Figure 1. Derivation of revealed regions map. Baseline ventilation binning maps are binarized into ventilation masks for each timepoint (top), and then combined via image registration to classify the image into three regions: newly ventilated (green), ventilated at both time points (“constant” in blue), and regions that were ventilated at baseline but are not at follow-up (“lost”, red). Arrows indicate a prominent persistent ventilation defect in the right lung, and in the left lung, ventilation defects at baseline that are newly revealed at follow-up.

Figure 2. Example histograms corresponding to the classifications in the revealed regions map at baseline (top row) and follow-up (bottom row). Vertical bars indicate the mean value in each region. In this subject, the lost region (red) had a higher mean baseline RBC:Gas ratio than both the constant region (blue) or the overall ventilated volume (gray). At follow-up, the newly revealed regions (green) had a higher mean RBC:Gas ratio than the constant region, but the RBC:Gas ratio of the overall ventilated lung still dropped from 0.15 to 0.11.

Figure 2: Central slices from 3D SPGR ¹⁹F-MR images of inhaled PFP in twelve representative participants. Equivalent slices pre- and post-bronchodilator from the 3D datasets are displayed. The original magnitude images are displayed, where no image processing has been applied other than selection of an appropriate lower windowing threshold to minimise visibility of background noise.

Figure 3: The distribution of %VV measurements from ¹⁹F-MRI images of inhaled perfluoropropane in patients with asthma (N = 14) and patients with COPD (N = 12), displayed beside %VV measurements of 38 healthy volunteers measured during an earlier phase of this study.⁷ Diamonds label the group means, with the 25th percentile, median, and 75th percentiles marked by the three horizonal lines in each box plot. A significant change in %VV was measured between pre- and post-bronchodilator measurements within the asthma cohort, though not for patients with COPD.

Figure 1. Boxplots of HVP and RBC:Barrier measurements from the 1-year follow-up (a,b), and within-patient differences in these metrics over the 1-year monitoring period (c,d). Significant differences (P<0.05) between medication groups are indicated by asterisks. Note that no significant differences were found between groups at baseline (data not shown).

Figure 2. Example parametric maps of ventilation (a) and RBC:Barrier ratio (b) from an IPF patient not taking anti-fibrotic medication (male, age 65) and an IPF patient treated with anti-fibrotics (male, age 60) at baseline and after 1 year. Note the reduction in HVP over time in the no anti-fibrotic case (a, left), with HVP apparently maintained with anti-fibrotic medication (a, right). Similarly, RBC:Barrier appears to decrease over time in the no anti-fibrotic patient (b, left), and recover in the patient taking anti-fibrotics (b, right).

Figure 1- A schematic depiction of the multi-breath sequence (top) in which local signal dynamics of agent wash-in, washout and response to dissolved species saturation (image series within colored borders) is quantified to yield a parametric representation of each voxel’s residual gas volume, ventilation during tidal breathing, and gas exchange.

Figure 2- Depiction of the cross-modality registration scheme utilized for quantifying pre/post functional change. CT images (upper left and 1H MRI (upper middle) are approximately coregistered using Affine transformations and Contrast-Limited Adaptive Histogram Equalization (CLAHE, left). The required transformation is then applied to the HXe image to overlay on the segmented CT (upper right).

Figure 1: Comparison of ventilation parameters of a 67-years-old COPD patient (left column – registration to expiration, middle column – registration to middle respiratory state, right column – registration towards inspiration). Note similar pattern of hypoventilated areas in RVent maps and areas with abnormal ventilation dynamics in CC maps for all three different registrations. RVent / CC: 0.22 ml/ml / 0.54 for registration towards expiration, 0.21 ml/ml / 0.59 for registration towards middle respiratory state and 0.27ml/ml / 0.55 for registration towards inspiration.

Figure 2: Comparison of perfusion parameters (Q_N and QDP) of a 75-years-old CTEPH patient in representative dorsal slice (left column – registration to expiration, middle column – registration to middle respiratory state, right column – registration towards inspiration). Note hypoperfused areas in the left lung, which are marked as defect areas in perfusion defect percentage map. Q_N / QDP: 2.7% / 47.7% for registration towards expiration, 2.0% / 57.6% for registration towards middle respiratory state and 1.9% / 57.8% for registration towards inspiration.

