Random effects during training: Implications for deep learning-based medical image segmentation.

BACKGROUND: A single learning algorithm can produce deep learning-based image segmentation models that vary in performance purely due to random effects during training. This study assessed the effect of these random performance fluctuations on the reliability of standard methods of comparing segmentation models. METHODS: The influence of random effects during training was assessed by running a single learning algorithm (nnU-Net) with 50 different random seeds for three multiclass 3D medical image segmentation problems, including brain tumour, hippocampus, and cardiac segmentation. Recent literature was sampled to find the most common methods for estimating and comparing the performance of deep learning segmentation models. Based on this, segmentation performance was assessed using both hold-out validation and 5-fold cross-validation and the statistical significance of performance differences was measured using the Paired t-test and the Wilcoxon signed rank test on Dice scores. RESULTS: For the different segmentation problems, the seed producing the highest mean Dice score statistically significantly outperformed between 0 % and 76 % of the remaining seeds when estimating performance using hold-out validation, and between 10 % and 38 % when estimating performance using 5-fold cross-validation. CONCLUSION: Random effects during training can cause high rates of statistically-significant performance differences between segmentation models from the same learning algorithm. Whilst statistical testing is widely used in contemporary literature, our results indicate that a statistically-significant difference in segmentation performance is a weak and unreliable indicator of a true performance difference between two learning algorithms.

Multi-Modal in Vitro Experiments Mimicking the Flow Through a Mitral Heart Valve Phantom.

PURPOSE: Fluid-structure interaction (FSI) models are more commonly applied in medical research as computational power is increasing. However, understanding the accuracy of FSI models is crucial, especially in the context of heart valve disease in patient-specific models. Therefore, this study aimed to create a multi-modal benchmarking data set for cardiac-inspired FSI models, based on clinically important parameters, such as the pressure, velocity, and valve opening, with an in vitro phantom setup. METHOD: An in vitro setup was developed with a 3D-printed phantom mimicking the left heart, including a deforming mitral valve. A range of pulsatile flows were created with a computer-controlled motor-and-pump setup. Catheter pressure measurements, magnetic resonance imaging (MRI), and echocardiography (Echo) imaging were used to measure pressure and velocity in the domain. Furthermore, the valve opening was quantified based on cine MRI and Echo images. RESULT: The experimental setup, with 0.5% cycle-to-cycle variation, was successfully built and six different flow cases were investigated. Higher velocity through the mitral valve was observed for increased cardiac output. The pressure difference across the valve also followed this trend. The flow in the phantom was qualitatively assessed by the velocity profile in the ventricle and by streamlines obtained from 4D phase-contrast MRI. CONCLUSION: A multi-modal set of data for validation of FSI models has been created, based on parameters relevant for diagnosis of heart valve disease. All data is publicly available for future development of computational heart valve models.

Validation of fluid-structure interaction simulations of the opening phase of phantom mitral heart valves under physiologically inspired conditions.

BACKGROUND AND OBJECTIVE: Atrioventricular valve disease is a common cause of heart failure, and successful surgical or interventional outcomes are crucial. Patient-specific fluid-structure interaction (FSI) modeling may provide valuable insights into valve dynamics and guidance of valve repair strategies. However, lack of validation has kept FSI modeling from clinical implementation. Therefore, this study aims to validate FSI simulations against in vitro benchmarking data, based on clinically relevant parameters for evaluating heart valve disease. METHODS: An FSI model that mimics the left heart was developed. The domain included a deformable mitral valve of different stiffnesses run with different inlet velocities. Five different cases were simulated and compared to in vitro data based on the pressure difference across the valve, the valve opening, and the velocity in the flow domain. RESULTS: The simulations underestimate the pressure difference across the valve by 6.8-14 % compared to catheter measurements. Evaluation of the valve opening showed an underprediction of 5.4-7.3 % when compared to cine MRI, 2D Echo, and 3D Echo data. Additionally, the simulated velocity through the valve showed a 7.9-8.4 % underprediction in relation to Doppler Echo measurements. Qualitative assessment of the velocity profile in the ventricle and the streamlines of the flow in the domain showed good agreement of the flow behavior. CONCLUSIONS: Parameters relevant to the diagnosis of heart valve disease estimated by FSI simulations showed good agreement when compared to in vitro benchmarking data, with differences small enough not to affect the grading of heart valve disease. The FSI model is thus deemed good enough for further development toward patient-specific cases.

