Mapping of thoracic aorta growth rate on serial self-navigated 3D whole-heart magnetic resonance angiographies by image registration

Introduction

Accurate and reproducible assessment of aortic diameters and their growth rate is of key importance for the management of patients with thoracic aortic aneurysms [1,2]. It has been recently shown that image registration permits the assessment of progressive aortic dilation on ECG-gated contrast-enhanced CT angiography, outperforming manual quantification and allowing for 3D aortic size and growth mapping [3]. However, exposure to radiation makes it convenient to limit the use of CT for serial follow-up, especially in young patients. Self-navigated 3D whole-heart CMR acquisitions provides excellent image quality overcoming these limitations [4].

Purpose

To evaluate the accuracy and reproducibility of registration-based assessment of aortic dilation using self-navigated 3D whole-heart CMR acquisitions.

Methods

Fifteen patients with two self-navigated 3D whole-heart CMR images obtained at least 1 year apart were included. Aortic root and thoracic aorta diameters were measured by 2 independent observers both manually (multiplanar reconstruction) and with the registration-based technique. To perform registration-based assessment, the aorta was semi-automatically segmented and typical anatomical landmarks were placed by each observer at baseline [3]. Geometrical mapping between baseline and follow-up acquisitions was obtained using deformable image registration, and applied to the baseline aortic surface points to obtain their location at follow-up. Finally, aortic diameters and their growth rate were automatically measured and used to calculated 3D aortic dilation maps. Agreement between techniques and their inter-observer reproducibility were calculated.

Results

Patients age was 27.2±14.5 years and 40% were male. Mean follow-up duration was 2.7±1.6 years. Compared to manual assessment, the registration-based technique presented low bias and excellent agreement for aortic diameters (Table 1), and low bias and moderate agreement for growth rates both in the aortic root and the thoracic aorta (Table, Fig. 1A). The techniques presented similar inter-observer reproducibility in the assessment of aortic diameters (Table 1), while the registration-based method demonstrated much higher inter-observer reproducibility in the assessment of growth rates in the aortic root and the thoracic aorta (Table 1, Fig. 1A and B). Three-dimensional mapping of thoracic aortic diameters and growth was highly reproducible (mean regional ICC=0.90 for diameters; 0.82 for growth rate).

Conclusion

The assessment of the dilation rate of the thoracic aorta via registration of serial self-navigated 3D whole-heart CMR acquisitions is accurate and reproducible in the aortic root and the thoracic aorta. Thus, it allows to assess local aortic growth without the drawbacks of CT.

Funding Acknowledgement

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Instituto de Salud Carlos III and Ministerio de Ciencia e Innovaciόn (Spain)

Circumferential wall shear stress predicts co-localized progressive dilation in bicuspid aortic valve patients

Background

Bicuspid aortic valve (BAV), a congenital heart defect, is associated with ascending aorta (AAo) dilation. Whether the high prevalence of dilation in BAV patients is related to alteration of aortic blood flow and thus in wall shear stress (WSS) [1,2], which have been associated with aortic wall degeneration [3], or intrinsic abnormalities of the aortic wall, such as altered aortic stiffness [4], has not been established. Recently, a technique for the semi-automatic quantification of progressive aortic dilation maps via image registration has been introduced [5].

Purpose

To test whether ascending aorta WSS predicts co-localized progressive dilation in BAV patients.

Methods

Forty BAV patients free from moderate and severe aortic valve regurgitation (regurgitant fraction <16%) and stenosis (maximum velocity at the aortic valve <3m/s), with no previous aortic or aortic valve surgery or replacement and included in a double-blind clinical trial (BICATOR, NCT02679261) were enrolled. All patients underwent a baseline 4D flow CMR study to assess aortic hemodynamics, followed by two contrast-enhanced computed tomography angiographies to quantify progressive dilation. WSS was computed at 64 pre-specified standardized ascending aortic regions, automatically obtained dividing the ascending aorta into 8 equidistant longitudinal sections which were further divided along the circumference into 8 equal regions (I = inner, L = left, O = outer and R = right) [2]. WSS was also projected into axial and circumferential directions, as previously described [1,2]. Progressive dilation was assessed in terms of growth rate (GR), i.e. increase in diameter divided by follow-up duration [mm/year], following a previously described methodology [5], at the same 64 pre-specified ascending aortic locations. A two-tailed p-value <0.05 was considered statistically significant.

Results

Demographic and clinical characteristics of the patients are shown in Table 1. WSS and growth rate maps are shown in Figure 1. Follow-up duration was 44.8±2.6 months. Growth rate (Figure 1A) was heterogeneously distributed, being highest (up to 0.26 mm/year) in the outer region of the mid AAo and in the inner region of the proximal-mid AAo. Circumferential WSS showed highest values in the outer region of the mid AAo (Figure 1C) while WSS (magnitude) and its axial component (Figure 1B and D) presented maximum values in the right region of the mid AAo. Maps of statistically significant association between GR and WSS values showed circumferential WSS to be correlated with GR in regions where progressive dilation was fastest, while WSS magnitude and its axial component resulted in limited associations with GR maps.

Conclusions

Circumferential wall shear stress predicts location-matched progressive dilation in bicuspid aortic valve patients.

Funding Acknowledgement

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): This study has received funding from the Instituto de Salud Carlos III (PI17/00381). Guala A. has received funding from Spanish Ministry of Science, Innovation and Universities (IJC2018-037349-I). Table 1. DemographicsFigure 1. GR and WSS maps and correlations

Accurate and reproducible aortic growth rate mapping via registration of serial contrast-enhanced computed tomography angiograms

Introduction

Accurate assessment of aortic diameters and growth rates is key for clinical management of patients with aortic aneurysms [1]. Manual assessment on multiplanar reformatted views of computed tomography angiograms (CTA) is recommended [1], although its reproducibility in the assessment of growth rates has not been reported [2]. Image registration has been proposed to provide 3D maps of aortic diameters and growth [3], but its accuracy and reproducibility have not been established.