Figure 4: Median T₁ was significantly lower in patients with CTEPH/CTED and PH-lung compared to controls. ∆M0 was significantly higher in patients with CTEPH/CTED than controls.

Figure 2: Segmentations of perfused and non-perfused regions were applied to M0 and T₁ maps, to calculate median M0 and T1 in these regions. Example maps are shown for a patient with CTEPH, a patient with PH lung and a control patient.

Figure 4. Case examples with impeded correlation between Delta-VDP and Delta-LCI. (See red dots in figure 2 and 3)

A) At baseline a large ventilation defect, due to mucus plugging is visible (red arrow). At 14 months follow-up, the VDP has reduced, but LCI stayed stable. Baseline (delta) FEV1: -1.7 (+0.9), LCI: 9.6 (-0.2), VDP: 28.4% (-7.9).

B) At follow-up, an increased VDP is visible (red arrowhead). The LCI decreased (improved). Due to possibly more mucus plugging the LCI is “blind” to closed lung areas. Baseline (delta) FEV1, z-score: -0.1 (-0.7), LCI: 10.4 (-1.8), VDP: 20.4% (+5.6).

Figure 3. Arrow plot of baseline LCI (x-axis) and VDP (y-axis) to follow-up LCI and VDP. Redline marks the upper limit of normality. The arrow points from baseline to follow-up. Overall good agreement in the delta-VDP and delta-LCI is visible. Two outliers are marked as red dots and are shown in figure 4.

Note. – VDP: ventilation defect percentage; LCI: Lung clearance index

Figure 3: Coronal 3D UTE images in a patient with lung nodules and a healthy volunteer (A) before (B) after retrospective binning and (C) following concomitant field corrections. Blue arrows show some of the regions of improvements with self-gating-based retrospective binning. Red circle shows improvement in delineation of the lung nodule. Red arrows highlight some visible improvements in image sharpness with concomitant field corrections. Significant improvements in image sharpness are demonstrated with both binning and concomitant field corrections.

Figure 4: Representative example of axial 3D UTE images in a LAM patient (A) before and (B) after concomitant field correction. Blue arrows show improvements in cyst delineation and visibility with concomitant field corrections. Red arrows show improvements in vessel sharpness.

Coronal plane slice of a healthy volunteer with the Time to Peak (TTP) and Full Width Half Max (FWHM) biomarker on the left and right, respectively. The scans were repeated twice for reproducibility, with the first and second rows of the image corresponding to repeated scans of the volunteer. For the TTP biomarker, a value of five is to be expected as the fifth phase corresponds to exhalation, representing the highest signal intensity value. For the FWHM biomarker, a value of six is expected, as it would capture the phases before and after full exhalation.

Sagittal plane slice of a healthy volunteer with the Time to Peak (TTP) and Full Width Half Max (FWHM) biomarker on the left and right, respectively. The scans were repeated twice for reproducibility, with the first and second rows of the image corresponding to repeated scans of the volunteer. The results of the sagittal place slice support the findings of the coronal plane slice visualization.

Figure 1. 62-year-old female with invasive adenocarcinoma (L to R: thin-section CT, DWI, ADC map, MTR_asym map fused with T2WI, and SUV_max map fused with CT) and determined as recurrence group.

Thin-section CT demonstrates a nodule in the right upper lobe. ADC of this nodule was 1.05×10-3mm2/s. SUV_max was 2.6. APTw image shows low MTR_asym with the value of 1.25. This case was false-negative on PET/CT, and true-positive on DWI and APTw image. When combined both indexes, PET/CT with APTw image was diagnosed as recurrence group and determined as true-positive case.

Figure 2. Results of multivariate regression analysis for distinguishing recurrence from non-recurrence groups in NSCLC patients.

MTR_asym at 3.5ppm and SUV_max were determined as significant predictor for distinguishing recurrence from non-recurrence groups (p<0.05).