Factors affecting performance of fetal blood T2 measurements for noninvasive estimation of oxygen saturation.

PURPOSE: To ultimately make accurate and precise fetal noninvasive oxygen saturation (sO2 ) measurements by T2 -prepared bSSFP more widely available by systematically assessing error sources in order to potentially reduce perinatal mortality in cardiovascular malformations and fetal growth restriction. METHODS: T2 -prepared bSSFP data were acquired in phantoms; in flowing blood in adults in the superior sagittal sinus, ascending and descending aorta, and main pulmonary artery; and in the fetal descending aorta and umbilical vein. T2 was assessed in relation to T2 two- or three-parameter curve-fitting techniques, SSFP readout, refocusing time delay (τ), constant and pulsatile blood flow, and impact of T1 recovery. Further, fetal T2 and sO2 variability were quantified in the descending aorta and umbilical vein in healthy fetuses and fetuses with cardiovascular malformation (gestational weeks 32-38). RESULTS: In phantoms, three-parameter fitting was accurate irrespective of phase FOV (<4 ms; i.e., <2%), and T2 was overestimated (up to 23 ms/10%; p = 0.001) beyond ±30 Hz off-resonance. In the adult aorta, T2 was underestimated during higher blood flow velocities and pulsatility for τ = 16 ms (-41 ms/-17%; p = 0.008). In fetuses, two-parameter fitting overestimated T2 compared with three-parameter fitting (+33 ms/+18%; p = 0.03). T2 variability was 18 ms/15% in the fetal descending aorta and 28 ms/14% in the umbilical vein. The resulting estimated sO2 variability was ∼10% (15% of sO2 value) in the fetal descending aorta. CONCLUSIONS: Errors due to T2 -fitting techniques, off-resonance, flow velocity, and insufficient T1 recovery between image acquisitions could be mitigated by using three-parameter fitting with included saturation-prepared images approximating infinite T2 -preparation time, adequate shimming covering the fetus and placenta, and by modifying acquisition parameters. Variability in fetal blood T2 and sO2 , however, indicate that it is currently not feasible to use these methods for prediction of disease.

4D Flow cardiovascular magnetic resonance consensus statement: 2023 update.

Hemodynamic assessment is an integral part of the diagnosis and management of cardiovascular disease. Four-dimensional cardiovascular magnetic resonance flow imaging (4D Flow CMR) allows comprehensive and accurate assessment of flow in a single acquisition. This consensus paper is an update from the 2015 '4D Flow CMR Consensus Statement'. We elaborate on 4D Flow CMR sequence options and imaging considerations. The document aims to assist centers starting out with 4D Flow CMR of the heart and great vessels with advice on acquisition parameters, post-processing workflows and integration into clinical practice. Furthermore, we define minimum quality assurance and validation standards for clinical centers. We also address the challenges faced in quality assurance and validation in the research setting. We also include a checklist for recommended publication standards, specifically for 4D Flow CMR. Finally, we discuss the current limitations and the future of 4D Flow CMR. This updated consensus paper will further facilitate widespread adoption of 4D Flow CMR in the clinical workflow across the globe and aid consistently high-quality publication standards.

Comparison of 2D and 4D Flow MRI in Neonates Without General Anesthesia.

