Schematization of the processing pipeline used to compute B₀ map in Femoral Cartilage using qDESS. The coil-combined phase difference between the second and first echo is computed using the Hermitian inner product¹⁰. Femoral cartilage is segmented using a deep learning model with the DSOMA framework¹³. The FC segmented 3D wrapped phase difference is unwrapped with PRELUDE¹⁴ and then B⁰ is computed according to eq. 1. The 3D B₀ map is projected onto a 2D space for visualization using DOMSA.

Left (A1) and right (B1) unrolled 2D B₀ projection maps in FC obtained using the WASSR, qDESS and 2-GRE methods for a simultaneous bilateral knee acquisition. A pixel-wise unrolled 2D difference map with WASSR is also displayed. BA plots for the left (A2) and right (B2) knees to evaluate the agreement between the WASSR method and the qDESS and 2-GRE methods, respectively. The plots were made using B0 values sampled from six slices in the FC (2 in the lateral condyles, 2 in the trochlea, and 2 in the medial condyles). In the right knee, registration errors caused artifacts in the trochlea region.

Figure 2. A 15-year old boy with stage III JOCD lesion. (A) The first echo of the T2*-weighted images showing lesion region (arrow). (B) The corresponding color-coded apparent diffusion coefficient (ADC) map. The zoomed in box of the lesion and control regions in (A) depicts the evaluated regions: progeny lesion (light green), interface (red), adjacent parent bone (cyan), distant parent bone (magenta), adjacent cartilage (dark blue), contralateral cartilage (yellow), and contralateral parent bone (dark green).

Figure 3. The average of the median ADC values measured in ROIs at JOCD stage I (n=7), stage II (n=20), stage III (n=12) and stage IV (n=7). Decreased ADC values were observed in patients with JOCD stage IV compared to JOCD stages I-III in the interface, lesion and adjacent parent bone.

Fig.1 Scheme of the MIXTURE (Multi-Interleaved X-prepared tse with inTUitive RElaxometry).

(a) T2-mapping was performed using T2-prepared 3D segmented turbo spin- echo (TSE) with variable refocusing pulse trains.(b) Two images with different TE (TE = 0 and 50ms) were acquired with interleaved acquisition. To obtain the compatible contrasts with routine TSE images, TSE shot#1 did not apply any pre-pulses (as “PDW) and shot#2 applied both SPAIR and T2pre (as “fat-suppressed T2W”).

Fig.4 Simultaneous multi-contrast and quantitative images of MIXTURE from a patient with knee joint pain.

Upper: 57y male, knee osteoarthritis, osteonecrosis, bone bruise, intraosseous cyst.Lower: 65y male, idiopathic osteonecrosis.

A) Axial 3D-UTE images of the proximal patellar tendon in a patient with patellar tendinopathy.

B) Selected voxels for tissue-specific T2* analysis in the degenerative tissue of the patellar tendon. Mean T2* decreased from 19.9±7.3 ms (baseline) to 17.2±5.8 ms (12 weeks) to 16.8±4.9 ms (24 weeks).

C) Mono-exponential T2* maps, on a scale from dark blue (short T2* relaxation times) to red (long T2* relaxation times).

D) Bi-exponential fitting maps, displaying the percentage of short T2* components on a scale from dark blue (0% short T2* components) to red (100% short T2* components).

Relation between the change in T2* relaxation times and symptom severity. Symptom severity was assessed using the validated VISA-P questionnaire (scale 0-100), where 100 represents no pain, unrestricted function and maximum ability to play sports.

Fig. 3. FF histograms and MRI-based FF map in 4 examples of soleus muscle with very similar mean FF value, in 4 different NMDs (DMD, DYS, IMB and IMNM). Values for number of pixels in ROI (#pixels), mean FF (M), median FF (Mdn), FF standard deviation (std), kurtosis (K) and skewness (S) are presented as mean ± standard deviation, of all included muscles or ROIs (indicated by n).