Figure 1. 72-year old invasive adenocarcinoma with N2 disease

cDWI at each b value demonstrates right hilar, right lower paratracheal, subcarina lymph node metastases (arrow, large arrow and arrow head). However, PET/CT shows only right hilar lymph node metastasis (arrow). On the other hand, aDWI and ADC demonstrate right hilar and right lower paratracheal lymph node metastases (arrow and large arrow). This case was assessed as N2 case on each cDWI and aDWI and considered as true-positive. However, this case was evaluated as N1 case on PET/CT and considered as false-negative case.

Figure 2. Result of ROC analysis for distinguishing metastatic from non-metastatic lymph nodes.

Area under the curve (AUC) of CR₆₀₀ (AUC=0.87) was significantly larger than that of SUV_max (AUC=0.77, p=0.003), CR₄₀₀ (AUC=0.79, p<0.0001) and CR₈₀₀ (AUC=0.81, p<0.0001). In addition, AUC of CR₈₀₀ was also significantly larger than that of CR₄₀₀ (p=0.02).

Figure.1: A 58-year-old male patient with lung adenocarcinoma. The white arrow points to the lesion.ADC (a), α (b), DDC (c) pseudo-color image and pet image (d).

Figure.2：Violin case diagram.α value of AC（0.64±0.19） and SCC（0.81±0.10）group（A）, DDC value of AC [(2.14±0.83)×10^-3 mm² /s ]and SCC[(1.33±0.30)×10^-3 mm² /s] group（B）. SUVmax value of AC (7.36±4.61) and SCC(12.56±5.85) group（C）, AC: adenocarcinoma, SCC: Squamous cell carcinoma.

Figure 1.A 61-year-old man with left lung squamous cell carcinoma. The red arrows point to the lesion. A. PET image. B. D_t pseudo colored map. C. D_p pseudo colored map. D. F pseudo colored maps. E. sADC pseudo colored map. F. Hematoxylin and eosin(HE)x 100.

Figure 2.A 56-year-old man with left lung tuberculous granuloma. The red arrows point to the lesion.A. PET image. B. D_t pseudo colored map. C. D_p pseudo colored map. D. F pseudo colored maps. E. sADC pseudo colored map. F. Hematoxylin and eosin(HE)x 100.

Figure 3. Representative IVIM-DWI analysis for lepidic predominant adenocarcinoma and micropapillary predominant adenocarcinoma. (A) f map from a lepidic predominant adenocarcinoma patient; (B) f map from a micropapillary predominant adenocarcinoma patient.

Figure 4. Representative OE-UTE-MRI analysis for lepidic predominant adenocarcinoma and micropapillary predominant adenocarcinoma. (A) lesion PSE map from a lepidic predominant adenocarcinoma patient; (B) lesion PSE map from a micropapillary predominant adenocarcinoma patient.

Fig. 1:A 47-years-old female with pathologically confirmed Small cell lung cancer in the right lung. A: T2WI, B: T1WI+C, C: Native T1 mapping, D:T1 mapping+C.

Fig. 2: A 67-years-old female with pathologically confirmed adenocarcinoma in the right lung. A: T2WI, B: T1WI+C, C: Native T1 mapping, D: T1 mapping+C.

Figure1.The GRASP DCE-derived quantitative values distribution of malignancy and benignity in lung lesions. The mean values of Ve in Adeno-Ca and benign groups are 0.360±0.167 and 0.428±0.114. (ANOVA, p<0.05)

Figure 3. A 72-year-old male with a 6cm mass in the right upper lobe. This mass was diagnosed as adenocarcinoma IIIb (cT4N2M0). (A): Free-breathing Golden-angle RAdial Sparse Parallel (GRASP)-DCE MRI post-processed image showing mean Ktrans of 0.264 min-1, Kep of 1.769 min-1 and Ve of 0.149. The iAUC value of Time-intensity curve for this mass was 0.125. (B):There were 3 early phases images from top to the bottom after contrast injection showing the mass continuous heterogeneous enhancement.

Figure.1 :A 62-year-old man with left Lung adenocarcinoma. A is T2w map,B is D pseudo colored map，C is f pseudo colored map, D is D* pseudo colored maps，E is Calculated B0 map,F is MTRasym(3.5ppm) pseudo colored maps.