Purpose

To quantify accuracy and inter-observer reproducibility of aortic root and thoracic aorta diameters and growth rate by registration of serial CTAs compared to current standard.

Methods

Forty non-operated patients with ≥2 contrast-enhanced ECG-gated CTA acquired at least 6 months apart were included. Aortic diameters and growth rates were measured in the aortic root and thoracic aorta by two independent observers, both with the current standard and with the registration-based technique. To perform registration-based assessment, each observer semi-automatically segmented the aorta at baseline and located typical anatomical landmarks (Fig. 1A). Then, deformable image registration was used to map baseline and follow-up CT scans and deformation was applied to the baseline aortic surface points to obtain their location at follow-up (Fig. 1B). Finally, aortic root diameters and growth rate and 3D maps of thoracic aortic diameters and growth rate were automatically obtained (Fig. 1C). Agreement between techniques and their inter-observer reproducibility were calculated.

Results

Follow-up duration was 3.3±1.5 years (range 0.52–6.2). Compared with manual assessment, registration-based aortic diameters presented low bias and excellent agreement in the aortic root (0.42 mm, ICC=0.99) and the thoracic aorta (0.55 mm, ICC=0.99), and similar inter-observer reproducibility (ICC=0.99 for both). Compared with manual assessment, registration-based growth rates presented low bias and good agreement in the aortic root (0.12 mm/y, ICC=0.84) and the thoracic aorta (0.03 mm/y, ICC=0.77) (Fig. 2A), and much higher inter-observer reproducibility (ICC=0.96 vs 0.68 in the aortic root, ICC=0.96 vs 0.80 in the thoracic aorta) (Fig. 2B and C). Registration-based aortic growth rates reproducibility at 6 months follow-up was comparable to that obtained by manual assessment at 2.7 years (LoA = [−0.01, 0.33] and LoA = [−0.13, 0.21], respectively). Aortic diameters and growth rate 3D maps were highly reproducible (ICC>0.9) in the whole thoracic aorta.

Conclusions

Progressive aortic dilation assessment via registration of CTAs is accurate and more reproducible than the current standard even over follow-ups as short as 6 months, and further provides robust 3D mapping of aortic diameters and growth rates. Its application may provide new insights in aneurysms pathophysiology and improve the clinical management of these patients.

Funding Acknowledgement

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): This study has received funding from the Instituto de Salud Carlos III (PI17/00381). Guala A. has received funding from Spanish Ministry of Science, Innovation and Universities (IJC2018-037349-I). Figure 1. Methodology.Figure 2. Growth rate comparison.

Automatic segmentation of the aorta on multi-center and multi-vendor phase-contrast enhanced magnetic resonance angiographies and the advantages of transfer learning

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Guala A. received funding from the Spanish Ministry of Science, Innovation and Universities

Background

Phase-contrast (PC) enhanced magnetic resonance (MR) angiography (MRA) is a class of angiogram exploiting velocity data to increase the signal-to-noise ratio, thus avoiding the administration of external contrast agent, normally used to segment 4D flow MR data. To train deep-learning algorithms to segment PC-MRA a large amount of manually annotated data is needed: however, the relatively novelty of the sequence, its rapid evolution and the extensive time needed to manually segment data limit its availability.

Purpose

The aim of this study was to test a deep learning algorithm in the segmentation of multi-center and multi-vendor PC-MRA and to test if transfer learning (TL) improves performance.

Methods

A large dataset (LD) of 262 and a small one (SD) of 22 PC-MRA, acquired without contrast agent at 1.5 T in a General Electric and a Siemens scanner, respectively, were manually annotated and divided into training (232 and 15 cases) and testing (30 and 7) sets. They both included PC-MRA of healthy subjects and patients with aortic diseases (excluding dissections) and native aorta. A convolutional neural networks (CNN) based on nnU-Net framework [1] was trained in the LD and another in the SD. The left ventricle was removed semi-automatically from the DL segmentations of the LD as it was not relevant for this application. Networks were then tested on the test sets of the dataset there were trained and the other dataset to assess generalizability. Finally, a fine-tuning transfer learning approach was applied to LD network and the performance on both test sets were tested. Dice score, Hausdorff distance, Jaccard score and Average Symmetrical Surface Distance were used as segmentation quality metrics.

Results

LD network achieved good performance in LD test set, with a DS of 0.904, ASSD of 1.47, J of 0.827 and HD of 6.35, which further improve after removing the left ventricle in the post-processing to a DS of 0.942, ASSD of 0.93, J of 0.892 and HD of 3.32. SD network results in an average DS of 0.895, ASSD of 0.59, J of 0.812 and HD of 2.05. Once tested on the testing set of the other dataset, LD network resulted in a DS of 0.612 while SD network in DS of 0.375, thus showing limited generalizability. However, the application of transfer learning to LD network resulted in the improvement of the evaluation metrics on the SD from a DS of 0.612 to 0.858, while slightly worsening in the first one without post-processing to 0.882.

Conclusions

nnU-net framework is effective for fast automatic segmentation of the aorta from multi-center and multi-vendor PC-MRA, showing performance comparable with the state of the art. The application of transfer learning allows for increased generalization to data from center not included in the original training. These results unlock the possibility for fully-automatic analysis of multi-vendor multi-center 4D flow MR.