BACKGROUND: Neonates with critical congenital heart disease require early intervention. Four-dimensional (4D) flow may facilitate surgical planning and improve outcome, but accuracy and precision in neonates are unknown. PURPOSE: To 1) validate two-dimensional (2D) and 4D flow MRI in a phantom and investigate the effect of spatial and temporal resolution; 2) investigate accuracy and precision of 4D flow and internal consistency of 2D and 4D flow in neonates; and 3) compare scan time of 4D flow to multiple 2D flows. STUDY TYPE: Phantom and prospective patients. POPULATION: A total of 17 neonates with surgically corrected aortic coarctation (age 18 days [IQR 11-20]) and a three-dimensional printed neonatal aorta phantom. FIELD STRENGTH/SEQUENCE: 1.5T, 2D flow and 4D flow. ASSESSMENT: In the phantom, 2D and 4D flow volumes (ascending and descending aorta, and aortic arch vessels) with different resolutions were compared to high-resolution reference 2D flow. In neonates, 4D flow was compared to 2D flow volumes at each vessel. Internal consistency was computed as the flow volume in the ascending aorta minus the sum of flow volumes in the aortic arch vessels and descending aorta, divided by ascending aortic flow. STATISTICAL TESTS: Bland-Altman plots, Pearson correlation coefficient (r), and Student's t-tests. RESULTS: In the phantom, 2D flow differed by 0.01 ± 0.02 liter/min with 1.5 mm spatial resolution and -0.01 ± 0.02 liter/min with 0.8 mm resolution; 4D flow differed by -0.05 ± 0.02 liter/min with 2.4 mm spatial and 42 msec temporal resolution, -0.01 ± 0.02 liter/min with 1.5 mm, 42 msec resolution and -0.01 ± 0.02 liter/min with 1.5 mm, 21 msec resolution. In patients, 4D flow and 2D flow differed by -0.06 ± 0.08 liter/min. Internal consistency in patients was -11% ± 17% for 2D flow and 5% ± 13% for 4D flow. Scan time was 17.1 minutes [IQR 15.5-18.5] for 2D flow and 6.2 minutes [IQR 5.3-6.9] for 4D flow, P < 0.0001. DATA CONCLUSION: Neonatal 4D flow MRI is time efficient and can be acquired with good internal consistency without contrast agents or general anesthesia, thus potentially expanding 4D flow use to the youngest and smallest patients. EVIDENCE LEVEL: 1 TECHNICAL EFFICACY: Stage 2.

Fetal 3D cardiovascular cine image acquisition using radial sampling and compressed sensing.

PURPOSE: To explore a fetal 3D cardiovascular cine acquisition using a radial image acquisition and compressed-sensing reconstruction and compare image quality and scan time with conventional multislice 2D imaging. METHODS: Volumetric fetal cardiac data were acquired in 26 volunteers using a radial 3D balanced SSFP pulse sequence. Cardiac gating was performed using a Doppler ultrasound device. Images were reconstructed using a parallel-imaging and compressed-sensing algorithm. Multiplanar reformatting to standard cardiac views was performed before image analysis. Clinical 2D images were used for comparison. Qualitative and quantitative image evaluation were performed by two experienced observers (scale: 1-4). Volumes, mass, and function were assessed. RESULTS: Average scan time for the 3D imaging was 6 min, including one localizer. A 2D imaging stack covering the entire heart including localizer sequences took at least 6.5 min, depending on planning complexity. The 3D acquisition was successful in 7 of 26 subjects (27%). Overall image contrast and perceived resolution were lower in the 3D images. Nonetheless, the 3D images had, on average, a moderate cardiac diagnostic quality (median [range]: 3 [1-4]). Standard clinical 2D acquisitions had a high cardiac diagnostic quality (median [range]: 4 [3, 4]). Cardiac measurements were not different between 2D and 3D images (all p > 0.16). CONCLUSION: The presented free-breathing whole-heart fetal 3D radial cine MRI acquisition and reconstruction method enables retrospective visualization of all cardiac views while keeping examination times short. This proof-of-concept work produced images with diagnostic quality, while at the same time reducing the planning complexity to a single localizer.