Figure.2: The MTRasym (3.5ppm), D, D*, and f values of adenocarcinoma and squamous cell carcinoma groups. (A) MTRasym (3.5ppm) = (22.04±12.87)%, (8.25±6.77)% ; (B) D = (1.33±0.44) ×10−³ mm²/s , (0.89± 0.16)×10−³ mm²/s ; (C) D* = (67.34±68.22)×10−³ mm²/s , (25.66±24.43)×10−³ mm²/s; (D) f = (0.34±0.18), (0.26±0.09) .

A-D A adenocarcinoma in the left upper lobe (A) An axial ADC mapping. (B) D map. (C) T2 map. (D) Hematoxylin-eosin staining confirms the nodule which is associated with nuclear polymorphism, higher cellularity and nuclear-to-cytoplasmic ratios which make ADC value, D value, and T2 value down.E-H A ball pneumonia in the right middle lobe. (E) An axial ADC mapping. (F) D map. (G) T2 map. (H) Hematoxylin-eosin staining confirms the nodule was characterized by more extracellular fluid spec and tissue fluid which would make ADC value, D value and T2 value rise.

Figure 1. The IVIM quantitative values distribution of malignancy and benignity in lung lesions. The mean ADC value of benign lesions is 1.846±0.403×10-3 mm2/s (ANOVA, p<0.01)

Figure 3. 72-year-old male with a 6cm mass in the right upper lobe. This mass was diagnosed as adenocarcinoma IIIb (cT4N2M0).ROI area of the mass is 2.341 cm2.(A): IVIM-f shows a value of 0.033×10-3mm2/s around the mass. (B): IVIM-D* shows a value of 0.030×10-3mm2/s around the mass. (C):IVIM-D shows a value of 1.041×10-3mm2/s around the mass. (D): ADC shows a value of 1.066×10-3mm2/s around the mass.

Figure 1: Example images from CT, UTE MRI, and ¹²⁹Xe DW-MRI for one IPF patient. (a) Baseline images where CALIPER ILD%, UTE signal, ¹²⁹Xe ADC, and Lm_D are elevated in the lower zone compared to the global mean value. (b) Images in the same patient after 1 year where increases in the global mean value are observed for all imaging metrics.

Figure 2: (a) Scatter plot demonstrating a significant correlation between lower zone normalized UTE signal and lower zone CALIPER ILD% values across both baseline and 1 year scans. Scatter plots comparing lower zone normalized UTE signal with lower zone ¹²⁹Xe ADC (b) and Lm_D (c) values.

Statistical analysis of UIP and NSIP

58-Figure 1 A 58-year-old female with UIP mainly had line-shape enhancement or patchy enhancement. The Ktrans and Kep values were low in the drawn ROI on the slice of the largest lesion area. The pseudo-color map only including lesions, parts of healthy lungs and muscles mainly displayed in blue.

Figure 1. Examples of CALIPER CT (a), ¹²⁹Xe ADC (b) and ¹²⁹Xe Lm_D (c) images.

Figure 2. Correlation between global ¹²⁹Xe RBC:TP and CALIPER reticulation % (a), fibrosis % (b) and vessel related structures % (c) (n=13).

Figure 3: Two example T₁ maps of patients with IPAH with no lung disease seen on CT, and with emphysema seen on CT (left) and two box plots of mean and median lung T₁ in patients with PH (right). There are significant differences in median parenchyma T₁ between patients with emphysema and patients with fibrosis and centrilobular ground glass that are not seen when using mean T₁ as an average metric.

Figure 2: Example T₁ maps in a patient with PH due to lung disease and pulmonary arterial hypertension. Box plots of mean and median lung T₁ in patients with PH and control groups both show significantly lower lung parenchyma T₁ in patients with PH due to lung disease, than patients without PH, patients with PAH or healthy volunteers.

Figure 1. (a) Ultrashort echo-time dynamic contrast enhancement MRI image and signal enhancement evolution in the lung before (baseline) and after gadolinium injection in (A) SS and SSBN3 rats (control- 0 Gy),(B) SS rats radiated at 0 (n = 4) and 13 Gy (n = 5), and (C) SSBN3 rats radiated at 0 (n = 4) and 13 Gy (n = 4). Error bars represent mean ± SEM. (D) Survival curve Leg-out PBI with supportive care in SS and SS-BN3 rats.