Fig. 1. Illustration of the individually drawn muscle ROIs across 5 slices in the thigh and the leg: AL, adductor longus, AM, adductor magnus, BF, biceps femoris, ED, extensor digitorum, GL, gastrocnemius lateralis, GM, gastrocnemius medialis, GRA, gracilis, PER, peroneus, RF, rectus femoris, SAR, sartorius, SM, semimembranosus, SOL, soleus, ST, semitendinosus, TA, tibialis anterior, TP, tibialis posterior, VI, vastus intermedius, VL, vastus lateralis, VM, vastus medialis.

Figure 2: Images are from A. a representative patients (M, 15 years) from Response group with osteosarcoma in right Tibia. B. a representative patients (M, 14 years) from Non-Response group with osteosarcoma in right femur.

For A and B: a, c) Estimated T1 maps and b, d) Histogram of T1 values in tumor at baseline and follow-up respectively.

A comparatively higher reduction tumor T1 can be observed after chemotherapy for Responder than Non-responder. Histogram of tumor T1 for Responder was low picked at baseline and became more positively skewed after chemotherapy than Non-responder.

Figure 3: ROC curve analysis using statistically significant (p<0.5) histogram parameters of T1 value in tumor a) at baseline and b) its relative percentage changes (Δ) at follow-up.

a) At baseline T1-mean and T1-skewness in combination produced AUC=0.86, sensitivity=88%, specificity=72% in identifying chemotherapeutic response.

b) After chemotherapy, ΔT1-mean and ΔT1-skewness in combination produced AUC=0.92, sensitivity=89%, specificity=92% in identifying chemotherapeutic response.

Figure 1: Images of an ACL-MD patient who is a 61-year-old female with the OSag fs （Oblique Sagittal fat suppression Proton Density weighted image）, sagittal 3D-FIESTA, T1mapping, T2mapping, T2*mapping, arthroscopy and pathology.

Figure 2: The curves of T1, T2, and T2* values in the ACL-MD group.

Figure 5. The structural connection heatmap of meniscus (c) obtained by the automatic parcellation (a) and tractography (b).

Figure 1. The automatic segmentation process used in this study, from the acquired DWI (a) to the 9 different areas of meniscus (e). Both Radial Segmentation (c) and Rotational Segmentation (d) were derived from the binary mask (b). These two methods were further combined to divide the whole meniscus to 9 regions (e). R-R: Red-Red zone; R-W: Red-White zone; W-W: White-White zone.

Figure 1: The schematic demonstration of the CNN framework implementing RELAX. A cyclic workflow was constructed to enforce self-supervised learning. The physics models and additional constraints can be incorporated into the framework to guide the learning of CNN mapping function to extract the latent image parameter maps from undersampled images.

Figure 3: Comparison of T1 maps generated from different reconstruction methods for another testing knee dataset at R=5. The deep learning-based methods, including both MANTIS and RELAX, removed most of the artifacts and showed a similar reconstruction performance, which outperformed conventional constrained reconstruction k-t SLR. The absolute error maps were amplified by five times for display purposes to show the method difference.

Figure 1: A) Coronal slices at different echo time extracted from the 3D uTE. B) Parametric maps from a uTE-CSE acquisition: Fat-water decomposition allowed to reconstruct adiposity map (PDFF) and long T2 species T2* map. C) uTE Sub (weighted substraction of long and short echo image) and porosity maps which respectively indicate average pore size volume and bone porosity, and short T2* map of cortical bone.

Figure 3: Comparison between PDFF measurements obtained by uTE and CSE in different hip bone subregions.

Figure 1. Scheme of the MIXTURE T2 mapping.

T2-mapping is performed using T2-prepared 3D segmented turbo spin- echo (TSE) with variable refocusing pulse trains. Two images with different TE (TE = 0 and 50ms) were acquired with interleaved acquisition. To obtain the compatible contrasts with routine TSE images, TSE shot#1 did not apply any pre-pulses (as “PDW) and shot#2 applied both SPAIR and T2prep (as “fat-suppressed T2W”).

Figure 2. Scheme of the MIXTURE T1ρ mapping.

T1ρ mapping was performed using a twice-refocused spin-lock (SL) prepared segmented TSE sequence with spectral fat suppression. Three images with different SL preparation times (SL = 0, 25, and 50ms) were acquired with interleaved acquisition. Amplitude of the SL pulse was set to 500 Hz. TSE shot#1 did not apply any pre-pulses and later shots applied both SPAIR and SLprep.