Increased biventricular hemodynamic forces in precapillary pulmonary hypertension.

Precapillary pulmonary hypertension (PHprecap) is a condition with elevated pulmonary vascular pressure and resistance. Patients have a poor prognosis and understanding the underlying pathophysiological mechanisms is crucial to guide and improve treatment. Ventricular hemodynamic forces (HDF) are a potential early marker of cardiac dysfunction, which may improve evaluation of treatment effect. Therefore, we aimed to investigate if HDF differ in patients with PHprecap compared to healthy controls. Patients with PHprecap (n = 20) and age- and sex-matched healthy controls (n = 12) underwent cardiac magnetic resonance imaging including 4D flow. Biventricular HDF were computed in three spatial directions throughout the cardiac cycle using the Navier-Stokes equations. Biventricular HDF (N) indexed to stroke volume (l) were larger in patients than controls in all three directions. Data is presented as median N/l for patients vs controls. In the RV, systolic HDF diaphragm-outflow tract were 2.1 vs 1.4 (p = 0.003), and septum-free wall 0.64 vs 0.42 (p = 0.007). Diastolic RV HDF apex-base were 1.4 vs 0.87 (p < 0.0001), diaphragm-outflow tract 0.80 vs 0.47 (p = 0.005), and septum-free wall 0.60 vs 0.38 (p = 0.003). In the LV, systolic HDF apex-base were 2.1 vs 1.5 (p = 0.005), and lateral wall-septum 1.5 vs 1.2 (p = 0.02). Diastolic LV HDF apex-base were 1.6 vs 1.2 (p = 0.008), and inferior-anterior 0.46 vs 0.24 (p = 0.02). Hemodynamic force analysis conveys information of pathological cardiac pumping mechanisms complementary to more established volumetric and functional parameters in precapillary pulmonary hypertension. The right ventricle compensates for the increased afterload in part by augmenting transverse forces, and left ventricular hemodynamic abnormalities are mainly a result of underfilling rather than intrinsic ventricular dysfunction.

Real-time imaging of respiratory effects on cerebrospinal fluid flow in small diameter passageways.

PURPOSE: Respiration-related CSF flow through the cerebral aqueduct may be useful for elucidating physiology and pathophysiology of the glymphatic system, which has been proposed as a mechanism of brain waste clearance. Therefore, we aimed to (1) develop a real-time (CSF) flow imaging method with high spatial and sufficient temporal resolution to capture respiratory effects, (2) validate the method in a phantom setup and numerical simulations, and (3) apply the method in vivo and quantify its repeatability and correlation with different respiratory conditions. METHODS: A golden-angle radial flow sequence (reconstructed temporal resolution 168 ms, spatial resolution 0.6 mm) was implemented on a 7T MRI scanner and reconstructed using compressed sensing. A phantom setup mimicked simultaneous cardiac and respiratory flow oscillations. The effect of temporal resolution and vessel diameter was investigated numerically. Healthy volunteers (n = 10) were scanned at four different respiratory conditions, including repeat scans. RESULTS: Phantom data show that the developed sequence accurately quantifies respiratory oscillations (ratio real-time/reference QR  = 0.96 ± 0.02), but underestimates the rapid cardiac oscillations (ratio QC  = 0.46 ± 0.14). Simulations suggest that QC can be improved by increasing temporal resolution. In vivo repeatability was moderate to very strong for cranial and caudal flow (intraclass correlation coefficient range: 0.55-0.99) and weak to strong for net flow (intraclass correlation coefficient range: 0.48-0.90). Net flow was influenced by respiratory condition (p < 0.01). CONCLUSIONS: The presented real-time flow MRI method can quantify respiratory-related variations of CSF flow in the cerebral aqueduct, but it underestimates rapid cardiac oscillations. In vivo, the method showed good repeatability and a relationship between flow and respiration.