Figure 2. (A) A typical image of ultrashort echo-time dynamic contrast enhancement MRI image in the lung (five ROIs in the lung used to understand the perfusion in the lung) and (B) kinetic DCE raw data (without motion correction – black circles) and filtered data for respiratory motion correction (red line).

Figure 4. Comparison of right and left lung xt-PCA tidal breathing signal in a healthy volunteer (top row) and in a patient with right diaphragmatic palsy (DP) (bottom row). The dysfunction of the right lung in the patient manifests as a disproportionate diaphragmatic excursion with a compensatory larger signal in the left lung.

Figure 3. Results of xt-PCA diaphragm motion extraction in a healthy volunteer (top row) and a postsurgical patient following robotic plication and chest wall reconstruction (bottom row). A comparison of tidal breathing and maximum inspiration and expiration shows a number of potentially useful features for diagnosis, including amplitude, frequency and regularity. Both figures are animated to the same time scale and included on each figure is the mean signal (black) and ±1σ standard deviation (grey).

Figure 1. Workflow of spatial comparison method of ¹H and ³He MRI ventilation.

Figure 2. Corresponding coronal slices for three patients of ³He (top row)) and ¹H (bottom row) MRI ventilation after registration. The red arrows depict defects and regions that are visually similar on both scans. Voxel-wise Spearman's correlation coefficients (ρ) are provided for each case.

Fig. 1. Second measurements vs. first measurements of tissue-to-gas, RBC-to-gas, and RBC-to-tissue, in healthy subjects (blue) and in COPD subjects (red), as well as intraclass correlation coefficient (ICC), coefficient of variation (CV), and coefficient of repeatability (CR) for each of the three gas uptake measures.

Fig. 3. Percent change in gas uptake ratios vs. percent change in lung volume, for healthy subjects (blue) and COPD subjects (red), with linear fits to all subjects (black line), healthy subjects (blue line), and COPD subjects (red line), as well as correlation coefficients ρ, slopes, and y-intercepts for each of the fits.

Figure 1: Multi-breath wash-in and wash-out of inhaled PFP/O₂. A central coronal slice from a 3D ¹⁹F-MR image acquired during each breath hold is displayed. Images were acquired on alternate inhalations (ie. after 1, 3, 5, etc. inhalations of PFP/O₂). PFP/O₂ wash-in is visible over the initial breaths. The inhalation rig valve was then switched to deliver room air for visualisation of PFP/O₂ gas wash-out. 1a. PFP/O₂ wash-in and wash-out in a healthy volunteer. 1b. PFP/O₂ wash-in in P1 (nb. wash-out images were not acquired in this participant). 1c. PFP/O₂ wash-in and wash-out in P2.

Figure 3: 3D voxel-wise exponential time constant estimation, where τ₁ represents the number of wash-in respiratory cycles required to increase the ¹⁹F-MRI signal to 1 - 1/e (≈63.2%) of its calculated maximum value. τ₂ represents the number of respiratory cycles required to reduce the ¹⁹F-MRI signal to 1/e (≈36.8%) of its calculated maximum value. Maps for wash-in τ₁(3a., 3c., and 3d.) and wash-out τ₂ (3b. and 3e.) are shown.

Figure 1.(A) CT images at expiration state, (B) CT images at inspiration state, (C) MR images at expiration state, (D) MR images at inspiration state. Yellow arrows indicate emphysema lesion. In VS-UTE, although there is a slight decrease in signal, it is not easy to accurately diagnose the lesion.

Figure 2. (A,B) CT image at inspiration state, (C,D) Ventilation map each slide using VS-UTE, (E,F) Ventilation flow map each slide using VS-UTE. Yellow arrows indicate emphysema lesion. Figure 2 show the left lower lobe had high correlation with CT and MRI ventilation map.

Figure 3. Example coronal slices of DL and SFCM segmentations for three cases with different image resolutions and diseases compared to the expert segmentations. DSC and Avg HD values are given for each case.

