Noninvasive quantification of cerebrovascular pressure changes using 4D Flow MRI

Automated intracranial vessel segmentation of 4D flow MRI data in patients with atherosclerotic stenosis using a convolutional neural network

Introduction

Intracranial 4D flow MRI enables quantitative assessment of hemodynamics in patients with intracranial atherosclerotic disease (ICAD). However, quantitative assessments are still challenging due to the time-consuming vessel segmentation, especially in the presence of stenoses, which can often result in user variability. To improve the reproducibility and robustness as well as to accelerate data analysis, we developed an accurate, fully automated segmentation for stenosed intracranial vessels using deep learning.

Methods

154 dual-VENC 4D flow MRI scans (68 ICAD patients with stenosis, 86 healthy controls) were retrospectively selected. Manual segmentations were used as ground truth for training. For automated segmentation, deep learning was performed using a 3D U-Net. 20 randomly selected cases (10 controls, 10 patients) were separated and solely used for testing. Cross-sectional areas and flow parameters were determined in the Circle of Willis (CoW) and the sinuses. Furthermore, the flow conservation error was calculated. For statistical comparisons, Dice scores (DS), Hausdorff distance (HD), average symmetrical surface distance (ASSD), Bland-Altman analyses, and interclass correlations were computed using the manual segmentations from two independent observers as reference. Finally, three stenosis cases were analyzed in more detail by comparing the 4D flow-based segmentations with segmentations from black blood vessel wall imaging (VWI).

Results

Training of the network took approximately 10 h and the average automated segmentation time was 2.2 ± 1.0 s. No significant differences in segmentation performance relative to two independent observers were observed. For the controls, mean DS was 0.85 ± 0.03 for the CoW and 0.86 ± 0.06 for the sinuses. Mean HD was 7.2 ± 1.5 mm (CoW) and 6.6 ± 3.7 mm (sinuses). Mean ASSD was 0.15 ± 0.04 mm (CoW) and 0.22 ± 0.17 mm (sinuses). For the patients, the mean DS was 0.85 ± 0.04 (CoW) and 0.82 ± 0.07 (sinuses), the HD was 8.4 ± 3.1 mm (CoW) and 5.7 ± 1.9 mm (sinuses) and the mean ASSD was 0.22 ± 0.10 mm (CoW) and 0.22 ± 0.11 mm (sinuses). Small bias and limits of agreement were observed in both cohorts for the flow parameters. The assessment of the cross-sectional lumen areas in stenosed vessels revealed very good agreement (ICC: 0.93) with the VWI segmentation but a consistent overestimation (bias ± LOA: 28.1 ± 13.9%).

Discussion

Deep learning was successfully applied for fully automated segmentation of stenosed intracranial vasculatures using 4D flow MRI data. The statistical analysis of segmentation and flow metrics demonstrated very good agreement between the CNN and manual segmentation and good performance in stenosed vessels. To further improve the performance and generalization, more ICAD segmentations as well as other intracranial vascular pathologies will be considered in the future.

Differentiation Between the Low and High Trans-Stenotic Pressure Gradient in Patients With Idiopathic Intracranial Hypertension Using 4D Flow MRI-Derived Hemodynamic Parameters

BACKGROUND

Trans-stenotic pressure gradient (TPG) measurement is essential for idiopathic intracranial hypertension (IIH) patients with transverse sinus (TS) stenosis. Four-D flow MRI may provide a noninvasive imaging method for differentiation of IIH patients with different TPG.

PURPOSE

To investigate the associations between 4D flow parameters and TPG, and to evaluate the diagnostic performance of 4D flow parameters in differentiating patients with high TPG (GroupHP) from low TPG (GroupLP).

STUDY TYPE

Prospective.

POPULATION

31 IIH patients with TS stenosis (age, 38 ± 12 years; 23 females) and 5 healthy volunteers (age, 25 ± 1 years; 2 females).

FIELD STRENGTH/SEQUENCE

3T, 3D phase contrast MR venography, and gradient recalled echo 4D flow sequences.