Figure 4: Example of a CT-image displaying the lumbar spine (L2-S1). All visible IVDs have been injected with a contrast media visible in white. The contrast media is spread through posterior annular tears and their location is correctly determined by the proposed algorithm in the T2W MR image and highlighted in red.

Figure 3: Confusion matrix presenting the classification results based on validation data in a 10-fold stratified cross-validation. True Positive: 100%, False Positive: 6.7%, True Negative: 100%, False Negative: 0%

FAC maps of a patient with Osteoporosis: mono-unsaturated fat (A), poly-unsaturated fat (B), saturated fat (C) and unsaturated fat (D).

PUF on the second lumbar vertebrae of the three groups

Figure 1 UTE T2* Mapping of cartilaginous endplate. (a) the cartilage endplates of L1/2-L5/S1 were divided into three equal parts from front to back, and the ROI of each part was delineated, including ASEP, MSEP, PSEP, AIEP, MIEP and PIEP. Then, the superior and inferior cartilaginous endplates (SEP, IEP) were completely delineated, that is, the complete measurement was carried out without triserification, and then the T2* value of each part was measured in false color figure b and c

FIG. 2 Bar plots of T2* in the Pfirrmann graded cartilaginous endplate in four groups. Except for Grade I, the T2* value of Grade III-V in the superior endplate was inferior than that in the inferior endplate. T2* value of superior endplate decreases gradually from Grade I-IV. The value of T2* in the inferior and rear end of the superior endplate was inferior than that in the front middle. T2* in the posterior part of the superior endplate was higher than that in the inferior end plate for Grade I-II and opposite for Grade III-IV.

Fig1. Average T2* values of the AS group were statistically higher than those of the control group.

Note. — A total of 30 AS patients with 60 sacroiliac joints; n, numbers of sacroiliac joints

Figure 1(a): The least squares fit of known PEG and HA phantom densities vs. 1H and 31P MRI derived intensities after B1 correction with 3-point calibration. (b). Rat femur 1H images (water+fat suppressed) and 31P images, the same slice 1H & 31P images (blue color). (c). EBM values from Gravimetry and MRI analysis.

Male, 35 years old, participated in the full marathon twice, and his medial half month later his foot was torn horizontally. Fig. ① - ⑧ showed 3d Cube-Pd and Magic (T1WI, IW, T2WI, stir, t1mapping, pdmapping, T2mapping) respectively. ① 3D cube-pd showed linear hyperintensity in the posterior horn of the medial meniscus, magiciw, T2WI and stir sequences showed mild hyperintensity in the posterior horn of the meniscus. figure ⑦ Pd-mapping. The signal intensity of medial meniscus was slightly increased in T1 mapping and figure ⑧ in T2WI mapping.

Male, 45 year old, participated in the whole marathon for 5 times, suffered from patellar cartilage injury. Fig. ① 3D cube-pd shows the absence of the inferior part of patellar cartilage, bone marrow edema under the articular surface, magec T1WI sequence, IW and T2WI show the absence of cartilage, and the bone marrow edema under the articular surface is not clear. Figure ⑦ PD mapping showed partial cartilage defect and slight increase of bone marrow edema signal. T2WI mapping showed that the signal intensity in some areas increased, but no obvious signal increase in bone marrow edema

Figure 2. ROIs on a healthy volunteer (a) and 4 OA patients (b-d). Patient b has OA developed around ROI 3 and 4. Patient c has almost no cartilage left. Patient d has OA in ROI 1 and patient e has OA in ROI 1 and 2.

Figure 4. Fitted pb values comparing healthy volunteer (a) and 4 OA patients (b-d), arranged in the same order as in figure 2. Note that fitted pb values have a similar trend compared to figure 3. Fitted pb values of OA patients are generally lower than the healthy volunteer. Fitted pb values can also be correlated to cartilage regions with obvious OA symptom.

