) T1 map. The lesion (arrow) showed quantitative T1 relaxation time (T1), T2 relaxation time (T2), proton density (PD) and apparent diffusion coefficient (ADC) values, and the quantifications by Reader A: T2 value=83 ± 11ms；PD value = 59.4 ± 10.3pu; T1 value = 1.26 ± 0.31 10^-3ms; ADC = 0.85 ^10-3mm²/s and Reader B: T2 value = 83 ± 8ms；PD value = 59.3 ± 10.5pu; T1 value = 1.25 ± 0.30 10^-3ms; ADC = 0.83 10^-3mm²/s.

Figure 3: Receiver operating characteristic (ROC) curves of univariable analysis (a, b) and multivariable analysis (c,d). T1 = T1 relaxation time, T2 = T2 relaxation time, PD = proton density, ADC = apparent diffusion coefficient.

Figure 2. Example of Kety-Tofts standard model parameter maps in a single slice of a breast cancer DCE-MRI exam, zoomed to the tumor region of interest (ROI). (A) and (D) K^trans maps from fits to the signal intensity and concentration, respectively. (B) and (E) v_e maps from fits to the signal intensity and concentration, respectively. (C) and (F) Maps of the estimated noise from each curve in dB (larger values represent noisier fits.)

Figure 1. (A) Example of a simulated image intensity curve with added Gaussian noise from the repeat K^trans, v_e curve fits (N=1000 for each parameter pair and noise level.) (B) Same noise-added samples as (A) in blue, but now black line is fit to the signal intensity curve. (C) Blue circles are calculated contrast-agent concentration computed from the signal intensity samples in (A) and (B), and the black line is fit to concentration samples. In this instance of added noise, the concentration fit resulted in a K^trans with 20% error, while the intensity fit resulted in a K^trans with a 10% error.

FIGURE 1. Log-log plot of R₁ against B₀. Blue: human; Red: rat; Green: mouse. Each symbol represents one study. Size of circle reflects number of subjects (some smaller symbols are occluded by larger symbols). Solid black line: Eq.1. Dashed black line: Eq.2 with R_1,inf=0.213s^-1. The dotted line shows, purely for illustrative purposes, a fit to Eq.2 where R_1,inf was fixed arbitrarily at a higher value of 0.5s^-1.

FIGURE 2. Plot of R₁ against B₀. Each symbol represents one study. Solid black line: Eq.1. Dashed black line: Eq.2 with R_1,inf.=0.213s^-1. Dotted line: R_1,inf.

Table 3. Repeatability and reproducibility of VFA T₁ values in the liver, compared to literature values in phantoms and other organs.

Figure 1. Box plots showing median and inter-quartile ranges of (a) bias, (b) repeatability of VFA T₁ values in all volunteers, averaged over both field strengths for each vendor, and (c) reproducibility calculated across all vendors as well as vendor pairs.* indicates significant difference p < 0.05

Table 2 R2* values for liver measured by UTE and mDIXON-quant by two observers

Figure 2. correlation between R2* values measured by UTE and mDIXON-Quant.

Fig. 1. R₂* as function of nanoparticle concentration at different concentrations of added FeCl₃. Dashed lines are the linear regression lines, its equations and coefficients of determination R² for linear correlations are given in the colors of corresponding symbols and lines.

Fig. 2. R₂* as function of FeCl₃ concentration at different concentrations of added nanoparticles. Dashed lines are the linear regression lines, its equations and coefficients of determination R² for linear correlations are given in the colors of corresponding symbols and lines.

Fig. 1 - Magnitude images of a 25 year old sedated patient obtained with a) 2D GRE, b) 3D Cartesian Dixon, and c) Free-breathing 3D Radial Dixon acquisitions. The 2D GRE and 3D Dixon Cartesian-based techniques exhibited motion artifacts (white arrows) whereas free-breathing 3D Radial Dixon showed noticeably less motion artifacts.