Computational Fluid Dynamics Support for Fontan Planning in Minutes, Not Hours: The Next Step in Clinical Pre-Interventional Simulations.

Computational fluid dynamics (CFD) modeling may aid in planning of invasive interventions in Fontan patients. Clinical application of current CFD techniques is however limited by complexity and long computation times. Therefore, we validated a "lean" CFD method to magnetic resonance imaging (MRI) and an "established" CFD method, ultimately aiming to reduce complexity to enable predictive CFD during ongoing interventions. Fifteen Fontan patients underwent MRI for CFD modeling. The differences between lean and established approach, in hepatic and total flow percentage to the left pulmonary artery (%LPA), power loss and relative wall shear stress area were 1.5 ± 4.0%, -0.17 ± 1.1%, -0.055 ± 0.092 mW and 1.1 ± 1.4%. Compared with MRI, the lean and established method showed a bias in %LPA of -1.9 ± 3.4% and -1.8 ± 3.1%. Computation time was for the lean and established approach 3.0 ± 2.0 min and 7.0 ± 3.4 h, respectively. We conclude that the proposed lean method provides fast and reliable results for future CFD support during interventions.

A multispeaker dataset of raw and reconstructed speech production real-time MRI video and 3D volumetric images.

Real-time magnetic resonance imaging (RT-MRI) of human speech production is enabling significant advances in speech science, linguistics, bio-inspired speech technology development, and clinical applications. Easy access to RT-MRI is however limited, and comprehensive datasets with broad access are needed to catalyze research across numerous domains. The imaging of the rapidly moving articulators and dynamic airway shaping during speech demands high spatio-temporal resolution and robust reconstruction methods. Further, while reconstructed images have been published, to-date there is no open dataset providing raw multi-coil RT-MRI data from an optimized speech production experimental setup. Such datasets could enable new and improved methods for dynamic image reconstruction, artifact correction, feature extraction, and direct extraction of linguistically-relevant biomarkers. The present dataset offers a unique corpus of 2D sagittal-view RT-MRI videos along with synchronized audio for 75 participants performing linguistically motivated speech tasks, alongside the corresponding public domain raw RT-MRI data. The dataset also includes 3D volumetric vocal tract MRI during sustained speech sounds and high-resolution static anatomical T2-weighted upper airway MRI for each participant.

Multicenter Consistency Assessment of Valvular Flow Quantification With Automated Valve Tracking in 4D Flow CMR.

OBJECTIVES: This study determined: 1) the interobserver agreement; 2) valvular flow variation; and 3) which variables independently predicted the variation of valvular flow quantification from 4-dimensional (4D) flow cardiac magnetic resonance (CMR) with automated retrospective valve tracking at multiple sites. BACKGROUND: Automated retrospective valve tracking in 4D flow CMR allows consistent assessment of valvular flow through all intracardiac valves. However, due to the variance of CMR scanners and protocols, it remains uncertain if the published consistency holds for other clinical centers. METHODS: Seven sites each retrospectively or prospectively selected 20 subjects who underwent whole heart 4D flow CMR (64 patients and 76 healthy volunteers; aged 32 years [range 24 to 48 years], 47% men, from 2014 to 2020), which was acquired with locally used CMR scanners (scanners from 3 vendors; 2 1.5-T and 5 3-T scanners) and protocols. Automated retrospective valve tracking was locally performed at each site to quantify the valvular flow and repeated by 1 central site. Interobserver agreement was evaluated with intraclass correlation coefficients (ICCs). Net forward volume (NFV) consistency among the valves was evaluated by calculating the intervalvular variation. Multiple regression analysis was performed to assess the predicting effect of local CMR scanners and protocols on the intervalvular inconsistency. RESULTS: The interobserver analysis demonstrated strong-to-excellent agreement for NFV (ICC: 0.85 to 0.96) and moderate-to-excellent agreement for regurgitation fraction (ICC: 0.53 to 0.97) for all sites and valves. In addition, all observers established a low intervalvular variation (≤10.5%) in their analysis. The availability of 2 cine images per valve for valve tracking compared with 1 cine image predicted a decreasing variation in NFV among the 4 valves (beta = -1.3; p = 0.01). CONCLUSIONS: Independently of locally used CMR scanners and protocols, valvular flow quantification can be performed consistently with automated retrospective valve tracking in 4D flow CMR.