Machine learning to automatically detect anatomical landmarks on phase-contrast enhanced magnetic resonance angiography

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Guala A. received funding from the Spanish Ministry of Science, Innovation and Universities

Introduction

The heterogeneous characteristic of the thoracic aorta implies that all biomarkers with potential for risk stratification need to be references to a specific location. This is the case, for example, of diameter [1], stiffness [2] and wall shear stress [3]. This is normally achieved by the manual identification of a limited number of key anatomic landmarks [4], which is a time-demanding task and may impact biomarkers accuracy and reproducibility. Automatic identification of these anatomic landmarks may speed-up the analysis and allow for the creation of fully automatic image analysis pipelines. Machine learning (ML) algorithms might be suitable for this task.

Purpose

The aim of this study was to test the performance of a ML algorithm in localizing key thoracic anatomical landmarks on phase-contrast enhanced magnetic resonance angiograms (PC-MRA).

Methods

PC-MRA of 323 patients with native aorta and aortic valve and a variety of aortic conditions (141 bicuspid aortic valve patients, 60 patients with degenerative aortic aneurysms, 82 patients with genetic aortopathy and 40 healthy volunteers) were included in this study. Four anatomical landmarks were manually identified on PC-MRA by an experienced researcher: sinotubular junction, the pulmonary artery bifurcation and the first and third supra-aortic vessel braches. A reinforcement learning algorithm (DQN), combining Q-learning with deep neural networks, was trained. The algorithm was tested in a separate set of 30 PC-MRA with similar distribution of aortic conditions in which human intra-observer reproducibility was quantified. The distance between points was used as quality metric and human annotation was considered as ground-truth. Repeated-measures ANOVA was used for statistical testing.

Results

ML algorithm resulted in performance similar to the intra-observer variability obtained by the experienced human reader in the identification of the sinotubular junction (11.1 ± 8.6 vs 11.0 ± 8.1 mm, p = 0.949) and first (6.8 ± 5.6 vs 6.6 ± 3.9 mm, p = 0.886) and third (8.4 ± 7.4 vs 6.8 ± 4.0 mm, p = 0.161) supra-aortic vessels branches. However, the algorithm did not reach human-level performance in the localization of the pulmonary artery bifurcation (15.2 ± 13.1 vs 10.2 ± 7.0 mm, p = 0.008). The time needed to the ML algorithm to locate all points ranged between 0.8 and 1.6 seconds on a standard computer while manual annotation required around two minutes to be performed.

Conclusions

The rapid identification of key aortic anatomical landmarks by a reinforced learning algorithm is feasible with human-level performance. This approach may thus be used for the design of fully-automatic pipeline for 4D flow CMR analysis.

Bicuspid aortic valve fusion length correlates with maximum aortic diameter and heamodynamic abnormalities: a 4D flow CMR study

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Guala A. received funding from the Spanish Ministry of Science, Innovation and Universities.

Background

Bicuspid aortic valve (BAV), a congenital heart defect, is associated with ascending aorta dilation, possibly via alteration of aortic blood flow [1]. In BAV abnormal flow condition have been associated with aortic extracellular matrix dysregulation and elastic fiber degeneration [2]. Current morphological classification of BAV patients with aortic valve with a single fusion between two adjacent leaflets does not allow for risk stratification.

Purpose

This research work tested whether the extent of fusion between leaflets is related to AAo diameter and flow alterations.

Methods

Ninety BAV patients free from moderate and severe aortic valve disease and with no previous aortic or aortic valve surgery or replacement were prospectively enrolled. A comprehensive magnetic resonance protocol comprised a stack of double-oblique 2D balanced steady-state free-precession (bSSFP) cine CMR of the aortic valve, which was used to measure the length of the fusion between leaflets, a cine CMR at the level of the pulmonary bifurcation to assess aortic diameter and 4D flow MRI sequence to assess flow characteristics and regional stiffness [3]. Jet angle and flow radial displacement, quantifying the extent of flow eccentricity, and systolic flow reversal ratio (SFRR), assessing the relative amount of backward flow during systole, were computed at 8 equidistant planes in the ascending aorta and 4 equidistant planes in the aortic arch [4]. A two-tailed p-value < 0.05 was considered statistically significant.

Results

The length of leaflet fusion varied widely (median 7.7 mm, inter-quartile range [5.5; 10.2]), Table 1). In bivariate analysis, fusion length was also associated to ascending aortic diameter (R = 0.391, p < 0.001), age (R = 0.313, p = 0.005) and body surface area (R = 0.396, p < 0.001). It was also positively related to flow abnormalities: like displacement in the proximal and distal ascending aorta, jet angle in the mid ascending aorta, and SFRR in the ascending aorta and the aortic arch (see Figure 1). The association between fusion length and ascending aorta diameter persisted in multivariate analysis after correction for age (p = 0.006).

Conclusions

Bicuspid aortic valve fusion extent varies greatly and it is associated with aortic diameter, possibly through flow alterations. Prospective longitudinal studies are needed to establish whether fusion length may allow for risk stratification in bicuspid aortic valve patients.

Aortic root longitudinal strain by speckle-tracking echocardiography predicts progressive aortic root dilation in Marfan syndrome patients

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Spanish Ministry of Science, Innovation and Universities; Instituto de Salud Carlos III

Introduction

In Marfan syndrome (MFS) patients reduced longitudinal strain of the ascending aorta (AAo) as measured by applying feature-tracking on cine cardiac magnetic resonance (CMR) images predicts aortic root dilation and aortic events during the follow-up. Speckle-tracking is well established for cardiac deformation assessment but proximal aorta applications are challenging due to limited wall thickness and substantial cardiac motion. Moreover, echocardiography is widely used in the clinical assessment aortic diseases.

Purpose

We aimed to test a speckle-tracking tool for root longitudinal strain analysis in terms of comparison with CMR-derived AAo longitudinal strain and reproducibility and as predictor of dilation in MFS patients.

