Deep Learning-Based Analysis of Aortic Morphology From Three-Dimensional MRI.

BACKGROUND: Quantification of aortic morphology plays an important role in the evaluation and follow-up assessment of patients with aortic diseases, but often requires labor-intensive and operator-dependent measurements. Automatic solutions would help enhance their quality and reproducibility. PURPOSE: To design a deep learning (DL)-based automated approach for aortic landmarks and lumen detection derived from three-dimensional (3D) MRI. STUDY TYPE: Retrospective. POPULATION: Three hundred ninety-one individuals (female: 47%, age = 51.9 ± 18.4) from three sites, including healthy subjects and patients (hypertension, aortic dilation, Turner syndrome), randomly divided into training/validation/test datasets (N = 236/77/78). Twenty-five subjects were randomly selected and analyzed by three operators with different levels of expertise. FIELD STRENGTH/SEQUENCE: 1.5-T and 3-T, 3D spoiled gradient-recalled or steady-state free precession sequences. ASSESSMENT: Reinforcement learning and a two-stage network trained using reference landmarks and segmentation from an existing semi-automatic software were used for aortic landmark detection and segmentation from sinotubular junction to coeliac trunk. Aortic segments were defined using the detected landmarks while the aortic centerline was extracted from the segmentation and morphological indices (length, aortic diameter, and volume) were computed for both the reference and the proposed segmentations. STATISTICAL TESTS: Segmentation: Dice similarity coefficient (DSC), Hausdorff distance (HD), average symmetrical surface distance (ASSD); landmark detection: Euclidian distance (ED); model robustness: Spearman correlation, Bland-Altman analysis, Kruskal-Wallis test for comparisons between reference and DL-derived aortic indices; inter-observer study: Williams index (WI). A WI 95% confidence interval (CI) lower bound >1 indicates that the method is within the inter-observer variability. A P-value <0.05 was considered statistically significant. RESULTS: DSC was 0.90 ± 0.05, HD was 12.11 ± 7.79 mm, and ASSD was 1.07 ± 0.63 mm. ED was 5.0 ± 6.1 mm. A good agreement was found between all DL-derived and reference aortic indices (r >0.95, mean bias <7%). Our segmentation and landmark detection performances were within the inter-observer variability except the sinotubular junction landmark (CI = 0.96;1.04). DATA CONCLUSION: A DL-based aortic segmentation and anatomical landmark detection approach was developed and applied to 3D MRI data for achieve aortic morphology evaluation. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 2.

Four-dimensional flow cardiovascular magnetic resonance aortic cross-sectional pressure changes and their associations with flow patterns in health and ascending thoracic aortic aneurysm.

BACKGROUND: Ascending thoracic aortic aneurysm (ATAA) is a silent and threatening dilation of the ascending aorta (AscAo). Maximal aortic diameter which is currently used for ATAA patients management and surgery planning has been shown to inadequately characterize risk of dissection in a large proportion of patients. Our aim was to propose a comprehensive quantitative evaluation of aortic morphology and pressure-flow-wall associations from four-dimensional (4D) flow cardiovascular magnetic resonance (CMR) data in healthy aging and in patients with ATAA. METHODS: We studied 17 ATAA patients (64.7 ± 14.3 years, 5 females) along with 17 age- and sex-matched healthy controls (59.7 ± 13.3 years, 5 females) and 13 younger healthy subjects (33.5 ± 11.1 years, 4 females). All subjects underwent a CMR exam, including 4D flow and three-dimensional anatomical images of the aorta. This latter dataset was used for aortic morphology measurements, including AscAo maximal diameter (iDMAX) and volume, indexed to body surface area. 4D flow MRI data were used to estimate 1) cross-sectional local AscAo spatial (∆PS) and temporal (∆PT) pressure changes as well as the distance (∆DPS) and time duration (∆TPT) between local pressure peaks, 2) AscAo maximal wall shear stress (WSSMAX) at peak systole, and 3) AscAo flow vorticity amplitude (VMAX), duration (VFWHM), and eccentricity (VECC). RESULTS: Consistency of flow and pressure indices was demonstrated through their significant associations with AscAo iDMAX (WSSMAX:r = -0.49, p < 0.001; VECC:r = -0.29, p = 0.045; VFWHM:r = 0.48, p < 0.001; ∆DPS:r = 0.37, p = 0.010; ∆TPT:r = -0.52, p < 0.001) and indexed volume (WSSMAX:r = -0.63, VECC:r = -0.51, VFWHM:r = 0.53, ∆DPS:r = 0.54, ∆TPT:r = -0.63, p < 0.001 for all). Intra-AscAo cross-sectional pressure difference, ∆PS, was significantly and positively associated with both VMAX (r = 0.55, p = 0.002) and WSSMAX (r = 0.59, p < 0.001) in the 30 healthy subjects (48.3 ± 18.0 years). Associations remained significant after adjustment for iDMAX, age, and systolic blood pressure. Superimposition of ATAA patients to normal aging trends between ∆PS and WSSMAX as well as VMAX allowed identifying patients with substantially high pressure differences concomitant with AscAo dilation. CONCLUSION: Local variations in pressures within ascending aortic cross-sections derived from 4D flow MRI were associated with flow changes, as quantified by vorticity, and with stress exerted by blood on the aortic wall, as quantified by wall shear stress. Such flow-wall and pressure interactions might help for the identification of at-risk patients.