Figure 1. Example UTE subtraction images of patella cartilage (A, B, TE1 = 0.1 ms, TE2 = 2.8 ms) , as well as UTE images at a series of TEs: (C) 0.1 ms, (D) 0.5 ms, (E) 2.8 ms, and (F) 5 ms. ROIs were delineated on deep (yellow) and superficial (blue) layers of patella cartilage in B and copied to C-F to measure the signal intensity. T2* values from single-component exponential fitting curves (G) were 3.08 ms deep patella cartilage and 13.28 ms superficial patella cartilage.

Table 1. Deep cartilage T2* values before and after activity

Figure 2. Representative T2-T2 relaxation spectra of bovine and ovine Achilles tendon obtained from 2D-ILT⁸ (panel a), and those reconstructed from 2D-SMA estimates (panel b) of two-dimensional proton T2-T2 relaxometry NMR data. Mixing time, $$$t_{m}$$$ , is 1, 10, and 100ms. T2 axes in 2D-ILT spectra are displayed in log10 scale (seconds). (2D-ILT = two-dimensional inverse Laplace transform; 2D-SMA = two-dimensional subband Steiglitz-McBride algorithm.)

Table 1. Mean exchange-rate estimates of T2-T2 correlation relaxometry data and the mean apparent and inherent bi-component estimates of proton NMR data of bovine (N=3) and ovine (N=3) Achilles tendon samples. Percentage increase in the apparent estimate value in the presence of exchange is also displayed, computed as: 100$$$\times$$$(apparent estimate - inherent estimate)/ inherent estimate.

Figure 1: Proposed image reconstruction framework. During training, an attacker introduces difficult to reconstruct features (δ) to under-sampled images (x): it maximizes a reconstruction error between x and a fully-sampled image (y), in the presence of δ, given a network parametrized by θ. Next, network's parameters (θ) are updated to minimize a robust training loss, which includes a reconstruction error and a robust training term that further penalizes reconstruction errors in the regions including abnormalities.

Figure 4 Abnormality reconstruction improved using our proposed ART strategy (based on the SqueezeNet classifier). Reconstruction of 4x undersampled MRI obtained using A) baseline Unet, B) proposed approach; C) fully-sampled MRI. D) Difference map between A) and B), shows clearer visibility of cartilage lesions and a better preservation of abnormal signal changes.

Figure.1 Distribution of lumbar vertebral body FF values in different age groups in males and females

Figure.2 The scatter chart of correlation between lumbar body FF value and BMI. Figure A shows that the FF value of male lumbar vertebrae is moderately correlated with BMI (r=0.668, P=0.005). Figure B shows that the FF value of female lumbar vertebrae has no correlation with BMI (r=0.214, P=0.395).

Figure 2

Example images of a 74 year old female patient with progressive multiple myeloma (first diagnosis 5 years and 3 months prior). The focal osteolytic lesion in the left dorsal iliac crest (arrow) is conspicuous in the CT as well as the VNCa image (mean VNCa attenuation 42.3 HU). In the T1w MRI image the lesion shows a homogeneous hypointense signal.

Figure 3

Example images of a 69 year old female patient with multiple myeloma in partial response (first diagnosis 3 years and 1 month prior). The focal osteolytic lesion in the right dorsal iliac crest is visible but not conspicuous in the CT image. The lesion is nicely depicted in the VNCa image (mean VNCa attenuation 7.6 HU). In the T1w MRI image the lesion shows moderate central hypointensity and a markedly hypointense rim which does not represent the typical appearance of an “active” lesion but is consistent with partial remission.

Table 1: Average qMT parameters (N=6 cadaver knee specimens) and biochemical content (N=70 excised cylindrical plugs). WM is wet mass

Table 2: Frequency and percentage of each grade for human cadaver menisci samples (N=161). G1=normal tissue, G2=mild degeneration, G3=moderate degeneration, G4=severe degeneration

Spatial distribution of increased 3T artifact compared to 1.5T at the central template slice. All implants are shown in the top left, with metal-on-poly, ceramic, and metal-on-metal continuing in a clockwise direction. Metal-on-metal and ceramic implants represent the extrema of field-dependent differences in artifact.

Evenly separated coronal slices of the left hip arthroplasty acetabular region template space.

