Four-Dimensional Flow MRI of Abdominal Veins: A Systematic Review

The aim of this systematic review is to provide an overview of the use of Four-Dimensional Magnetic Resonance Imaging of vector blood flow (4D Flow MRI) in the abdominal veins. This study was composed according to the PRISMA guidelines 2009. The literature search was conducted in MEDLINE, Cochrane Library, EMBASE, and Web of Science. Quality assessment of the included studies was performed using the QUADAS-2 tool. The initial search yielded 781 studies and 21 studies were included. All studies successfully applied 4D Flow MRI in abdominal veins. Four-Dimensional Flow MRI was capable of discerning between healthy subjects and patients with cirrhosis and/or portal hypertension. The visual quality and inter-observer agreement of 4D Flow MRI were rated as excellent and good to excellent, respectively, and the studies utilized several different MRI data sampling strategies. By applying spiral sampling with compressed sensing to 4D Flow MRI, the blood flow of several abdominal veins could be imaged simultaneously in 18–25 s, without a significant loss of visual quality. Four-Dimensional Flow MRI might be a useful alternative to Doppler sonography for the diagnosis of cirrhosis and portal hypertension. Further clinical studies need to establish consensus regarding MRI sampling strategies in patients and healthy subjects.


Introduction
Phase contrast Magnetic Resonance Imaging (MRI) is widely used for the clinical evaluation of blood flow in the heart and in large vessels [1]. Developments of the technique have enabled the measurement of time-resolved vector blood flow in three anatomic dimensions, also known as 4D Flow MRI. Studies have shown that 4D Flow MRI can provide full, velocity encoded, volumetric coverage of the heart, aorta and thoracic arteries [2][3][4][5][6][7]. This allows for the acquisition of quantitative hemodynamic parameters such as blood flow velocity and blood flow volume, while simultaneously visualizing the direction and velocity of blood flow as velocity vectors, by using streamlines and particle tracing. To achieve this, 4D Flow MRI uses the phase contrast technique, in which bipolar magnetic field gradients create a phase shift of the MR signal proportional to the flow velocity. The sensitivity of phase contrast MRI to flow velocities is controlled by the velocity encoding (venc) parameter, which must be set to the expected maximum flow velocity. After the acquisition of phase contrast MRI, flow visualization software is used to create volumetric blood flow images. [8]. The ability to combine hemodynamic data with visual assessment in a single scan sequence may permit the exploration of diagnostic markers in any vascular region of interest. As with other MRI scanning techniques, 4D Flow MRI can use different sampling methods, which may affect the scan time and image quality [9][10][11]. The results from several studies indicate that the hemodynamic parameters acquired by 4D Flow MRI have the potential to reveal arterial pathologies such as aortic stenosis and chronic obstructive pulmonary disease [3,6,12].
The utility of 4D Flow MRI in the abdominal area and for the venous system is less explored, as compared to cardiovascular applications. Even so, some studies have investigated 4D Flow MRI as a diagnostic tool for pathologies in the abdominal venous vasculature, such as cirrhosis and portal hypertension [13,14].
These pathologies can already be examined non-invasively by Doppler sonography. Though Doppler has a short examination time for single blood vessels [15], 4D Flow MRI may acquire volumetric flow data in several blood vessels simultaneously, and thus 4D Flow MRI may have a greater clinical applicability for the evaluation of pathologies of abdominal veins.
To the knowledge of the authors, no systematic review of 4D Flow studies outside the cardiovascular system has been published. The purpose of this systematic review is to create an overview of the published literature evaluating the feasibility of 4D Flow MRI as a diagnostic tool in abdominal veins.