ASSESSMENT

Scan-rescan reproducibility of 4D flow parameters were performed. The correlation between TPG and flow parameters was analyzed. The netflow and velocity difference between inflow plane, outflow plane, and the stenosis plane were calculated and compared between GroupHP and GroupLP.

STATISTICAL TESTS

Pearson's correlation or Spearman's rank correlation coefficient, Independent samples t-test or Wilcoxon rank-sum test, Intra-class correlation coefficient (ICC), Bland-Altman analyses, Receiver operating characteristic curves. A P value <0.05 was considered significant.

RESULTS

Significant correlations were found between TPG and netflow parameters including Favg,out-s, Favg,in-s, Fmax,out-s, and Fmax,in-s (r = 0.525-0.565). Significant differences were found in Favg,out-s, Fmax,out-s, Favg,in-s, and Fmax,in-s between GroupHP and GroupLP. Using the cut-off value of 2.19 mL/sec, the Favg,out-s showed good estimate performance in distinguishing GroupHP from GroupLP (AUC = 0.856). The ICC (ranged 0.905-0.948) and Bland-Altman plots indicated good scan-rescan reproducibility.

DATA CONCLUSIONS

4D flow MRI derived flow parameters showed good correlations with TPG in IIH patients with TS stenosis. Netflow difference between outflow and stenosis location at TS shows the good performance in differentiating GroupHP and GroupLP cases.

LEVEL OF EVIDENCE

2 TECHNICAL EFFICACY: Stage 2.

Evaluating the accuracy of cerebrovascular computational fluid dynamics modeling through time-resolved experimental validation

Accurate modeling of cerebral hemodynamics is crucial for better understanding the hemodynamics of stroke, for which computational fluid dynamics (CFD) modeling is a viable tool to obtain information. However, a comprehensive study on the accuracy of cerebrovascular CFD models including both transient arterial pressures and flows does not exist. This study systematically assessed the accuracy of different outlet boundary conditions (BCs) comparing CFD modeling and an in-vitro experiment. The experimental setup consisted of an anatomical cerebrovascular phantom and high-resolution flow and pressure data acquisition. The CFD model of the same cerebrovascular geometry comprised five sets of stationary and transient BCs including established techniques and a novel BC, the phase modulation approach. The experiment produced physiological hemodynamics consistent with reported clinical results for total cerebral blood flow, inlet pressure, flow distribution, and flow pulsatility indices (PI). The in-silico model instead yielded time-dependent deviations between 19-66% for flows and 6-26% for pressures. For cerebrovascular CFD modeling, it is recommended to avoid stationary outlet pressure BCs, which caused the highest deviations. The Windkessel and the phase modulation BCs provided realistic flow PI values and cerebrovascular pressures, respectively. However, this study shows that the accuracy of current cerebrovascular CFD models is limited.

Coupling synthetic and real-world data for a deep learning-based segmentation process of 4D flow MRI

BACKGROUND AND OBJECTIVE

Phase contrast magnetic resonance imaging (4D flow MRI) is an imaging technique able to provide blood velocity in vivo and morphological information. This capability has been used to study mainly the hemodynamics of large vessels, such as the thoracic aorta. However, the segmentation of 4D flow MRI data is a complex and time-consuming task. In recent years, neural networks have shown great accuracy in segmentation tasks if large datasets are provided. Unfortunately, in the context of 4D flow MRI, the availability of these data is limited due to its recent adoption in clinical settings. In this study, we propose a pipeline for generating synthetic thoracic aorta phase contrast magnetic resonance angiography (PCMRA) to expand the limited dataset of patient-specific PCMRA images, ultimately improving the accuracy of the neural network segmentation even with a small real dataset.

METHODS

The pipeline involves several steps. First, a statistical shape model is used to synthesize new artificial geometries to improve data numerosity and variability. Secondly, computational fluid dynamics simulations are employed to simulate the velocity fields and, finally, after a downsampling and a signal-to-noise and velocity limit adjustment in both frequency and spatial domains, volumes are obtained using the PCMRA formula. These synthesized volumes are used in combination with real-world data to train a 3D U-Net neural network. Different settings of real and synthetic data are tested.