Methods

Thirty-five MFS patients diagnosed by original GHENT criteria, with maximum aortic root diameter of 45 mm and free from previous aortic dissection or cardiac/aortic surgery and non-severe aortic regurgitation were consecutive enrolled and followed-up. CMR and echocardiography were performed less than 2 months apart. Baseline and final aortic root diameter were measured on CMR images. To quantify aortic root cyclic elongation by echocardiography, two regions of interests were manually created covering both walls in a parasternal long-axis view and tracked along the cardiac cycle (Figure 1). Longitudinal strain was computed as the average of maximum increase in relative distance of several sub-regions covering both walls. CMR-derived AAo longitudinal strain was available in 29 patients. Intra-observer reproducibility was tested in 15 patients via intraclass correlation coefficient (ICC) for single-rater absolute agreement.

Results

Aortic root longitudinal strain by echocardiography was mildly related to CMR-derived AAo longitudinal strain (R = 0.27) and was larger compared to CMR-derived values (16.2 ± 6.0 vs 11.3 ± 4.3). Reproducibility was high, with ICC of 0.811, R = 0.802, p < 0.001. After a mean follow up of 76 ± 13 months, aortic root diameter grew in 20 patients with a rate of 0.29± 0.24 mm/year. Overall mean growth-rate was 0.87 ± 0.33 mm/year. In multivariable analysis corrected for age and baseline aortic root diameter, baseline longitudinal strain by echocardiography was independently and inversely related to progressive dilation (p = 0.033).

Conclusions

The measurement of aortic root longitudinal strain by speckle-tracking echocardiography is feasible. Aortic root longitudinal strain is an independent predictor of progressive dilation in MFS patients. This may permit the improvement of risk-stratification in aortic diseases in large scale studies.

Abstract Figure 1

Leaflets fusion length in bicuspid aortic valve is related to ascending and aortic root dilation and ascending aorta wall shear stress

Funding Acknowledgements

Type of funding sources: Public Institution(s). Main funding source(s): Spanish Ministry of Science, Innovation and Universities, Instituto de Salud Carlos III

Background

Bicuspid aortic valve (BAV) is the most common congenital heart defect, consisting in the fusion of two aortic valve leaflets. Altered flow patterns have been related to aortic wall degeneration in BAV patients and may be responsible for the high prevalence of aortic disease in these patients. A number of studies on excised BAV or using advanced imaging modalities reported a wide variability of fusion extent between leaflet, but no previous study assessed whether leaflet fusion length may be used to stratify BAV patients.

Purpose

We aimed to test whether leaflet fusion extent can be quantified by cardiac magnetic resonance imaging (CMR) and whether it is related to aortic dilation and flow abnormalities in non-dysfunctional BAV.

Methods

One hundred and twenty BAV adults with no previous aortic or aortic valve surgery or significant valvular disease were consecutively enrolled. Patients with two sinuses of Valsalva (true BAV) or fusion of the left and non-coronary cusps, both being rare forms of BAV, were excluded. Twenty-eight healthy volunteers were also included for comparison. A 4D flow CMR sequence was acquired and circumferential wall shear stress and pulse wave velocity were assessed in the ascending aorta. A stack of double-oblique cine images of the aortic valve were used to quantify the length of the fusion between leaflets.

Results

The length of the fusion between leaflets was effectively measured in 112/120 patients (93%). Reproducibility was good (ICC = 0.826). Fusion length varied greatly (range 2.3 – 15.4 mm, 7.8 ± 3.2 mm, tertiles cut-off points were 6 and 9.3 mm). After correction for age, BSA, stroke volume and BAV fusion morphotype, fusion length was independently associated with diameter at the sinus of Valsalva (p = 0.002). Moreover, once corrected for age, stroke volume and ascending aorta pulse wave velocity, fusion length was positively related to ascending aorta diameter (p = 0.028). The comparison of maps of circumferential peak-systolic WSS in healthy volunteers (left) and BAV patients pertaining to the three leaflet fusion length tertiles is shown in Figure 1. Circumferential WSS progressively increase with larger fusion length. This trend was statistically significant (p < 0.05) in the right and outer regions of the proximal and mid ascending aorta.

Conclusions

Bicuspid aortic leaflet fusion length varies considerably and it is independently associated with ascending aorta and aortic root dilation, possibly through flow alterations.

Abstract Figure 1

Regional curvature in thoracic aortic aneurysms of different aetiologies and its relationship with established risk factors

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Spanish Ministry of Science, Innovation and Universities ; Instituto de Salud Carlos III

Introduction

The aorta is a 3D hollow, curvilinear elastic structure whose diseases have life-threatening consequences. Despite much effort has been paid to study aortic diameter, diameter is a poor predictor of events. Conversely, much less is known about aortic curvature, its distribution in the thoracic aorta and the potential impact of risk factors in aneurysms associated with different conditions. Currently, 4D flow magnetic resonance imaging (4D flow CMR) allows to obtain 3D geometry, 4D flow data and regional aortic stiffness.

Purpose

We aim to study regional aortic curvature in thoracic aorta aneurysms of different aetiologies and define its relationship with established risk factors.

Methods

One-hundred twenty patients (40 for each group, selected out of prospective cohorts of 156 bicuspid aortic valve – BAV-, 77 Marfan –MFS- and 67 patients with a degenerative aneurysm – TAVdeg-) were matched for age, sex and BSA via propensity score with 40 healthy volunteers (HV). The thoracic aorta was semi-automatically segmented from angiograms and the centreline was computed. Local curvature was assessed at 20 planes covering the thoracic aorta from the sinotubular junction to the proximal descending aorta (DAo) at the level of the pulmonary artery bifurcation. Local curvature was normalized by subject mean thoracic aorta curvature. Length was measured as centreline length. Aortic stiffness was measured in the DAo by pulse wave velocity (PWV). Aneurysm was defined by z-score ≥ 2 using diameters measured by double-oblique cine CMR.