Deep Learning-based Automated Aortic Area and Distensibility Assessment: the Multi-Ethnic Study of Atherosclerosis (MESA).

This study details application of deep learning for automatic segmentation of the ascending and descending aorta from 2D phase-contrast cine magnetic resonance imaging for automatic aortic analysis on the large MESA cohort with assessment on an external cohort of thoracic aortic aneurysm (TAA) patients. This study includes images and corresponding analysis of the ascending and descending aorta at the pulmonary artery bifurcation from the MESA study. Train, validation, and internal test sets consisted of 1123 studies (24,282 images), 374 studies (8067 images), and 375 studies (8069 images), respectively. The external test set of TAAs consisted of 37 studies (3224 images). CNN performance was evaluated utilizing a dice coefficient and concordance correlation coefficients (CCC) of geometric parameters. Dice coefficients were as high as 97.55% (CI: 97.47-97.62%) and 93.56% (CI: 84.63-96.68%) on the internal and external test of TAAs, respectively. CCC for maximum and minimum and ascending aortic area were 0.969 and 0.950, respectively, on the internal test set and 0.997 and 0.995, respectively, for the external test. The absolute differences between manual and deep learning segmentations for ascending and descending aortic distensibility were 0.0194 × 10-4 ± 9.67 × 10-4 and 0.002 ± 0.001 mmHg-1, respectively, on the internal test set and 0.44 × 10-4 ± 20.4 × 10-4 and 0.002 ± 0.001 mmHg-1, respectively, on the external test set. We successfully developed a U-Net-based aortic segmentation and analysis algorithm in both MESA and in external cases of TAA.

Analysis of aortic pressure fields from 4D flow MRI in healthy volunteers: Associations with age and left ventricular remodeling.

BACKGROUND: Aging-related arterial stiffness is associated with substantial changes in global and local arterial pressures. The subsequent early return of reflected pressure waves leads to an elevated left ventricular (LV) afterload and ultimately to a deleterious concentric LV remodeling. PURPOSE: To compute aortic time-resolved pressure fields of healthy subjects from 4D flow MRI and to define relevant pressure-based markers while investigating their relationship with age, LV remodeling, as well as tonometric augmentation index (AIx) and pulse wave velocity (PWV). STUDY TYPE: Retrospective. POPULATION: Forty-seven healthy subjects (age: 49.5 ± 18 years, 24 women). FIELD STRENGTH/SEQUENCE: 3 T/4D flow MRI. ASSESSMENT: Spatiotemporal pressure fields were computed by integrating velocity-derived pressure gradients using Navier-Stokes equations, while assuming zero pressure at the sino-tubular junction. To quantify aortic pressure spatiotemporal variations, we defined the following markers: 1) volumetric aortic pressure propagation rates ΔP E1 /ΔV and ΔP E2 /ΔV, representing variations of early and late systolic relative pressure peaks along the aorta, respectively, according to the cumulated aortic volume; 2) ΔA PE1-PE2 defined in four aortic regions as the absolute difference between early and late systolic relative pressure peaks amplitude. STATISTICAL TESTS: Linear regression, Wilcoxon rank sum test, Bland-Altman analysis, and intraclass correlation coefficients (ICC). RESULTS: Spatiotemporal variations of aortic pressure peaks were moderately to highly reproducible (ICC ≥0.50) and decreased significantly with age, in terms of absolute magnitude: ΔP E1 /ΔV (r = 0.70, P < 0.005), ΔP E2 /ΔV (r = -0.45, P < 0.005) and ΔA PE1-PE2 (|r| > 0.39, P < 0.005). ΔP E1 /ΔV was associated with LV remodeling (r = 0.53, P < 0.001) and ascending aorta ΔA PE1-PE2 was associated with AIx (r = -0.59, P < 0.001). Both associations were independent of age and systolic blood pressures. Only weak associations were found between pressure indices and PWV (r ≤ 0.40). DATA CONCLUSION: 4D flow MRI relative aortic pressures were consistent with physiological knowledge as demonstrated by their significant volumetric and temporal variations with age and their independent association with LV remodeling and augmentation index. Level of Evidence 2 Technical Efficacy Stage 3 J. Magn. Reson. Imaging 2019;50:982-993.