RESULTS

Incorporating synthetic data into the training set significantly improved the segmentation performance compared to using only real data. The experiments with synthetic data achieved a DICE score (DS) value of 0.83 and a better target reconstruction with respect to the case with only real data (DS = 0.65).

CONCLUSION

The proposed pipeline demonstrated the ability to increase the dataset in terms of numerosity and variability and to improve the segmentation accuracy for the thoracic aorta using PCMRA.

Super-resolution 4D flow MRI to quantify aortic regurgitation using computational fluid dynamics and deep learning

Changes in cardiovascular hemodynamics are closely related to the development of aortic regurgitation (AR), a type of valvular heart disease. Metrics derived from blood flows are used to indicate AR onset and evaluate its severity. These metrics can be non-invasively obtained using four-dimensional (4D) flow magnetic resonance imaging (MRI), where accuracy is primarily dependent on spatial resolution. However, insufficient resolution often results from limitations in 4D flow MRI and complex aortic regurgitation hemodynamics. To address this, computational fluid dynamics simulations were transformed into synthetic 4D flow MRI data and used to train a variety of neural networks. These networks generated super-resolution, full-field phase images with an upsample factor of 4. Results showed decreased velocity error, high structural similarity scores, and improved learning capabilities from previous work. Further validation was performed on two sets of in vivo 4D flow MRI data and demonstrated success in de-noising flow images. This approach presents an opportunity to comprehensively analyse AR hemodynamics in a non-invasive manner.

Cerebrovascular super-resolution 4D Flow MRI - Sequential combination of resolution enhancement by deep learning and physics-informed image processing to non-invasively quantify intracranial velocity, flow, and relative pressure

The development of cerebrovascular disease is tightly coupled to regional changes in intracranial flow and relative pressure. Image-based assessment using phase contrast magnetic resonance imaging has particular promise for non-invasive full-field mapping of cerebrovascular hemodynamics. However, estimations are complicated by the narrow and tortuous intracranial vasculature, with accurate image-based quantification directly dependent on sufficient spatial resolution. Further, extended scan times are required for high-resolution acquisitions, and most clinical acquisitions are performed at comparably low resolution (>1 mm) where biases have been observed with regard to the quantification of both flow and relative pressure. The aim of our study was to develop an approach for quantitative intracranial super-resolution 4D Flow MRI, with effective resolution enhancement achieved by a dedicated deep residual network, and with accurate quantification of functional relative pressures achieved by subsequent physics-informed image processing. To achieve this, our two-step approach was trained and validated in a patient-specific in-silico cohort, showing good accuracy in estimating velocity (relative error: 15.0 ± 0.1%, mean absolute error (MAE): 0.07 ± 0.06 m/s, and cosine similarity: 0.99 ± 0.06 at peak velocity) and flow (relative error: 6.6 ± 4.7%, root mean square error (RMSE): 0.56 mL/s at peak flow), and with the coupled physics-informed image analysis allowing for maintained recovery of functional relative pressure throughout the circle of Willis (relative error: 11.0 ± 7.3%, RMSE: 0.3 ± 0.2 mmHg). Furthermore, the quantitative super-resolution approach is applied to an in-vivo volunteer cohort, effectively generating intracranial flow images at <0.5 mm resolution and showing reduced low-resolution bias in relative pressure estimation. Our work thus presents a promising two-step approach to non-invasively quantify cerebrovascular hemodynamics, being applicable to dedicated clinical cohorts in the future.

Relative pressure estimation from 4D flow MRI using generalized Bernoulli equation in a phantom model of arterial stenosis

OBJECTIVE

Arterial stenosis is a significant cardiovascular disease requiring accurate estimation of the pressure gradients for determining hemodynamic significance. In this paper, we propose Generalized Bernoulli Equation (GBE) utilizing interpolated-based method to estimate relative pressures using streamlines and pathlines from 4D Flow MRI.

METHODS

4D Flow MRI data in a stenotic phantom model and computational fluid dynamics simulated velocities generated under identical flow conditions were processed by Generalized Bernoulli Equation (GBE), Reduced Bernoulli Equations (RBE), as well as the Simple Bernoulli Equation (SBE) which is clinically prevalent. Pressures derived from 4D flow MRI and noise corrupted CFD velocities were compared with pressures generated directly with CFD as well as pressures obtained using Millar catheters under identical flow conditions.