Fig. 2 - Linear regression analysis of the HIC values obtained using (a) 3D Cartesian Dixon vs 3D Radial Dixon (b), 3D Cartesian Dixon vs 2D GRE method, and (c) 3D Radial Dixon vs 2D GRE methods.

Figure 2, Scatter plots show a significantly positive correlation between Fibroscan and T1 values (r=0.398, P=0.04)

Figure 1, Examples of region of interest (ROI) placement in a patient graded as G1S1. Three ROIs were placed in the different area of liver to measure the T1 relaxation time（● represent the ROI ）

Figure 1: T1 map using the (a) preRF B1+ map and the (b) DAM based GRE-EPI B1+ map to correct the nominal FAs in the SPGR sequence. The liver T1 is overall more homogenous and lower using the DAM GRE-EPI B1+ map. (c) MOLLI T1 map for matching slice location. T1 colormap scale [700 1200] ms.

Figure 3: Correlation between masked T1 and B1+ values. (a) preRF B1+ shows a significant correlation, with a positive slope of 26.8 ms/10% change in B1+ responsible for the difference in T1 of 143 ms between the upper and lower part of the liver. The DAM GRE-EPI (b) removes most of the B1+-related variation from the T1 map showing a negligible residual correlation between T1 and B1+.

FIGURE 2. Flowchart of the proposed algorithm. a: Multi-echo gradient echo images with different flip angles; b: candidate phasor solution classification; c-d: Pixels were then classified into “smooth” and “non-smooth” and the “smooth” pixels were grouped into different sub-regions according to the spatial connectivity; e:Phasor solution of each sub-region were determined using dual flip angle prior information and combined as in f. g: Fat water components under different FA, and water components under different FAs could be used for further T1 quantification.

FIGURE 3 Timing diagram of the pulse sequence. The sequence consists of three blocks. A: B1+ mapping block based on DREAM for T1 mapping (Optional). B & C: multiple echo GRE with 3D acquisition using k-space linear reordering under different FAs. The equally spaced echoes are acquired with monopolar readout mode. The echo spacing (∆TE) are kept the same in B and C, but the echo times of C are moved shTE to B. The pre-dephasing gradients indicated by the blue trapezoids have the same amplitudes and ramp-up time. For simplicity, the slice selection and phase encoding gradients are omitted.

Figure 3, Correlations (Spearman’s) between R1ρ and LIC at different FSLs.

Figure 1, An example of quantitative T1ρ maps with UTE-T1ρ sequence at different FSL.

Figure 3, Correlations (Spearman) between R1ρ and LIC for RARE-R1ρ (a) and UTE-R1ρ (b).

Figure 2, Comparison of mean values of RARE-R1ρ and UTE-R1ρ. Error bar indicate the standard deviation.

Figure 4: Scatter plot of T2 estimates from 21 neoplasms. (A) T2 estimation from the gProCo (which does not take into account each component’s slice profile) underestimates the T2 for the two hemangiomas from the subjects in Figure 3 (arrows). These two lesions fall within the range of malignancies (red box) determined from malignancies without PV. (B) The csProCo corrected signal model gives good separation between the hemangiomas and malignant lesions. Note that the two hemangiomas (arrows) now fall in the expected range of benign lesions.

Figure 3: (A) T2-weighted images from RADTSE-VFA for three subjects where the imaging slice was chosen to be at the edge of the lesion, thereby affected by partial volume. (B) The gProCo model under-estimates the hemangiomas whereas the csProCo, which takes into account individual slice profiles for liver and lesion, yields T2 estimates that match the expected T2 of the lesion (based on T2 values in the center of the slice, where PV is minimized).

Figure 3. Results from two healthy volunteers demonstrate the feasibility of combining SMS with CSE-MRI to achieve rapid, whole-liver, motion robust tissue quantification. The far-left image shows the physical locations of the simultaneously excited slices. The middle set of images shows the multi-echo separated images, followed by the estimated PDFF and R2* maps for both slices.