Figure 4. a) Comparison of segmentation performance of DL and SFCM for all scans in the testing set and for each acquisition protocol. Means are given; the best result for each metric is in bold. b) Comparison of DL performance for each of the three acquisition protocols using DSC (left) and Average boundary Hausdorff distance (right). Significances of differences between acquisitions were assessed using a Mann–Whitney U test.

Figure 2. Example images from our validation set. A is the ground truth (the original breath-hold ZTE image), B is the test image (artefact augmented breath-hold ZTE) and C is the prediction of the network.

Figure 3. Example images from our test set for one volunteer. On the top there is the same slice in 4 different respiratory phases from the original 4D-ZTE acquisition and on the bottom there are the same phases predicted by the network.

Figure 1: An example of the segmentation performance. Three consecutive slices of a 3D image are shown in column (a) together with the manual ground truth in (b) and the DL output in (c), with parenchyma in green and airways in yellow. The example shows the quality of the predictions and the challenges the system has to face, where a small portion of one airway in the mid slice was excluded from both labels.

Figure 2: Examples of HP gas images where some ambiguous features, circled in orange, are present. These features resemble airway segments and, potentially, could be misclassified. In the upper row the HP gas images are shown, in the lower row the DL output is displayed, which shows that the features were correctly identified as belonging to the lung ventilated region.

Figure 1: Representation of the pipeline of processes which formed the framework which was applied to model dynamic OE-MRI scans.

Figure 3: Comparison of predicted (calculated using a value for the maximum theoretical ΔPO₂) T₁(t) (a) and ΔPO₂(t) (b) to values extracted from the dynamic series. The extracted T₁ and ΔPO₂ values change by less than predicted. Plateau T_1,oxy = (1147 ± 3) ms; ΔPO₂ = (201 ± 9) mmHg. Vertical lines indicate a switch from the simulated subject breathing air to 100% O₂ (4.2 minutes) or 100% O₂ to air (15.4 minutes).

Figure 1: Representative ¹⁹F gas images acquired with (a) 2D-GRE and (b) 3D-SoS. The yellow squares approximate the 9×9 ROI and locations in the phantom and background used to measure mean signal and background noise for SNR measurements.

Figure 2: (a) Simulated durations for three clinically relevant in-plane resolutions for 2D-GRE and 3D-SoS at a constant TR of 15 ms (10 slices). All other parameters are as stated in Table 1 for each sequence. A simulation with the minimum TR for 3D-SoS is also shown. (b) Achievable number of signal averages (NSA) in an assumed 20 second breath-hold, rounded to the nearest integer using scan durations in (a). (c) Relative SNR accounting for changes in spatial resolution, NSA in 20 sec, and difference in TE. All SNRs normalized to the 64×64 2D-GRE.

Figure 1: Bronchial Tree Simulation (R_2b) We use a recent lung XCAT phantom to generate a susceptibility map with the entire bronchial tree. This is used to compute a high resolution RDF map, that is smoothed using Fourier truncation. Several timepoints are then simulate, and exponential fitting is used to estimate R_2b maps. RDF maps and R_2b maps are shown with an overlaid lung outline. Dashed lines indicate the location of orthogonal sections.

Figure 2: Alveolus Simulation (R_2b). We simulate face-center cubic packing, using (a) a 9x9x9 lattice of fundamental blocks (3x3x3 is illustrated). Each fundamental block represents 0.5x0.5x0.5 mm³ with 5 isotropic resolution. We then simulate the RDF, and extract the (b) central fundamental block from the RDF map. We exclude air-filled spheres, and use only parenchyma voxels to simulate dephasing and perform estimation. A = alveolus.

Figure 3: Combined functional maps overlaid on a cross-section shown for 1.5 T (a) and 3 T (b). At 1.5T, both phase-cycled maps provide similar contrast and prominent structures (blue arrows) compared to constant phase maps while phase-cycled perfusion map suffers from overall lower values due to averaging. At 3T, both phase-cycled maps show similar contrast and prominent structures (blue arrows) compared to 1.5T. However, the constant phase is less able to reproduce these structures since it is more prone to field inhomogeneity artefacts (white arrows).