RESULTS

It was found that SBE and RBE methods underestimated the relative pressure for lower flow rates while overestimating the relative pressure at higher flow rates. Specifically, compared to the reference pressure, SBE underestimated the maximum relative pressure by 22[Formula: see text] for a pulsatile flow data with peak flow rate [Formula: see text] and overestimated by around 40[Formula: see text] when [Formula: see text]. In contrast, for GBE method the relative pressure values were overestimated by 15[Formula: see text] with [Formula: see text]and around 10[Formula: see text] with [Formula: see text].

CONCLUSION

GBE methods showed robust performance to additive image noise compared to other methods. Our findings indicate that GBE pressure estimation over pathlines attains the highest level of accuracy compared to GBE over streamlines, and the SBE and RBE methods.

Inverse problems in blood flow modeling: A review

Mathematical and computational modeling of the cardiovascular system is increasingly providing non-invasive alternatives to traditional invasive clinical procedures. Moreover, it has the potential for generating additional diagnostic markers. In blood flow computations, the personalization of spatially distributed (i.e., 3D) models is a key step which relies on the formulation and numerical solution of inverse problems using clinical data, typically medical images for measuring both anatomy and function of the vasculature. In the last years, the development and application of inverse methods has rapidly expanded most likely due to the increased availability of data in clinical centers and the growing interest of modelers and clinicians in collaborating. Therefore, this work aims to provide a wide and comparative overview of literature within the last decade. We review the current state of the art of inverse problems in blood flows, focusing on studies considering fully dimensional fluid and fluid-solid models. The relevant physical models and hemodynamic measurement techniques are introduced, followed by a survey of mathematical data assimilation approaches used to solve different kinds of inverse problems, namely state and parameter estimation. An exhaustive discussion of the literature of the last decade is presented, structured by types of problems, models and available data.

Altered Aortic Hemodynamics and Relative Pressure in Patients with Dilated Cardiomyopathy

Ventricular-vascular interaction is central in the adaptation to cardiovascular disease. However, cardiomyopathy patients are predominantly monitored using cardiac biomarkers. The aim of this study is therefore to explore aortic function in dilated cardiomyopathy (DCM). Fourteen idiopathic DCM patients and 16 controls underwent cardiac magnetic resonance imaging, with aortic relative pressure derived using physics-based image processing and a virtual cohort utilized to assess the impact of cardiovascular properties on aortic behaviour. Subjects with reduced left ventricular systolic function had significantly reduced aortic relative pressure, increased aortic stiffness, and significantly delayed time-to-pressure peak duration. From the virtual cohort, aortic stiffness and aortic volumetric size were identified as key determinants of aortic relative pressure. As such, this study shows how advanced flow imaging and aortic hemodynamic evaluation could provide novel insights into the manifestation of DCM, with signs of both altered aortic structure and function derived in DCM using our proposed imaging protocol.

Accelerated dual-venc 4D flow MRI with variable high-venc spatial resolution for neurovascular applications

Purpose

Dual‐velocity encoded (dual‐venc or DV) 4D flow MRI achieves wide velocity dynamic range and velocity‐to‐noise ratio (VNR), enabling accurate neurovascular flow characterization. To reduce scan time, we present interleaved dual‐venc 4D Flow with independently prescribed, prospectively undersampled spatial resolution of the high‐venc (HV) acquisition: Variable Spatial Resolution Dual Venc (VSRDV).

Methods

A prototype VSRDV sequence was developed based on a Cartesian acquisition with eight‐point phase encoding, combining PEAK‐GRAPPA acceleration with zero‐filling in phase and partition directions for HV. The VSRDV approach was optimized by varying z, the zero‐filling fraction of HV relative to low‐venc, between 0%–80% in vitro (realistic neurovascular model with pulsatile flow) and in vivo (n = 10 volunteers). Antialiasing precision, mean and peak velocity quantification accuracy, and test–retest reproducibility were assessed relative to reference images with equal‐resolution HV and low venc (z = 0%).