Results

Matching was successful in all groups with the exception of a residual age difference between HV and TAVdeg. Curvature in HV showed a fairly smooth transition between the straighter ascending aorta (AAo) and DAo to a more curved aortic arch, with a peak in the mid aortic arch (Figure 1A).  Conversely, all patients’ groups presented a peak in curvature in the proximal DAo and a decreased local curvature in the aortic arch and mid DAo close to the level of the pulmonary artery. BAV and TAVdeg patients showed also increased curvature in the mid AAo, were dilation is prevalent. Conversely, in the same area MFS showed a reduced curvature and limited prevalence of aneurysm. In the overall population, age, AAo and root diameters, mean blood pressure, DAo PWV and aortic length, all established risk factors for aortic events, were inversely related to curvature in the distal AAo and aortic arch (Figure 1B).

Conclusions

Aneurysms related to different aetiologies show similar abnormalities in aortic curvature, with limited curvature in the aortic arch and a peak soon after the third supra-aortic vessel. Age, aortic diameter, length, stiffness and blood pressure, all known risk factors, are all related to reduced curvature in the distal ascending aorta and aortic arch.

Abstract Figure.

Reinforcement machine learning-based aortic anatomical landmarks detection from phase-contrast enhanced magnetic resonance angiography

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Spanish Ministry of Science, Innovation and Universities; La Marató de TV3

Introduction

Automatic analysis of medical imaging data may improve their clinical impact by reducing analysis time and improving reproducibility. Many medical imaging data, like 4D-flow magnetic resonance imaging (MRI), are often quantified regionally, implying the need for anatomical landmark identification to locate correspondences in the extracted data and compare among patients. Machine learning (ML) techniques hold potential for automatic analysis of medical imaging. Phase-contrast enhanced magnetic resonance angiography (PC-MRA) is a class of angiograms not requiring the administration of contrast agents.

Purpose

We aimed to test whether a machine learning algorithm can be trained to identify key anatomical cardiovascular landmarks on PC-MRA images and compare its performance with humans.

Methods

Three-hundred twenty-three aortic PC-MRA were manually annotated with the location of 4 landmarks: sinotubular junction, pulmonary artery bifurcation and first and third supra-aortic vessels (Figure 1), often used to separate the aorta in sub-regions. Patients included in the training dataset comprised healthy volunteers (40), bicuspid aortic valve patients (141), patients with degenerative aortic disease (60) and patients with genetically-triggered aortic disease (82), all without previous aortic surgery and with native aortic valve. PC-MRA images and manual annotations were used to train a DQN, a reinforcement learning algorithm that combines Q-learning with deep neural networks. The agents can navigate the images and optimally find the landmarks by following the policies learned during training. Data from thirty patients, distributed in terms of aortic condition as the training set, unseen by the algorithm in the training phase, were used to quantify intra-observer reproducibility and to assess ML algorithm performance. Distance between points was used as metric for comparisons, original human annotation was used as ground-truth and repeated-measures ANOVA was used for statistical testing.

Results

Human and machine learning performed similarly in the identification of the sinotubular junction (distance between points of 11.0 ± 8.1 vs. 11.1 ± 8.6 mm, respectively, p = 0.949) and first (6.6 ± 3.9 vs. 6.8 ± 5.6 mm, p = 0.886) and third (6.8 ± 4.0 vs. 8.4 ± 7.4 mm, p = 0.161) supra-aortic vessels branches but human annotation outperformed ML landmark detection in the identification of the pulmonary artery bifurcation (10.2 ± 7.0 vs. 15.2 ± 13.1 mm, p = 0.008). Computation time for landmark detection by ML was between 0.8 and 1.6 seconds on a standard computer while human annotation took approximatively two minutes.

Conclusions

ML-based aortic landmarks detection from phase-contrast enhanced magnetic resonance angiography is feasible and fast and performs similarly to human. Reinforced learning anatomical landmark identification unlock automatic extraction of a variety of regional aortic data, including complex 4D flow parameters.

Abstract Figure

Semi-automatic quantification of aortic root progressive dilation by automatic co-registration of computed tomography angiograms: a preliminary comparison with manual assessment in Marfan patients

Funding Acknowledgements

Type of funding sources: Public Institution(s). Main funding source(s): Spanish Ministry of Science, Innovation and Universities Instituto de Salud Carlos III

Background. Dilation of the aortic root is a key feature of Marfan syndrome and it is related to the occurrence of aortic events and death. On top of maximum diameter, rapid annual growth rate is suggested by guidelines for indication of aortic root replacement. Current gold-standard for aortic root diameter assessment is manual quantification on multiplanar reformatted 3D computed tomography (CT) or magnetic resonance angiogram. However, inter- and intra-observer reproducibility are limited and different measurement methods, i.e. cusp-to-cusp and cusp-to-commissure, may be used in different clinical centres, leading to difficulties in the clinical assessment of progressive dilation.

Purpose. We aimed to test whether aortic root growth rate during follow-up can be reliably quantified by semi-automatic co-registration of two CT angiograms.