Figure 2: Ventilation and perfusion maps of subgroups at 3 T for constant phase (a) and phase-cycled (b) bSSFP acquisitions. As expected, constant phase acquisition generates similar functional maps throughout the experiment whereas phase-cycled acquisition is able to generate functional maps with different information by changing the RF phase. At 3 T, the perfusion maps with conventional $$$\Delta\phi = \pi$$$ acquisition suffer from field inhomogeneity and are less comprehensive compared to phase-cycled acquisitions.

Figure 3. Illustrative coronal FB bSTAR image reconstructions in a healthy volunteer: using all acquired data (no gating) - composite (a), data binned to end inspiration - volume 1 (b), data binned to intermediate respiratory state - volume 5 (c) and data binned to end expiration - volume 10 (d). For comparison breath-hold bSTAR acquisition was performed (e).

Figure 4. Comparison between free-breathing bSTAR composite (a), free-breathing bSTAR expiratory reconstruction (b), and expiratory breath-hold bSTAR image. Diagram (d) shows signal amplitude profiles obtained in the region indicated by the yellow box (signal amplitude was averaged along the x-axis).

Figure 1: Workflow of improved iMoCo reconstruction

Figure 3: Comparison of reconstruction result using original iMoCo, iMoCo1, iMoCo2 and iMoCo3. iMoCo2 and iMoCo3 enabled better suppression of streaking artifacts (red arrows).

Figure 3: Free-breathing 3D UTE images of 4 consecutive slices acquired without fat suppression (a) and with fat suppression (b). Fat around the heart is suppressed in (b) with better visualization of pulmonary arteries and myocardium.

Figure 4: Maximum Intensity Projection of a segmented lung volume from fat-suppressed 3D UTE.

Figure 1: obtained population AIF, with AUC normalised to 1. Gray shaded area indicates the ± standard deviation area.

Figure 4: examples of blood volume (first row), MTT (second row) and blood flow (third row) calculated using an individually-measured AIF (left) and the population AIF (right) in a 52 years old male patient affected by CTD-ILD.

Figure 2: Proton fraction maps for EX and IN in coronal and axial slice orientation. The grey lines indicate the position of the slice in the respective other orientation.

Table 1: Proton fraction (PF) and SNR values (mean ± std) for the coronal (a) and axial (b) slice orientation.

a) absolute perfusion values and b) perfusion phase as determined by SENCEFUL in MRI acquisitions using a Cartesian or radial read-out scheme at different echo times. c) Distribution of the perfusion phase values in the lung.

a) ventilation maps as determined by SENCEFUL in MRI acquisitions using a Cartesian or radial read-out scheme at different echo times. b) Distribution of the absolute ventilation values in the lung.

Figure 5. The example of MIP lung images reconstructed from the end expiration phase in axial (A), sagittal (B) and coronal (C) planes. The small pulmonary vessels are well-displayed in these images.

Figure 3. The estimation of respiration using the phase of DC. The data were acquired from a 37-year-old healthy male volunteer. Panel A shows the raw phase of DC (i.e., the blue curve) received from a coil close to the liver. The red curve in panel A is the corresponding low-pass filtered phase representing the respiration. The cut-off frequency is ranged from 0.1 to 0.5 Hz. The respiration is segmented into five phases (separated by the green lines) as shown in panel B. Phase 1 and 5 are corresponding to the end of inspiration and expiration respectively.

Figure 2: Signal enhancement in FB T1 weighted 3D stack of spiral UTE images. A) T1w UTE images at normoxia, and hyperoxia for two repetitions B) Histogram of signal intensity for normoxia and hyperoxia images within the lung for two repetitions shows excellent repeatability. C) PSE maps for the two repitions estimated from images in A) using Eq 1.

Figure 4: Representative temporal SNR map from subject 2. Images reconstructed from (top) BH data and (bottom) FB data. (left) Images before registration and (right) images after registration. tSNR is significantly higher for FB (un-registered: 23.8 ± 13.7, registered: 27.1+/-13.6) data compared to BH (un-registered: 9.6 ± 1.5, registered: 12.6 ± 2.6). Registration improves tSNR, as expected.