Figure 3. Representative ex vivo human cortical bone sample (first row) and in vivo tibial cortical bone T₁ maps of four healthy volunteers (second to fifth row) generated using the UTE-AFI-VTR (first column), the sf-UTE-AFI-VTR (second column) and the UTE-AFI-STR (third column) methods. Similar appearances on the T₁ maps are seen with each of the three methods.

Figure 1. 3D UTE-based AFI and VTR sequences. The conventional 3D UTE sequence is used for multiple-TR data acquisition (a). The 3D UTE-AFI sequence acquires data with two interleaved TRs (TR₁ and TR₂) to generate B₁ or F_z maps (b). Diagram of a single UTE unit in (a) and (b) shown in (c). In (c), a short rectangular RF pulse with a nominal TE of 32 μs is used for signal excitation and is followed by 3D spiral sampling. The spiral trajectories are arranged with 3D conical view ordering (d). RF = radiofrequency; DAW = data acquisition window.

Cartilage segmentation and quantification using the optimized protocols on both knees of a healthy rabbit. (A) 3D map of the femoral (blue) and tibial (red) cartilage, right knee. (B) Cartilage thickness and volume quantification table, right knee. (C) Histograms of T1 values from each knee joint showing the bilateral similarity. (D) Relaxation map superimposed on one sagittal slice of DESS image.

(A) T1 maps on sagittal axis for three different MP2RAGE test scan conditions (respectively: PAT factor 6, 3, 2, Echo Time: 3.85ms, 4.38ms, 4.38ms), superimposed on the morphological DESS. ROIs are drawn in the cartilage 3-dimensionally using a semi-automatic algorithm for segmentation on the DESS images. (B) Calculated T1 values for three different MP2RAGE test scan conditions.

Fig.4 T2 mapping of knee cartilage near metal plate in the tibial plateau (patient 2).

ME, Multi-echo; DE, Dual-echo

Fig.3 T2 mapping of knee cartilage near metal plate in the tibial plateau (patient 1).

ME, Multi-echo; DE, Dual-echo

Figure 3: Results from the linear regression analysis for the female (red) and male (blue) sub-cohort for a) water T2 vs age, b) water T2 vs. PDFF and c) PDFF vs. age. Water T2 as a function of both age (a) and PDFF (b) was longer in females when compared to the males. Furthermore, water T2 showed a negative correlation with both age (a) and PDFF (b) for both sexes.

Figure 2: Example real spectra including the obtained signal model fitting (fitted model and residual signal): a) 38-year-old female (BMI: 31.3 kg/m², water T2: 33.3 ms, PDFF: 40%) and b) 66-year-old male (BMI: 38.6 kg/m², water T2: 13.9 ms, PDFF: 49%). In the spectra in a) the water peak (4.67 ppm) is more pronounced (lower PDFF and shorter water T2) in comparison to the spectra in b).

Table 2. T1ρ values (in ms) of patients in the operated and contralateral knees. Significantly elevated T1ρ were observed in LFC, LT, MFC and MT.

Table 3. T2 values (in ms) of patients in the operated and contralateral knees. Significantly elevated T1ρ were observed in MFC, MT and TRO.

In order to place the ROIs in the same locations on both T1ρ and T2 maps, ROIs were drawn manually on the PD weighted images and automatically copy –pasted on to T1ρ and T2 maps. The automatically measured T1ρ, R1ρ and T2 TRs of each ROI were plotted in Microsoft Excel. All of image processing described above was performed using software in Matlab (Mathworks, Natick, MA) that was developed and implemented in-house.

In the analysis of the association between VAS and T1ρ, R1ρ and T2 values, VAS was significantly correlated and negatively with the R1 ρ values in central and posterior portion of MFC.

Figure 2. a:The comparison of POS between injured and uninjured group. *P＜0.05. b: The comparison of POS-injury among 3 groups (Group 1: negative, Group 2: positive，Group 3：severe positive). * vs Group 1, P＜0.05; # vs Group 2, P＜0.05.

Figure 3. a:The Pearson’s correlation between POS-injury and IKDC-jnjury. b: The Pearson’s correlation between POS-injury and KT1000.

Figure 2. The 4 echoes of UTE and 6 echoes of GRE images with different TE of extended (a and b) and flexed (c and d) knee. The corresponding T2* mapping shown in e), f), g), and h) respectively.