Methods. Seven Marfan syndrome patients, free from previous aortic surgery, with a total of 11 pairs of CT were identified. Manual assessment of six aortic root diameters (right-non coronary -RN- , right-left -RL- and left-non coronary -LN- cusp-to-cusp and R, L and N cusp-to-commissure) was obtained from all CTs by an experienced researcher blind to semi-automatic results. The thoracic aorta and the outflow tract were semi-automatically segmented in the baseline CT and commissure and cusps were manually located. A 10 mm-thick region of interest containing the aortic wall was automatically generated from segmentation boundary. Co-registration was obtained with three, fully-automatic steps. Firstly, baseline and follow-up CT scans were aligned by means of a rigid registration. Then, scans were co-registered with multi-resolution affine followed by b-spline non-rigid registrations based on mutual information metric. The transformation pertaining to the location of baseline commissure and cusps points was used to locate the same points in the follow-up scan (Fig. 1 top).

Results. Follow-up duration was 35 ± 22 (range 12-70.3) months. Automatic quantification of diameter growth during the follow-up was obtained in 62 out of 66 (94%) diameter comparisons. High Pearson correlation coefficients (R) and ICC were found between manual and semi-automatic assessment of growth rate, both for cusp-to-cusp and cusp-to-commissure diameters: R = 0.727 and ICC = 0.678 for RN; R = 0.822 and ICC = 0.602 for RL; R = 0.648 and ICC = 0.668 for LN; R = 0.726 and ICC = 0.711 for R; R = 0.911 and ICC = 0.895 for L and R = 0.553 and ICC = 0.482 for N. Scatter and Bland-Altman plots for all growth rates (Fig. 1) confirmed very good correlation (R = 0.810) but a slight tendency (R=-0.270) for underestimation at high growth rate.  No correlation was found between follow-up duration and difference between techniques (R = 0.06).

Conclusions. Semi-automatic quantification of aortic root growth rate by co-registration of pairs of CT angiograms is feasible for follow-up as short as one year. Larger studies are needed to confirm these preliminary data.

Abstract Figure. CT measurements. Automatic vs manual.

Relationship between aortic distensibility and aortic regurgitation depending on aortic valve anatomy. A CMR study

Background

Aortic regurgitation (AR) can be evaluated by cardiac magnetic resonance (CMR).The most commonly used method to quantify AR is direct measurement using phase contrast (PC) imaging, at the aortic root (as close as possible to the aortic valve), for the calculation of regurgitant fraction (RF). Aortic distensibility (AD) may affect aortic valve dynamics and, as a result, aortic regurgitation grade. However, the impact of aortic distensibility in this evaluation remains unknown.

Purpose

The aim of the study was to evaluate the relationship between AD and AR in patients with different aortic valve anatomy.

Methods

213 patients with different AR severity grades and aortic valve anatomy (tricuspid (TAV) and bicuspid valve (BAV) patients) were enrolled (32.2% female, 74% BAV, 55.5±15.4 years), excluding connective tissue disease. All patients underwent a CMR study with PC sequences for the evaluation of regurgitant fraction at the aortic valve level. AR was considered as mild (<15%), moderate (15–30%) or severe (>30%) depending on RF value. Furthermore we used cine-sequences to estimate aortic diameters and distensibilities, using Art Fun software. Distensibility was calculated as (change in aortic area between systole and diastole/diastolic area)/brachial pulse pressure.

Results

159 (73.7%) AR were mild, 30 (14.1%) moderate and 24 (11.3%) severe. RF significantly correlated with aortic root diameter (r=0.337, p<0.001) and did not correlate with AD at the level of proximal descending aorta (r=0.121 and p=0.107). Furthermore descendig aorta distensibility correlated with age (r=−0.631, p<0.001) and aortic root diameter (r=−0.224, p=0.002). Dividing population in two different groups, depending on aortic valve anatomy, in TAV patients RF continued to not correlate with AD (r=0.159, p=0.369). In contrast, RF in BAV patients was positively correlated with AD (r=0.223, p=0.007) even after adjustment for aortic diameter and age in a multiple regression model (p<0.001, R2=0.478).

Conclusions

In our study, aortic regurgitation is positively related to descending aorta distensibility in BAV patients, regardless of age and aortic root diameter. Thus, AD may play a role in the evaluation of AR in case of bicuspid valves. In contrast, in TAV patients, distensibility does not seem to influence the assessment of AR severity.

Descending aorta distensibility

Funding Acknowledgement

Type of funding source: Other. Main funding source(s): Research grant provided by the Cardiopath PhD program

Patients with blunt traumatic thoracic aortic injury treated with TEVAR present increased flow dynamics alterations and pulse wave velocity: a 4D flow CMR study

Background

Thoracic endovascular aortic repair (TEVAR) is widely used for the treatment of blunt traumatic thoracic aortic injuries. Aortic flow dynamics and mechanical implications of this intervention are poorly investigated and may be of particular interest in the long-term follow-up of these mostly young patients.

Purpose

To assess whether the presence of TEVAR in a cohort of otherwise healthy subjects was related to dilation of the proximal aorta or increase in aortic stiffness and flow alterations.

Methods

Nineteen patients who underwent TEVAR implantation after a traumatic injury of the thoracic descending aorta (DAo) (10.0±6.1 years from intervention) and 44 healthy volunteers (HV) underwent 4D flow CMR to compute ascending aorta (AAo) pulse wave velocity (PWV), a marker of aortic stiffness, systolic flow reversal ratio (SFRR), quantifying backward flow during systole and in-plane rotational flow (IRF), measuring in-plane strength of helical flow. IRF and SFRR were assessed at 20 planes between the sinotubular junction and the mid thoracic DAo. Aortic diameters were measured using double-oblique cine CMR.

Results

Patients with TEVAR and HV did not differ in age, sex, body surface area, blood pressure and DAo diameter distal to TEVAR (Table). However, TEVAR patients presented larger diameters at the sinus of Valsalva and AAo, increased AAo PWV and strong flow alterations: IRF was reduced from the distal AAo to the proximal DAo, while SFRR was increased in the whole thoracic aorta (Figure).