Achieving brain clearance and preventing neurodegenerative diseases-A glymphatic perspective.

Age-related neurodegenerative diseases are a growing burden to society, and many are sporadic, meaning that the environment, diet and lifestyle play significant roles. Cerebrospinal fluid (CSF)-mediated clearing of brain waste products via perivascular pathways, named the glymphatic system, is receiving increasing interest, as it offers unexplored perspectives on understanding neurodegenerative diseases. The glymphatic system is involved in clearance of metabolic by-products such as amyloid-ß from the brain, and its function is believed to lower the risk of developing some of the most common neurodegenerative diseases. Here, we present magnetic resonance imaging (MRI) data on the heart cycle's control of CSF flow in humans which corroborates findings from animal studies. We also review the importance of sleep, diet, vascular health for glymphatic clearance and find that these factors are also known players in brain longevity.

Doppler ultrasound cardiac gating of intracranial flow at 7T.

BACKGROUND: Ultra-high field magnetic resonance imaging (MR) may be used to improve intracranial blood flow measurements. However, standard cardiac synchronization methods tend to fail at ultra-high field MR. Therefore, this study aims to investigate an alternative synchronization technique using Doppler ultrasound. METHODS: Healthy subjects (n = 9) were examined with 7T MR. Flow was measured in the M1-branch of the middle cerebral artery (MCA) and in the cerebral aqueduct (CA) using through-plane phase contrast (2D flow). Flow in the circle of Willis was measured with three-dimensional, three-directional phase contrast (4D flow). Scans were gated with Doppler ultrasound (DUS) and electrocardiogram (ECG), and pulse oximetry data (POX) was collected simultaneously. False negative and false positive trigger events were counted for ECG, DUS and POX, and quantitative flow measures were compared. RESULTS: There were fewer false positive triggers for DUS compared to ECG (5.3 ± 11 vs. 25 ± 31, p = 0.031), while no other measured parameters differed significantly. Net blood flow in M1 was similar between DUS and ECG for 2D flow (1.5 ± 0.39 vs. 1.6 ± 0.41, bias ± 1.96SD: - 0.021 ± 0.36) and 4D flow (1.8 ± 0.48 vs. 9 ± 0.59, bias ± 1.96SD: - 0.086 ± 0.57 ml). Net CSF flow per heart beat in the CA was also similar for DUS and ECG (3.6 ± 2.1 vs. 3.0 ± 5.8, bias ± 1.96SD: 0.61 ± 13.6 µl). CONCLUSION: Gating with DUS produced fewer false trigger events than using ECG, with similar quantitative flow values. DUS gating is a promising technique for cardiac synchronization at 7T.

Blood flow imaging by optimal matching of computational fluid dynamics to 4D-flow data.