Table 3. In vivo repeatability and comparison of UTE and GRE T2* relaxation time (in ms) of each compartment for both extended and flexed knee joint.

Figure 2. Excellent T1ρ modeling is achieved for both normal cartilage (A, C) (T1ρ=32.3ms) and abnormal cartilage (B, D) (WORMS=2, T1ρ=40.8ms).

Figure 4. Boxplot of UTE-Cones-AdiabT1ρ values in different WORMS depth groups

Table 1. Comparison of average fat fraction in hamstrings muscles in the injured leg (HI) versus the contralateral leg (HC) across acquisition and quantification methods. Bold and italics indicate statistically significant difference in FF (p<0.05). Other muscle groups (quadriceps, other) were similarly compared but no significant differences were observed. In patients with hamstring autograft (right), substantial fatty infiltration was observed (red arrow).

Figure 2. Fat fraction (FF) maps from quantification methods (rows) for phantom and control at supine and prone positions (columns). Artifact is present in FattyRiot with complex processing, which appears to be position-dependent as it is similar across supine and prone orientations. Monopolar gradient with complex processing appears to reduce the amplitude of the artifact, but the pattern persists. Inline bipolar and FattyRiot monopolar magnitude are most consistent between subject positions.

Fig.2 Comparison of SNR and FA values between SPLICE-DTI with MultiVane-SPLICE-DTI.

Partial Maximum Intensity Projection image (20mm thickness) shows on A (SPLICE-DTI) and B (MultiVane-SPLICE-DTI).

Fig.3 Comparison of ICC (inter-rater reliability and intra-rater reliability) between SPLICE-DTI with MultiVane-SPLICE-DTI in 3 healthy volunteers.

Figure 1: Kaplan-Meier survival curves of the event-free-survival (EFS) demonstrate differences in patients in outcome (a) D*-25^th percentile and (b) Albumin. The significant differences are 0.0081 and 0.027 respectively by log-rank test.

.

Figure 1: UTE-MT imaging in lumbar of a 63‐year‐old male volunteer. Figure A: UTE-MT-ON; Figure B: UTE-MT-OFF; Figure C: UTE-MTR map

Figure 2: Correlation between disc herniation and (A) posterior longitudinal ligament, (B) nucleus pulposus UTE-MTR in 50 subjects.

The DTT abnormalities of symptomatic nerves can be classified into color hue reduction (a), sparse (b) and interruption (c). The visual abnormalities of tractography were found mainly in the extraforaminal (34 cases, 34.3%) and extraspinal (57 cases, 57.6%) region of compressed nerves.

The ROC curve for nerve edema had the highest AUC of 0.929 (95% CI 0.884 - 0.961); the cut point showed sensitivity and specificity of 85.9% and 100%, respectively. In addition to this, the ROC curve for distal FA and nerve diameter both had AUCs >0.80.

Figure 1: In vivo qualitative and quantitative imaging of the lumbar of a 56-year-old female volunteer using the 3D IR-UTE sequence. Figure A: ROI in the phantom; Figure B: ROI in the lumbar vertebrae; Figure C: The proton density map of the spine trabecular bone.

Figure 2: Linear regression in 30 subjects between (A) IR-UTE measured PD and QCT, (B) IR-UTE measured PD and DXA, and (C) IR-UTE measured PD and FRAX score.

Fig.3:Example of comparison between the left and right T₂ maps of FC where B₁ differs substantially between the left and right knee. Each row displays the quantitative parameter (B₁ and T₂ in the top and bottom panels, respectively) 2D unrolled projections of FC for the left and the mirrored right knee along with their pixel-wise difference and the BA plots between left and right parameter values in the six FC sub-regions. Overall, there is a relatively high difference between B₁ in the left and right knee. Applying the B₁ correction improves the expected symmetry in T₂.

Fig.1:Schematization of the pipeline used to process the longitudinal in-vivo bilateral acquisitions. Each time-point was registered to the baseline. The FC was manually segmented at the baseline time-point. For each time-point, the open-source DOSMA framework⁹ was used to compute T₂ map with and without B₁ correction and to visualize the 3D segmented volume projected onto a 2D space². The FC was automatically sub-divided into 3 different layers (total, deep and superficial) and 6 sub-regions (anterior/central/posterior for the medial/lateral sides).