Conclusions

In patients with blunt traumatic thoracic aortic injury treated with TEVAR the aorta proximal to TEVAR is dilated, stiffer and present potentially pathogenic flow conditions. Longitudinal studies are needed to assess whether these alterations have prognostic value and may improve clinical prevention and management of these patients.

Figure 1. IRF and SFRR in healthy vs TEVAR

Funding Acknowledgement

Type of funding source: Public grant(s) – National budget only. Main funding source(s): This study has been funded by Instituto de Salud Carlos III, Spanish Ministry of Science and Innovation (PI19/01480). Guala A. received funding from the Spanish Ministry of Science, Innovation and Universities (IJC2018-037349-I).

The length of the fusion between leaflets in bicuspid aortic valve is independently related to ascending aorta dilation and flow dynamics alterations assessed by 4D-flow CMR

Background

Aortic dilation in bicuspid aortic valve (BAV) patients has been related to altered flow patterns, which contribute to aortic wall degeneration. However, preventive aortic replacement is currently based on a diameter threshold. Several studies on excised BAV reported wide variability of fusion extent.

Purpose

To unveil whether leaflet fusion extent can be quantified by CMR and is related to aortic dilation and flow abnormalities in non-dysfunctional BAV.

Methods

One hundred and twenty adults with non-dysfunctional BAV and no previous aortic or aortic valve surgery and 28 healthy volunteers underwent double-oblique cine and 4D flow CMR. BAV patients with two sinuses of Valsalva or left and non-coronary cusps fusion were excluded. Peak systolic circumferential wall shear stress (WSSc) and pulse wave velocity (PWV) in the ascending aorta (AAo) were assessed by 4D flow CMR. Fusion length between leaflets was measured using a stack of double-oblique cine CMR images of the aortic valve.

Results

The length of the fusion was effectively measured in 112/120 (93%) patients with good reproducibility (ICC = 0.826) and showed great variability (range 2.3–15.4 mm, 7.8±3.2 mm and tertiles cut-off points 6 and 9.3 mm). In multivariate analysis adjusted for clinical and demographic characteristics and PWV, fusion length was independently associated with the diameter at the sinus of Valsalva (p=0.002) and the AAo (p=0.02) (Table). WSSc progressively increased with larger fusion length (Figure), with statistical significance (p<0.05) in the right and outer regions of the proximal and mid AAo.

Conclusions

Bicuspid aortic leaflet fusion length varies considerably, and it is independently associated with AAo and aortic root dilation, possibly through flow alterations.

Figure 1. Maps of circumferential WSS

Funding Acknowledgement

Type of funding source: Public grant(s) – National budget only. Main funding source(s): This study has been partially funded by Instituto Carlos III, Spanish Ministry of Science and Innovation (PI17/00381). Guala A. has received funding from the Spanish Ministry of Science, Innovation and Universities (IJC2018-037349-I).

The role of descending aorta diastolic reverse flow in the quantification of aortic regurgitation by CMR

Background

EACVI recommends the use of an “integrative approach”, using several parameters, in aortic regurgitation (AR) quantification. This approach is easily achieved by echocardiography although cardiovascular magnetic resonance (CMR) remains the gold standard for the quantification of regurgitant fraction (RF).

Purpose

The aim of the study was to analyze the accuracy of descending aorta (DA) diastolic reverse flow in the assessment of chronic AR severity by CMR to identify an additional parameter.

Methods

188 patients (34% female, 54.6±15.6 years) with different severity grades of chronic AR were enrolled. All patients underwent a CMR study. Aortic regurgitation was considered as absent (≤1%), mild (≤15%), moderate (≤15%) or severe (≥30%) depending on RF value at valve level. Furthermore, cine-sequences were used to estimate aortic diameters and distensibilities using Art Fun software. Velocity-time integral (VTI) of reverse flow in DA was calculated from maximum velocity curves by an in-house MatLab code.

Results

AR was absent in 21 (12%) patients, mild in 114 (62.9%), moderate in 23 (12.7%) and severe in 21 (11.6%).DA VTI reverse flow was significantly higher as was the RF at valve level (r=0.805, p<0.001) (IMG, Table). It also positively correlated withaortic root diameter (r=0.347, p<0.001) and DA distensibility (r=0.279, p<0.001). It did not correlate with age (r=−0.91, p=0.22). In a statistically significant multiple regression model (p<0.001, R2 = 0.697), although VTI reverse flow in DA correlated strongly with RF at valve level (p<0.001; beta = 0.733), it was also influenced by DA distensibility (p<0.001; beta = 0.197) and aortic root diameter (p<0.001; beta= 0.140).

Conclusions

VTI reverse flow in DA correlates strongly with the degree of AR and may be useful in the assessment of its severity. Neverthless, owing to the influence of other factors (aortic distensibility and aortic root diameter), it cannot be used as a single parameter in the quantification of AR severity by CMR.

Scatter Plot graphs

Funding Acknowledgement

Type of funding source: Other. Main funding source(s): research grant provided by the Cardiopath PhD program

P1601 Relationship between aortic distensibility and aortic regurgitation assessed by CMR in bicuspid valve patients

Funding Acknowledgements

Spanish Ministry of Economy and Competitiveness RTC-2016-5152-1, ISCIII PI17/00381, La Marató de TV3 (20151330), Eur FP7/People 267128 and CIBERCV

BACKGROUND

The severity of aortic regurgitation can be evaluated with cardiac magnetic resonance (CMR) through calculation of regurgitant fraction (RF) in phase contrast sequences acquired at the aortic root (as close as possible to the aortic valve). However, the impact of aortic distensibility in this evaluation remains unkown.