PURPOSE: Three-dimensional, time-resolved blood flow measurement (4D-flow) is a powerful research and clinical tool, but improved resolution and scan times are needed. Therefore, this study aims to (1) present a postprocessing framework for optimization-driven simulation-based flow imaging, called 4D-flow High-resolution Imaging with a priori Knowledge Incorporating the Navier-Stokes equations and the discontinuous Galerkin method (4D-flow HIKING), (2) investigate the framework in synthetic tests, (3) perform phantom validation using laser particle imaging velocimetry, and (4) demonstrate the use of the framework in vivo. METHODS: An optimizing computational fluid dynamics solver including adjoint-based optimization was developed to fit computational fluid dynamics solutions to 4D-flow data. Synthetic tests were performed in 2D, and phantom validation was performed with pulsatile flow. Reference velocity data were acquired using particle imaging velocimetry, and 4D-flow data were acquired at 1.5 T. In vivo testing was performed on intracranial arteries in a healthy volunteer at 7 T, with 2D flow as the reference. RESULTS: Synthetic tests showed low error (0.4%-0.7%). Phantom validation showed improved agreement with laser particle imaging velocimetry compared with input 4D-flow in the horizontal (mean -0.05 vs -1.11 cm/s, P < .001; SD 1.86 vs 4.26 cm/s, P < .001) and vertical directions (mean 0.05 vs -0.04 cm/s, P = .29; SD 1.36 vs 3.95 cm/s, P < .001). In vivo data show a reduction in flow rate error from 14% to 3.5%. CONCLUSIONS: Phantom and in vivo results from 4D-flow HIKING show promise for future applications with higher resolution, shorter scan times, and accurate quantification of physiological parameters.

Subacute vessel wall imaging at 7-T MRI in post-thrombectomy stroke patients.

PURPOSE: Reports from 3-T vessel wall MRI imaging have shown contrast enhancement following thrombectomy for acute stroke, suggesting potential intimal damage. Comparisons have shown higher SNR and more lesions detected by vessel wall imaging when using 7 T compared with 3 T. The aim of this study was to investigate the vessel walls after stent retriever thrombectomy using high-resolution vessel wall imaging at 7 T. METHODS: Seven patients with acute stroke caused by occlusion of the distal internal carotid artery (T-occlusion), or proximal medial cerebral artery, and treated by stent retriever thrombectomy with complete recanalization were included and examined by 7-T MRI within 2 days. The MRI protocol included a high-resolution black blood sequence with prospective motion correction (iMOCO), acquired before and after contrast injection. Flow measurements were performed in the treated and untreated M1 segments. RESULTS: All subjects completed the MRI examination. Image quality was independently rated as excellent by two neuroradiologists for all cases, and the level of motion artifacts did not impair diagnostic quality, despite severe motion in some cases. Contrast enhancement correlated with the deployment location of the stent retrievers. Flow data showed complete restoration of flow after treatment. CONCLUSION: Vessel wall imaging with prospective motion correction can be performed in patients following thrombectomy with excellent imaging quality at 7 T. We show that vessel wall contrast enhancement is the normal post-operative state and corresponds to the deployment location of the stent retriever.

A new vessel segmentation algorithm for robust blood flow quantification from two-dimensional phase-contrast magnetic resonance images.

Blood flow measurements in the ascending aorta and pulmonary artery from phase-contrast magnetic resonance images require accurate time-resolved vessel segmentation over the cardiac cycle. Current semi-automatic segmentation methods often involve time-consuming manual correction, relying on user experience for accurate results. The purpose of this study was to develop a semi-automatic vessel segmentation algorithm with shape constraints based on manual vessel delineations for robust segmentation of the ascending aorta and pulmonary artery, to evaluate the proposed method in healthy volunteers and patients with heart failure and congenital heart disease, to validate the method in a pulsatile flow phantom experiment, and to make the method freely available for research purposes. Algorithm shape constraints were extracted from manual reference delineations of the ascending aorta (n = 20) and pulmonary artery (n = 20) and were included in a semi-automatic segmentation method only requiring manual delineation in one image. Bias and variability (bias ± SD) for flow volume of the proposed algorithm versus manual reference delineations were 0·0 ± 1·9 ml in the ascending aorta (n = 151; seven healthy volunteers; 144 heart failure patients) and -1·7 ± 2·9 ml in the pulmonary artery (n = 40; 25 healthy volunteers; 15 patients with atrial septal defect). Interobserver bias and variability were lower (P = 0·008) for the proposed semi-automatic method (-0·1 ± 0·9 ml) compared to manual reference delineations (1·5 ± 5·1 ml). Phantom validation showed good agreement between the proposed method and timer-and-beaker flow volumes (0·4 ± 2·7 ml). In conclusion, the proposed semi-automatic vessel segmentation algorithm can be used for efficient analysis of flow and shunt volumes in the aorta and pulmonary artery.