Fig.1. (a) UTE (TE=30ms) images, showing reduced noise with PCA and MDPCA, and (b) improved fitting error with MDPCA in the T2* maps of cropped region corresponding to dashed box region (arrow – posterior horn of medial meniscus).

Table 1. Paired Comparisons of Mean T2* and Coefficient of Variation (CoV) for MEM, PCA-denoising and MD+PCA-denoising Algorithms. Note: * = p < 0.05.

T2 relaxation times of the menisci, both medial and lateral, of healthy volunteers were projected to a 2D plane to create T2 relaxation time maps. The map for Post-exercise was registered to Pre-exercise using Elastix. The mean T2 relaxation time in the medial and lateral menisci, as well as across both the menisci, was calculated.

T2 relaxation times were averaged across both the lateral and medial menisci. There was no significant difference in the average T2 relaxation times between pre-exercise and post-exercise scans for the Exercised and Rested groups. The Exercised leg group had slightly lower T2 times than the Rested group for both scans, but these differences were not statistically significant (p=0.96 for Pre-exercise and p=0.29 for Post-exercise).

Visualization of the fixed and the resulting registered volumes of Scaphoid and Lunate. The black lines depict the boundaries of the moving frames.

Kinematic profile of radial-ulnar deviation of scaphoid-lunate (SL) interval; (a) SL interval constructed by navigating two stablished points at the boundaries of the fixed images. (b) SL Center of Mass (COM) distance.

Figure 4. Synovitis detected by MRI in clinically inactive joints versus clinical synovitis. Values represent number of joints per patient. In many patients, there are additional joints with synovitis on MRI that were clinically unsuspected.

Figure 1. Bilateral elbow and right knee synovitis on post-contrast, water-only Dixon MR image (whole-body view).

Figure 3: Within the whole synovium, (a) MD had a significant positive correlation with K^trans and (b) FA had a significant negative correlation. For each standard deviation increase in K^trans, MD increased by 1.3x10^-4 mm²/s (approximately 7% of the median MD value) and FA decreased by 0.020 (approximately 6% of the median FA value).

Figure 4: Correlations between the diffusion parameters (MD and FA) and K^trans within the 67 osteophyte-adjacent ROIs (defined as the intersection of the synovium segmentation and manually drawn regions around each osteophyte). (a) MD did not have a significant correlation with K^trans (r = 077, p = 0.82) but (b) FA had a strong and significant negative correlation (r = -0.78, p = 0.0041).

Figure 5. Results from the fiber-based probabilistic tractography and Dice Coefficient analysis

Sagittal slices of the diffusion weighted images (grayscale) are displayed and overlaid with the resampled regions of interest (green, see Figure 2), as well as the reconstructed anterior cruciate ligament (cyan blue), estimated based on fiber-based probabilistic tractography. Both segmentation reconstructions are shown including voxels traversed by the two (left), or either (right), tractograms.

Figure 3. Three-dimensional high resolution structural imaging of the left knee

Three views (sagittal, coronal and axial) of the Double Echo Steady State (DESS) scan are shown highlighting the healthy anterior cruciate ligament (ACL) for this patient (top). The middle and bottom views show the placement of the region of interest seeds within the tibia and femur insertions of the ACL (green), as well as the manual segmentation of the ligament (blue) used for assessment of the tractography outputs.

Table 2: Agreement between qDESS and CE-MRI using Gwet's AC2 (95% LOA) for readers overall impression of synovitis and regional gradings

Figure 1: Example cases where radiologists tended to agree on severity of synovitis on CE-MRI, qDESS_Low, and qDESS_High. In the sagittal MR images of the suprapatellar pouch, mild uniform synovial thickening is demonstrated without obvious nodularity (grade 1). In the axial MR images of the medial and lateral recess, synovial thickening is more conspicuous with the suggestion of nodularity in the lateral recess images (grade 2). In the sagittal intercondylar notch images, there is nodular synovial enhancement anterior to the anterior horn of the lateral meniscus (grade 2)