Figure 2. Representative Regional Ventilation Map of Two Scans and Their Difference Using Three Registration Methods. Within each method, the first row and second row are from the first and second scan respectively while the third row is the difference of ventilation maps between the two after a simple registration. Regional ventilation of 0.1 corresponds to 10% of volume expansion with respect to the end-expiration state. For conciseness, every other respiratory phase is shown.

Figure 3. Split Violin Plot of the Regional Ventilation Distribution of All Subjects and Registration Methods. The discreet horizontal axis is the respiratory phases starting from the end-expiration phase, while the vertical axis is the regional ventilation. The two colors represent the 1st and 2nd scan respectively, and the shaded areas depict the regional ventilation distribution. The dotted lines are the quartiles of the distributions.

Figure 1: (a) Location of the different ROIs in the heart (red), lung (green), muscle (yellow) and background (orange) superimposed onto a pre CA image. In (b) the percentage change of signal intensity is shown.

Figure 2: Signal intensity changes during contrast agent injection. The time-frames in the graphs were reconstructed with different sliding window width ((a) 1000 projections, (b) 500 projections, (c) 250 projections and (d) 100 projections, t_W = 2.4s, 6s, 12s, 24s). The sliding window step was chosen as 25 projections (𝛥t = 600ms).

Figure 3 A-B. The azygos fissure on ZTE images. Both (A) axial and (B) coronal reconstructed image demonstrated the azygos fissure and vein (arrow).

Figure 2 A-B. Visibility of lung structures. (A) The posterior segmental bronchus (4th generation) of the right lower lobe (arrow) was well depicted by ZTE. (B) The 7th generation of the pulmonary artery in the right lower lobe (arrow) was well displayed using maximum intensity projection (MIP) by ZTE.

Figure 1: Example for BH (a) and SG (b) images in end-expiration and end-inspiration. It can be seen, that the breathing amplitude is bigger in the BH acquisition than in the SG. It is also visible, the BH images are sharper than the SG images. For both techniques a FV-map could be calculated. The proton fraction maps also shown. It is clearly difference between EX and IN visible.

Figure 2: For smoker and non-smoker the proton fraction values (the shown values are for EX) show a trend to increase from anterior to posterior. The values for the smokers are significant higher than for the non-smokers. The proton density in the BH images were a bit lower than in the SG images.

Figure 3. First row: T₁ images acquired with air and after a corresponding number of O₂ wash-in breaths (N). Lung mean T₁ values and MR-measured gas concentrations are given. Second row shows the following images respectively: morphological image, T₁ during room air and O₂ (full wash in) and oxygen enhanced and O₂ wash-in time-constant maps. Third row shows the images of ventilation and perfusion. This healthy volunteer has slightly increased ventilation in the left upper lung as compared to right. Similarly, the OE signal is higher in the upper left lung, and the wash-in shorter.

Figure 4. On the top: graph showing the pulmonary mean T₁ and gas concentrations throughout the first seven N₂ washout / O₂ wash-in breaths. O₂ delivery starts at breath number 0. To note the close relationship with MBW measured gas concentration of the exemplary MBW curve (cf. Figure 1); e.g. the O₂ concentration at the third breath as measured with the MRI is 80%, whereas with MBW it is ~77%. In the middle: maps of residual regional N₂ gas concentration after the corresponding number of oxygen wash-in breaths (N). On the bottom: morphological image to locate lung vessels and structures.

Figure 2: Exemplary image quality and the corresponding intensity profiles over the lung-liver interface for all three reconstruction techniques and the same acquisition. In expiration (top row) there is not substantial motion blur of the lung-liver interface, although the k-space gated image appears somewhat washed out. For inspiration, the nuSG reconstruction shows highest motion fidelity, followed by the image-gated reconstruction. The lung-liver interface is clearly blurred in the ksp-gated reconstruction.

Figure 3: Exemplary nuSG reconstructed image (right) and time course for uniform and non-uniform motion along the red intensity profile. The displacement of the lung-liver interface exhibits clearly non-uniform motion for the lower row (e.g. around frame 200 and between frame 500 and 600).