Comparison of different methods for the estimation of aortic pulse wave velocity from 4D flow cardiovascular magnetic resonance.

BACKGROUND: Arterial pulse wave velocity (PWV) is associated with increased mortality in aging and disease. Several studies have shown the accuracy of applanation tonometry carotid-femoral PWV (Cf-PWV) and the relevance of evaluating central aorta stiffness using 2D cardiovascular magnetic resonance (CMR) to estimate PWV, and aortic distensibility-derived PWV through the theoretical Bramwell-Hill model (BH-PWV). Our aim was to compare various methods of aortic PWV (aoPWV) estimation from 4D flow CMR, in terms of associations with age, Cf-PWV, BH-PWV and left ventricular (LV) mass-to-volume ratio while evaluating inter-observer reproducibility and robustness to temporal resolution. METHODS: We studied 47 healthy subjects (49.5 ± 18 years) who underwent Cf-PWV and CMR including aortic 4D flow CMR as well as 2D cine SSFP for BH-PWV and LV mass-to-volume ratio estimation. The aorta was semi-automatically segmented from 4D flow data, and mean velocity waveforms were estimated in 25 planes perpendicular to the aortic centerline. 4D flow CMR aoPWV was calculated: using velocity curves at two locations, namely ascending aorta (AAo) and distal descending aorta (DAo) aorta (S1, 2D-like strategy), or using all velocity curves along the entire aortic centreline (3D-like strategies) with iterative transit time (TT) estimates (S2) or a plane fitting of velocity curves systolic upslope (S3). For S1 and S2, TT was calculated using three approaches: cross-correlation (TTc), wavelets (TTw) and Fourier transforms (TTf). Intra-class correlation coefficients (ICC) and Bland-Altman biases (BA) were used to evaluate inter-observer reproducibility and effect of lower temporal resolution. RESULTS: 4D flow CMR aoPWV estimates were significantly (p < 0.05) correlated to the CMR-independent Cf-PWV, BH-PWV, age and LV mass-to-volume ratio, with the strongest correlations for the 3D-like strategy using wavelets TT (S2-TTw) (R = 0.62, 0.65, 0.77 and 0.52, respectively, all p < 0.001). S2-TTw was also highly reproducible (ICC = 0.99, BA = 0.09 m/s) and robust to lower temporal resolution (ICC = 0.97, BA = 0.15 m/s). CONCLUSIONS: Reproducible 4D flow CMR aoPWV estimates can be obtained using full 3D aortic coverage. Such 4D flow CMR stiffness measures were significantly associated with Cf-PWV, BH-PWV, age and LV mass-to-volume ratio, with a slight superiority of the 3D strategy using wavelets transit time (S2-TTw).

Value of aortic volumes assessed by automated segmentation of 3D MRI data in patients with thoracic aortic dilatation: A case-control study.