Figure 4. Comparison to a reference 3D CSE-MRI shows good agreement with the 2D SMS CSE-MRI protocol for estimated PDFF and R2*. Averages from each region of interest are shown. Note the negative values in the PDFF maps can likely be attributed to noise related bias common in low SNR CSE-MRI PDFF maps estimated with a fully complex signal model (see reference 6).

Figure 2: The top row presents T₁ maps (in ms) of the knee of a volunteer after B₁ correction with increasingly undersampled AFI-based B₁ maps (bottom row).

Figure 1: T₁ maps of the knee of a volunteer before and after the B₁ correction with fully sampled AFI-based B₁ maps. Notice the drop of the signal at the edge of the FoV that is corrected by the B₁ map. The images are scaled between 0 and 1500 ms.

Figure 3: Parametric tissue maps obtained with the proposed method.

Figure 1: Generated T1 and T2 maps using the proposed approach and measured average values inside the T1 and T2 spheres (points). The dashed line represents identity with respect to the known phantom values.

Fitting results showing the initial signal strength (proportional to the transverse magnetization, M₀) and R₂* maps for the fully sampled (R=1), twice under-sampled (R=2) and four-fold under-sampled (R=4) acquisitions in three orthogonal planes. Identical scaling was used for all M₀ [normalized signal intensity] and R₂* [ms-1] maps.

Magnitude images and phase maps of a sagittal slice. Fully sampled (R=1, top), two-fold (R=2), and four-fold (R=4) under-sampled (bottom) datasets are shown for each of the 5 echoes (left to right), prior to R₂* fitting.

Figure 3 – Representative T2* and PD maps estimated from 2 short (A and B) and 2 long (C and SR) acquisitions.

Figure 1 – Schematic representation of the super-resolution (SR) T2*-weighted acquisitions and model-based super-resolution reconstruction (SRR). 5 UTE Spiral VIBE datasets, consisting of 2 TEs each, were acquired with high in-plane and low through-plane resolution, while rotating around the frequency-encoding axis over angles of 0°, 36°, 72°, 108° and 144°. A model-based SRR framework, including a mono-exponential T2* relaxation model, was used to estimate high-resolution (HR) proton density (PD) and T2* maps directly from the series of low resolution T2*-weighted images.

Figure 4: Healthy volunteer scan of TSE Dixon sequence for water only images with corresponding T₂ map.

Figure 3: T_2water values for TSE Dixon compared to MRS for FFs of 0, 30 and 40% and agar concentrations of 2% (blue), 3% (green) and 4% (red).

Figure 2: k-Space orderings used for data collection. The a. linear side-to-side (LSS) and b. linear alternated center-out (LACO) are machine-provided orderings, used only in fully-sampled acquisitions. The two customized orderings we evaluated are c. balanced center-out (BCO) and d. center-random (CR). For the same SP and accelerated factor (AF), e. BCO and g. CR differ only on how the phase-encoding positions are ordered in each block. Both share the same initial positions, but in f. BCO the following points are closed to each other, while in h. CR the following points are random.

Figure 5: T1rho maps of a representative human knee joint (medial slice on the top, lateral slice on the bottom) obtained with different sampling patterns (SPs), such as optimal SP and Poisson disk (PD), and both customized orderings: balanced center-out (BCO) and center-random (CR).

FIGURE 1. Schematic comparison of the conventional model fitting and combined deep learning network SuperMAP with joint spatial-temporal under-sampling.

FIGURE 2. T1rho maps from 3 echoes using combined network SuperMAP and with single CNN network (RF 10.66), and the reference T1rho maps from eight fully sampled echoes.

Figure 4 Volunteer scan results from 7T. a) shows the nominal B₁ map, and b) shows the B₀ inhomogeneity map. As expected for 7T scanner, the B₀ inhomogeneity was much more severe compared to 3T. c) shows the NRMSE after fitting the echoes. Red arrows indicate the area with large inhomogeneity. Similar to the previous results, Prep6 yields the best result.