Validation and reproducibility of cardiovascular 4D-flow MRI from two vendors using 2 × 2 parallel imaging acceleration in pulsatile flow phantom and in vivo with and without respiratory gating.

BACKGROUND: 4D-flow magnetic resonance imaging (MRI) is increasingly used. PURPOSE: To validate 4D-flow sequences in phantom and in vivo, comparing volume flow and kinetic energy (KE) head-to-head, with and without respiratory gating. MATERIAL AND METHODS: Achieva dStream (Philips Healthcare) and MAGNETOM Aera (Siemens Healthcare) 1.5-T scanners were used. Phantom validation measured pulsatile, three-dimensional flow with 4D-flow MRI and laser particle imaging velocimetry (PIV) as reference standard. Ten healthy participants underwent three cardiac MRI examinations each, consisting of cine-imaging, 2D-flow (aorta, pulmonary artery), and 2 × 2 accelerated 4D-flow with (Resp+) and without (Resp-) respiratory gating. Examinations were acquired consecutively on both scanners and one examination repeated within two weeks. Volume flow in the great vessels was compared between 2D- and 4D-flow. KE were calculated for all time phases and voxels in the left ventricle. RESULTS: Phantom results showed high accuracy and precision for both scanners. In vivo, higher accuracy and precision ( P < 0.001) was found for volume flow for the Aera prototype with Resp+ (-3.7 ± 10.4 mL, r = 0.89) compared to the Achieva product sequence (-17.8 ± 18.6 mL, r = 0.56). 4D-flow Resp- on Aera had somewhat larger bias (-9.3 ± 9.6 mL, r = 0.90) compared to Resp+ ( P = 0.005). KE measurements showed larger differences between scanners on the same day compared to the same scanner at different days. CONCLUSION: Sequence-specific in vivo validation of 4D-flow is needed before clinical use. 4D-flow with the Aera prototype sequence with a clinically acceptable acquisition time (<10 min) showed acceptable bias in healthy controls to be considered for clinical use. Intra-individual KE comparisons should use the same sequence.

Independent validation of metric optimized gating for fetal cardiovascular phase-contrast flow imaging.

PURPOSE: To validate metric optimized gating phase-contrast MR (MOG PC-MR) flow measurements for a range of fetal flow velocities in phantom experiments. 2) To investigate intra- and interobserver variability for fetal flow measurements at an imaging center other than the original site. METHODS: MOG PC-MR was compared to timer/beaker measurements in a pulsatile flow phantom using a heart rate (∼145 bpm), nozzle diameter (∼6 mm), and flow range (∼130-700 mL/min) similar to fetal imaging. Fifteen healthy fetuses were included for intra- and interobserver variability in the fetal descending aorta and umbilical vein. RESULTS: Phantom MOG PC-MR flow bias and variability was 2% ± 23%. Accuracy of MOG PC-MR was degraded for flow profiles with low velocity-to-noise ratio. Intra- and interobserver coefficients of variation were 6% and 19%, respectively, for fetal descending aorta; and 10% and 17%, respectively, for the umbilical vein. CONCLUSION: Phantom validation showed good agreement between MOG and conventionally gated PC-MR, except for cases with low velocity-to-noise ratio, which resulted in MOG misgating and underestimated peak velocities and warranted optimization of sequence parameters to individual fetal vessels. Inter- and intraobserver variability for fetal MOG PC-MR imaging were comparable to previously reported values.