PURPOSE: The purpose of this study was to investigate the benefit of aortic volumes compared to diameters or cross-sectional areas on three-dimensional (3D) magnetic resonance imaging (MRI) in discriminating between patients with dilated aorta and matched controls. MATERIALS AND METHODS: Sixty-two patients (47 men and 15 women; median age, 66 years; age range: 33-86 years) with tricuspid aortic valve and ascending thoracic aorta aneurysm (TAV-ATAA) and 43 patients (35 men and 8 women; median age, 51 years; age range: 17-76 years) with bicuspid aortic valve and dilated ascending aorta (BAV) were studied. One group of 54 controls matched for age and sex to patients with TAV-ATAA (39 men and 15 women; median age, 68 years; age range: 33-81 years) and one group of 42 controls matched for age and sex to patients with BAV (34 men and 8 women; median age, 50 years; age range: 17-77 years) were identified. All participants underwent 3D MRI, used for 3D-segmentation for measuring aortic length, maximal diameter, maximal cross-sectional area (CSA) and volume for the ascending aorta. RESULTS: An increase in ascending aorta volume (TAV-ATAA: +107%; BAV: +171% vs. controls; P < 0.001) was found, which was three times greater than the increase in diameter (TAV-ATAA: +29%; BAV: +40% vs. controls; P < 0.001). In differentiating patients with TAV-ATAA from their controls, the indexed ascending aorta volume showed better performances (AUC, 0.935 [95% confidence interval (CI): 0.882-0.989]; accuracy, 88.7% [95% CI: 82.9-94.5]) than indexed ascending aorta length (P < 0.001), indexed ascending aorta maximal diameter (P = 0.003) and indexed ascending aorta maximal CSA (P = 0.03). In differentiating patients with BAV from matched controls, indexed ascending aorta volume showed significantly better performances performance (AUC, 0.908 [95% CI: 0.829-0.987]; accuracy, 88.0% [95% CI: 80.9-95.0]) than indexed ascending aorta length (P = 0.02) and not different from indexed ascending aorta maximal diameter (P = 0.07) or from indexed ascending aorta maximal CSA (P = 0.27) CONCLUSION: Aortic volume measured by 3D-MRI integrates both elongation and luminal dilatation, resulting in greater classification performance than maximal diameter and length in differentiating patients with dilated ascending aorta or aneurysm from controls.

Aortic Stiffness Measured from Either 2D/4D Flow and Cine MRI or Applanation Tonometry in Coronary Artery Disease: A Case-Control Study.

BACKGROUND AND OBJECTIVE: Aortic stiffness can be evaluated by aortic distensibility or pulse wave velocity (PWV) using applanation tonometry, 2D phase contrast (PC) MRI and the emerging 4D flow MRI. However, such MRI tools may reach their technical limitations in populations with cardiovascular disease. Accordingly, this work focuses on the diagnostic value of aortic stiffness evaluated either by applanation tonometry or MRI in high-risk coronary artery disease (CAD) patients. METHODS: 35 patients with a multivessel CAD and a myocardial infarction treated 1 year before were prospectively recruited and compared with 18 controls with equivalent age and sex distribution. Ascending aorta distensibility and aortic arch 2D PWV were estimated along with 4D PWV. Furthermore, applanation tonometry carotid-to-femoral PWV (cf PWV) was recorded immediately after MRI. RESULTS: While no significant changes were found for aortic distensibility; cf PWV, 2D PWV and 4D PWV were significantly higher in CAD patients than controls (12.7 ± 2.9 vs. 9.6 ± 1.1; 11.0 ± 3.4 vs. 8.0 ± 2.05 and 17.3 ± 4.0 vs. 8.7 ± 2.5 m·s-1 respectively, p < 0.001). The receiver operating characteristic (ROC) analysis performed to assess the ability of stiffness indices to separate CAD subjects from controls revealed the highest area under the curve (AUC) for 4D PWV (0.97) with an optimal threshold of 12.9 m·s-1 (sensitivity of 88.6% and specificity of 94.4%). CONCLUSIONS: PWV estimated from 4D flow MRI showed the best diagnostic performances in identifying severe stable CAD patients from age and sex-matched controls, as compared to 2D flow MRI PWV, cf PWV and aortic distensibility.

Comprehensive assessment of local and regional aortic stiffness in patients with tricuspid or bicuspid aortic valve aortopathy using magnetic resonance imaging.