Figure 1 Schematic of different T_1ρ preparations. For all preparations, 90 degree pulses had 400 us pulse duration, and all non-spin-lock RF pulses (blue) had the same amplitude.

Figure 3. Representative human T1ρ mapping results of different TSL sets. (A) TSL=0+. Patellar (PAT) cartilage ROI is shown in red, and muscle (MUS) ROI is shown in green; (C~H) Representative T1ρ maps obtained with TSL_sets 1~6, respectively. There is little difference between these maps. To demonstrate the spatial fidelity of these methods, the zoom-in figures of a small area (rectangle in (D)) are shown (B), each corresponding to TLS_set1 to 6 from left to right.

Figure 1. Representative phantom T1ρ mapping results. (A) Magnitude image of the center slice of the phantom, with labeled tubes #1-6 and ROIs; (B-C) B₀ and B₁ maps of the slice at -3.6 cm with non-uniform B₀ and B₁ distributions; (D-I) T1ρ maps of the same slice using different TSL PC schemes of TSL_sets 1~6, respectively. Their corresponding x- and y- line profiles from tube# 4 (as shown in (D)) of the T1ρ map are shown in the bottom row. The y-profile is shifted by 2-pixels for better visualization.

Figure 4, The mean T1, T2, and T1ρ relaxation times in phantoms for conventional and MRF methods. The dashed red line represents identity (y = x), while the black line is a linear fit of the data.

Figure 3, Phantom T1, T2, and T1ρ relaxation maps for conventional methods and the proposed T1ρ-MRF sequence.

In vivo $$$T_1$$$ map of a volunteer’s hand. The main structures visible in the reference anatomical image (2D GRE) are retrieved in the $$$T_1$$$ map. In this case, a shorter TI (25 ms) was chosen to better characterise the structures with a short $$$T_1$$$ .

Schematic representation of the Look-Locker-based $$$T_1$$$ mapping sequence. The interleaved scheme allows to change the number of slices without impacting the scan time. Nonetheless, more slices signify longer TIs, which can impact the retrieval of short relaxation times. The use of a saturation pulse was chosen to avoid waiting for full $$$T_1$$$ recovery, hence reducing the acquisition time.

Figure 2. Fat- and B1- corrected T1W is estimated using a two-step fitting. In the first step, data from pass 3 are used in a non-T1 weighted multi-echo SGRE fitting to determine an initial point for the joint estimation. In the second step, the derived signal model is used in a non-linear least squares fitting of all data to estimate parameters, including T1_W, T1_F, signal amplitude, signal phases (shared pass 1&2 / independent pass3), R2*, PDFF, B1 and B0. *For acquisitions without fat, B1 can be calculated analytically¹⁰ and fixed in the following joint estimation (shown in green).

Figure 5. T1_W estimates in gel agar phantom experiments (A) showed good agreement with the phantom design values (B) and decent agreement with STEAM-MRS measurements (C). Global B1 errors were introduced by manipulating the transmit gain (TG) at scan time. Plots (B,C) show that T1_W estimation bias is greatly reduced by the proposed simultaneous B1 estimation, but not entirely removed.

Summary of measured R2* values for both volunteers for vendor A and vendor B.

Top: phantom setup with body coil. Middle: sample ROI measuring R2* in one of the phantom’s vials. Bottom: T2 weighted image illustrating positioning of phantom with volunteer.

Figure 2: Experimental dataset: (A) GRE MRI of the OC in all anatomical planes. Here the optic tracts (OT) and optical nerves (ON) are indicated, together with the B₀ direction (yellow arrow). (B) The MR transversal images of the first three angular measurements before (top row) and after coregistration (bottom row) (left). Using the transformation matrix, the direction of B₀ per angular measurement were calculated (right).