PURPOSE

The aim of the study was to evaluate the relation between aortic distensibility and RF valve in bicuspid aortic valve patients.

METHODS

We enrolled bicuspid aortic valve patients without significant aortic stenosis (maximum velocity <2.5 m/s) and connective tissue disease. All patients underwent a CMR study with phase contrast sequences for evaluation of regurgitant fraction at the level of the aortic valve. Aortic regurgitation was considered as mild, moderate or severe depending on RF value (mild <15%; moderate 15-30%; severe >30%). Furthermore we used cine-sequences of aortic root, ascending and proximal descending aorta to estimate aortic diameters and distensibilities, using Art Fun software. Distensibility was calculated as (change in aortic area between systole and diastole/diastolic area)/brachial pulse pressure.

RESULTS

A total of 98 bicuspid aortic valve patients were included (30% female, 49.7 ± 14.5 years). 75 (76,5%) AR was mild, 17 (17,4%) moderate and 6 (6,1%) severe. RF valvewas significantly correlated with aortic root diameter (r= 0.430 y p < 0.001 )and aortic distensibility at the level of the ascending (r = 0.273 p =0.016) and descending aorta (r = 0.502 and p< 0.001). Aortic distensibility was positively correlated with RFvalve even after adjustment for aortic diameter ( p = 0.002 and p <0.001 respectively) . (Table) (IMG)

CONCLUSIONS

In our study, aortic regurgitation in bicuspid valve patients, evaluated by CMR using RF valve, is related to aortic distensibility. Thus, aortic distensibility should be included in the evaluation of aortic regurgitation by CMR as additional parameter. However, longitudinal studies are needed to evaluate the impact of including aortic distensibility in the evaluation of AR severity by CMR.

AR SEVERITY MILD MODERATE SEVERE Descending aorta distensibility(mean ± std. deviation) 2693,68 ± 997,5 3285,8 ±1952,7 5042,99 ±2873,44 Correlation between AR severity (by RFvalve) and descending aorta distensibility

Abstract P1601 Figure.

P1447 Ascending aorta longitudinal strain in bicuspid aortic valve patients: a comparison with healthy volunteers and patients with degenerative aortic aneurysm

Funding Acknowledgements

Spanish Ministry of Economy and Competitiveness RTC-2016-5152-1, ISCIII PI17/00381, La Marató de TV3 (20151330), Eur FP7/People 267128 and CIBERCV

Background

Histological findings of fibrillin-1 deficiency in bicuspid aortic valve (BAV) ascending aorta (AAo), as observed in Marfan (MFS), supported the existence of intrinsic aortic wall abnormalities, but recent studies reported the absence of an intrinsic impairment in stiffness. A recent study in MFS showed that AAo longitudinal strain was reduced in MFS and predicted dilation and aortic events. This parameter has not been studied in BAV.

Purpose

We investigated whether ascending aorta longitudinal strain is intrinsically altered in BAV with respect to tricuspid aortic valve (TAV) individuals.

Methods

80 BAV, 31 healthy volunteers (HV) and 29 TAV with AAo aneurysm, all without moderate valvular disease, were consecutively included. AAo dilation was defined as a z-score > 2. The 1.5T CMR protocol included a set of 2D cine CMR stacks covering the proximal aorta in saggital, coronal and axial views. AAo longitudinal strain was computed by an in-house Matlab code performing a feature tracking of the aortic valve in each of the cine images.

Results

Twenty (25%) of BAV had AAo dilation. AAo longitudinal strain was lower in non-dilated BAV compared to HV, but the difference was not significant in multivariate analysis adjusted for AAo diameter and systolic blood pressure. Similarly, the difference between dilated BAV and dilated TAV found in univariate analysis was not confirmed by multivariate analysis. On the other hand, both dilated BAV and TAV showed decreased AAo longitudinal strain compared to HV, which were confirmed in multivariate analyses.

Conclusions

AAo longitudinal strain, a marker of aortic stiffness with predictive value in MFS, is not altered in BAV patients compared to TAV matched for dilation prevalence. Reduced AAo longitudinal strain was independently associated with dilation in both BAV and TAV.

Table 1 HV vs. NON-DILATED BAV DILATED BAV vs DILATED TAV HV vs. DILATED BAV HV vs. DILATED TAV HV NON-DILATED BAV Univariate /multivariate p-value DILATED TAV DILATED BAV Univariate /multivariate p-value Univariate/ Multivariate p-value Univariate p-value N 31 20 29 60 Age [years 35 ± 8 49 ± 16 <0.001/ NS 66 ± 13 49 ± 14 <0.001 / <0.001 <0.001 / 0.052 <0.001 / NS Sex [% male] 42 35 0.629 24 42 0.097 / NS 0.969 0.149 BSA [m2] 1.83 ± 0.17 1.81 ± 0.14 0.702 1.95 ± 0.24 1.82 ± 0.22 0.015 / <0.001 0.881 0.030 / NS SBP [mmHg] 119 ± 11 132 ± 16 0.002 / 0.029 133 ± 17 138 ± 19 0.304 <0.001 / NS <0.001 / NS DBP [mmHg] 69 ± 11 73 ± 6 0.099 / NS 77 ± 9 79 ± 11 0.455 <0.001 / 0.016 0.004 / 0.023 Ascending aorta diameter [mm] 26 ± 4 33 ± 3 <0.001 / 0.006 46 ± 7 43 ± 6 0.032 / NS <0.001 / 0.001 <0.001 /0.007 AAo long strain [%] 10.5 ± 3.6 8.4 ± 4.1 0.067/ NS 5.9 ± 2.7 7.7 ± 3.6 0.023 / NS 0.001 / 0.002 <0.001 / 0.023 Demographics and uni- and multivariate analyses of AAo longitudinal strain