BACKGROUND: We aimed to provide a comprehensive aortic stiffness description using magnetic resonance imaging (MRI) in patients with ascending thoracic aorta aneurysm and tricuspid (TAV-ATAA) or bicuspid (BAV) aortic valve. METHODS: This case-control study included 18 TAV-ATAA and 19 BAV patients, with no aortic valve stenosis/severe regurgitation, who were 1:1 age-, gender- and central blood pressures (BP)-matched to healthy volunteers. Each underwent simultaneous aortic MRI and BP measurements. 3D anatomical MRI provided aortic diameters. Stiffness indices included: regional ascending (AA) and descending (DA) aorta pulse wave velocity (PWV) from 4D flow MRI; local AA and DA strain, distensibility and theoretical Bramwell-Hill (BH) model-based PWV, as well as regional arch PWV from 2D flow MRI. RESULTS: Patient groups had significantly higher maximal AA diameter (median[interquartile range], TAV-ATAA: 47.5[42.0-51.3]mm, BAV: 45.0[41.0-47.0]mm) than their respective controls (29.1[26.8-31.8] and 28.1[26.0-32.0]mm, p < 0.0001), while BP were similar (p ≥ 0.25). Stiffness indices were significantly associated with age (ρ ≥ 0.33), mean BP (arch PWV: ρ = 0.25, p = 0.05; DA distensibility: ρ = -0.30, p = 0.02) or AA diameter (arch PWV: ρ = 0.28, p = 0.03; DA PWV: ρ = 0.32, p = 0.009). None of them, however, was significantly different between TAV-ATAA or BAV patients and their matched controls. Finally, while direct PWV measures were significantly correlated to BH-PWV estimates in controls (ρ ≥ 0.40), associations were non-significant in TAV-ATAA and BAV groups (p ≥ 0.18). CONCLUSIONS: The overlap of MRI-derived aortic stiffness indices between patients with TAV or BAV aortopathy and matched controls highlights another heterogeneous feature of aortopathy, and suggests the urgent need for more sensitive indices which might help better discriminate such diseases.

3D aortic morphology and stiffness in MRI using semi-automated cylindrical active surface provides optimized description of the vascular effects of aging and hypertension.

BACKGROUND: Clinically, aortic geometry assessment is mainly based on the measurement of maximal diameters at different anatomic locations, which are subsequently used to indicate prophylactic aortic surgery. However, 3D evaluation of aortic morphology could provide volumetric quantification, which integrates both aortic dilatation and elongation and might thus be more sensitive to early geometric changes than diameters. Precise aortic morphology is also required for the calculation of pulse wave velocity (PWVMRI), an established marker of aortic stiffness. Accordingly, we proposed a 3D semi-automated analysis of thoracic aorta MRI data optimizing morphological and subsequent stiffness assessment. METHODS: We studied 74 individuals (40 males, 50 ± 12years): 21 healthy volunteers and 53 patients with hypertension in whom aortic 3D MRI angiography and 2D + t phase-contrast and cine imaging were performed. A semi-automated method was proposed for volumetric aortic segmentation and was evaluated by studying resulting measurements (length, diameters, volumes and PWVMRI) in terms of: 1) reproducibility, 2) correlations with well-established 2D aortic length and diameters, 3) associations with age, carotid-femoral PWV (cf-PWV) and presence of hypertension. RESULTS: The measurements obtained with the proposed method were reproducible (coefficients of variation ≤ 5.1%) and were highly correlated with 2D measurements (arch length: r = 0.80, Bland-Altman mean bias [limits]: 2.7 mm [-25; 30]; PWVMRI: r = 0.95, 0.22 m/s [-1.9; 2.4]). Higher or similar correlations with age were found for the proposed 3D method compared to the 2D approach (arch length: r = 0.47 (2D), r = 0.60 (3D); PWVMRI: r = 0.63 (2D), r = 0.64 (3D)). Moreover, a significant association was found between PWVMRI and cf-PWV (r = 0.49, p < 0.001). All aortic measurements increased with hypertension (p < 0.05) and with age: arch length (+9mm/decade); diameters: ascending (+1.2mm/decade) and descending aorta (+1.0mm/decade); volumes: ascending (+2.6mL/decade) and descending aorta (+4.0mL/decade); PWVMRI (+1.7  m s-1/decade). CONCLUSIONS: A semi-automated method based on cylindrical active surfaces was proposed for the 3D segmentation of the aorta using a single MRI dataset, providing aortic diameters at anatomical landmarks, aortic volumes and the aortic centerline length used for PWV estimation. Such measurements were reproducible and comparable to expert measurements, which required time-consuming centerline delineation. Furthermore, expected relationships with age and hypertension were found indicating the consistency of our measurements.