Results

In vitro results for all z demonstrated an antialiasing true positive rate at least 95% for RPEAK−GRAPPA = 2 and 5, with no linear relationship to z (p = 0.62 and 0.13, respectively). Bland–Altman analysis for z = 20%, 40%, 60%, or 80% versus z = 0% in vitro and in vivo demonstrated no bias >1% of venc in mean or peak velocity values at any RZF. In vitro mean and peak velocity, and in vivo peak velocity, had limits of agreement within 15%.

Conclusion

VSRDV allows up to 34.8% scan time reduction compared to PEAK‐GRAPPA accelerated DV 4D Flow MRI, enabling large spatial coverage and dynamic range while maintaining VNR and velocity measurement accuracy.

Click here for author‐reader discussions

On the numerical treatment of viscous and convective effects in relative pressure reconstruction methods

The mechanism of many cardiovascular diseases can be understood by studying the pressure distribution in blood vessels. Direct pressure measurements, however, require invasive probing and provide only single-point data. Alternatively, relative pressure fields can be reconstructed from imaging-based velocity measurements by considering viscous and inertial forces. Both contributions can be potential troublemakers in pressure reconstruction: the former due to its higher-order derivatives, and the latter because of the quadratic nonlinearity in the convective acceleration. Viscous and convective terms can be treated in various forms, which, although equivalent for ideal measurements, can perform differently in practice. In fact, multiple versions are often used in literature, with no apparent consensus on the more suitable variants. In this context, the present work investigates the performance of different versions of relative pressure estimators. For viscous effects, in particular, two new modified estimators are presented to circumvent second-order differentiation without requiring surface integrals. In-silico and in-vitro data in the typical regime of cerebrovascular flows are considered, allowing a systematic noise sensitivity study. Convective terms are shown to be the main source of error, even for flows with pronounced viscous component. Moreover, the conservation (often integrated) form of convection exhibits higher noise sensitivity than the standard convective description, in all three families of estimators considered here. For the classical pressure Poisson estimator, the present modified version of the viscous term tends to yield better accuracy than the (recently introduced) integrated form, although this effect is in most cases negligible when compared to convection-related errors.

Influence of Spatial Resolution and Compressed SENSE Acceleration Factor on Flow Quantification with 4D Flow MRI at 3 Tesla

Four-dimensional (4D) flow MRI allows quantifying flow in blood vessels–non invasively and in vivo. The clinical use of 4D flow MRI in small vessels, however, is hampered by long examination times and limited spatial resolution. Compressed SENSE (CS-SENSE) is a technique that can accelerate 4D flow dramatically. Here, we investigated the effect of spatial resolution and CS acceleration on flow measurements by using 4D flow MRI in small vessels in vitro at 3 T. We compared the flow in silicon tubes (inner diameters of 2, 3, 4, and 5 mm) measured with 4D flow MRI, accelerated with four CS factors (CS = 2.5, 4.5, 6.5, and 13) and three voxel sizes (0.5, 1, and 1.5 mm³) to 2D flow MRI and a flow sensor. Additionally, the velocity field in an aneurysm model acquired with 4D flow MRI was compared to the one simulated with computational fluid dynamics (CFD). A strong correlation was observed between flow sensor, 2D flow MRI, and 4D flow MRI (rho > 0.94). The use of fewer than seven voxels per vessel diameter (nROI) resulted in an overestimation of flow in more than 5% of flow measured with 2D flow MRI. A negative correlation (rho = −0.81) between flow error and nROI were found for CS = 2.5 and 4.5. No statistically significant impact of CS factor on differences in flow rates was observed. However, a trend of increased flow error with increased CS factor was observed. In an aneurysm model, the peak velocity and stagnation zone were detected by CFD and all 4D flow MRI variants. The velocity difference error in the aneurysm sac did not exceed 11% for CS = 4.5 in comparison to CS = 2.5 for all spatial resolutions. Therefore, CS factors from 2.5–4.5 can appear suitable to improve spatial or temporal resolution for accurate quantification of flow rate and velocity. We encourage reporting the number of voxels per vessel diameter to standardize 4D flow MRI protocols.