Figure 5: Orientation dependence of the parameters in M2 (β₀^M2 in the top row, β₁^M2 in the middle row and β₂^M2 in the bottom row) for the simulated data without (A) and with myelin water (B) contribution, and experimental (C) dataset. It is observed that the angular dependency is removed for β₀^M2 and β₁^M2 and transferred to β₂^M2 for all dataset and fibre’s dispersions. Importantly, the negative β₂^M2 values at small angles observed in the experimental dataset are only replicated if the myelin pool is added to the simulations.

Figure 2 – R₂*(A) and T₁ (B) estimates as function of residency time (range 100-500ms) and f_MW (range 2-20%) with fixed B_1eff=100%. R₂* increases as f_MW increases due to higher contribution of the myelin-water compartment (short T₂*) in the two-compartment model. Similarly, a decrease in T₁ is observed with increasing f_MW due to higher contribution of the myelin compartment (short T₁). C) Quantification of R₂* and T₁ variations as a function of residency time and myelin water fraction.

Figure 3 – A) T₁ dependence on B_1eff for different residency times (columns) and the three analysed strategies (rows). ESTATICS shows greater T₁ variation for low residency time =100ms, whereas per contrast shows high T₁ variation for long residency time = 500ms. B) T₁ variation as function of B_1eff for different residency times.

Fig.3. Signal profiles of 4 lines across the gel boundary. Dashed lines represent the average values across the lines. Voxels size is 400µm; b) From top: legend, positioning for MP2RAGE and MS-IR-EPI, the 4 lines sampling. Arrows indicate phase (phase 2 for MP2RAGE) direction. Both MPRAGE and MS-IR-EPI show a good profile, mostly affected by the true object shape/partial volume/k-space sampling effect. The difference in T1 estimation is expected, being MP2RAGE protocol suboptimal to assess a 1500ms T1 (our focus is on PSF here).

Fig.4. OTF, PSF and signal intensity profile estimated for an image profile with a gel boundary thickness similar to that of phantom #14. T1=1500ms. The true profile is superimposed. For sake of simplicity, full-Fourier versions were simulated. The estimated MS-IR-EPI and TI=2250 MP2RAGE profiles are close to the true profile. The estimated PSF and signal profile of MP2RAGE for TI=800ms show significant broadening, incompatible with imaging typical cortical myelinated layers

Table 1: The prostate T2 values from the multi-echo spin echo (MESE) imaging sequence and the prostate pseudo-T2 values from Autoref with different reference tissues, averaged over seven healthy volunteers.

Figure 2: The spread in mean prostate T2 and pseudo-T2 for all seven volunteers, both from MESE and Autoref with three pairs of reference tissues.

Figure 2: Box plots for T1, T2, and PD values of different tissues compared between different coils. VIS means considering the vertebral bodies and intervertebral discs together. * indicates p <0.05; ** indicates p <0.01; *** indicates p <0.001.

Figure 1: Quantitative maps of a 26-year-old female volunteer. (A-C) T1 map; (D-F) T2 map; (G-I) PD map. Maps from left to right were obtained using Spine DST, Body, and Flex Large, respectively.

Fig. 3: T_1ρ maps acquired under FSL = 100Hz (A) and 300Hz (B), with TSLs = [1ms, 11ms, 21ms, 31ms, 41ms]. The T_1ρ dispersion, i.e., T_1ρ = T_1ρ(300Hz) - T_1ρ(100Hz), is displayed in (C).

Fig. 2: T_1ρ imaging on a 43-yrs female healthy volunteer. (A) T₂-weighted image for structural information. (B) T_1ρ-weighted images at spin-lock times (TSLs) = [1ms, 11ms, 21ms, 31ms, 41ms] under a spin-lock frequency (FSL) of 300Hz.

Figure.1 T2 histograms and Gaussian distributions of AF and NP.

a-e: Pfirrmann grade I-Ⅴ discs. The peak value of the NP gradually decreased and shifted towards the peak of the AF with the increasing grades.

f: The peak value of the NP was significantly lower than the AF of grade Ⅳ disc.

Figure.3 Receiver operating characteristic (ROC) curves analysis of all quantitative parameters for distinguishing healthy discs from degeneration ones.