Search Strategy
The eligibility criteria and analysis in this review were performed according to the PRISMA guidelines 2009 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [16]. The literature search was conducted in the following databases: MEDLINE, Cochrane Library, EMBASE, and Web of Science. The intention of the search was to identify studies applying 4D Flow MRI in abdominal veins. Selection criteria for inclusion in this review were studies with human subjects, published in or after 2010, written in English, including implementation of 4D Flow MRI in the abdominal veins. The last search was performed on the 22 February 2021. Since the use of free text was different in the applied databases, it was necessary to tailor the search for each database. It should be noted that at the time of the search and of writing, 4D Flow MRI has not yet been made a MeSH term. Therefore, the search string required more free text to reduce the risk of missing relevant studies. The final search string can be found in Appendix A.

Study Selection
The included studies were filtered for duplicates by using Covidence, a web-based systematic review software. Two authors with 20 years and 1 year of experience in radiography (C.L. and S.H., respectively) reviewed the relevance of the studies, starting with the relevance of the title, then the relevance of the abstract and then the relevance of the full text. Disagreement on relevance and inclusion was resolved in consensus. As shown on the PRISMA flowchart of the inclusion process in Figure 1, the literature search yielded 781 publications. Of the yielded publications, 302 were duplicates and 455 were deemed irrelevant, on account of not applying 4D Flow MRI or not applying 4D Flow MRI in abdominal veins. After full-text assessment, three additional studies were deemed ineligible. Two did not include use of 4D Flow MRI, one had only 3 subjects.

Quality Assessment
Lastly, the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool was used to assess the risk of bias and applicability in the included studies [17]. Risks of bias and applicability were classified as high, low, or unclear by the same two authors who selected studies for inclusion.

Quality Assessment
Lastly, the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool was used to assess the risk of bias and applicability in the included studies [17]. Risks of bias and applicability were classified as high, low, or unclear by the same two authors who selected studies for inclusion.

Overview of Included Studies
A table (Table 1) of relevant data in each study was made using the following headings: author and publication year, aim of study, number of subjects, sampling methods, scan time, range of blood flow velocities for velocity encoding (venc), examined venous structures, and conclusion of study.

Overview of Included Studies
A table (Table 1) of relevant data in each study was made using the following headings: author and publication year, aim of study, number of subjects, sampling methods, scan time, range of blood flow velocities for velocity encoding (venc), examined venous structures, and conclusion of study.

Anatomic Coverage of Included Studies
All the included studies successfully applied 4D Flow MRI in abdominal veins. Vascular structures imaged in each study can be seen in Table 2.

Visual Quality, Inter-Observer Agreement, Sampling Method, and Scan Times
Nine out of 21 studies analyzed the visual quality of the acquired 4D Flow MRI scans [13,[18][19][20]23,24,27,29,30]. The nine studies reported a good to very good or excellent visual quality, though the left portal vein branch was reported to have a lower visual quality than other scanned veins in six of the seven studies that documented the visual quality in this vein [13,18,20,23,24,30]. Seven studies assessed inter-observer agreement. Five studies rated the inter-observer agreement as substantial, high or excellent [1,10,13,20,24,31]. One study rated it as good [18].
The studies that applied cartesian sampling had a scan time of 6 to 15 min [1,22,29]. Eight of the studies with applied radial sampling had a scan time of 10 to 12 min [14,21,[26][27][28]32], and one that applied time averaging had a scan time of down to three to four minutes [19]. The two studies with spiral sampling and compressed sensing applied had a scan time from 18 to 25 s [10,22].

Applicability and Risk of Bias
The QUADAS-2 assessment on risk of bias and concerns about applicability can be seen in Table 3 below. All studies were considered to have an overall low risk of bias, though several studies had an unclear risk in the index test and reference standard. Risk of bias in patient selection was considered high if the examined study did not make use of consecutively or randomly selected subjects, used a case-control design, or made inappropriate exclusions.
Risk of bias in the index test was considered high if the examined study interpreted the index test results with knowledge of the reference standard.
Risk of bias in the reference standard was considered high if the examined study interpreted the results of the reference standard with knowledge of the results the index test or used a reference standard that was unlikely to correctly classify the target condition. Risk of bias in flow and timing was considered high if the examined study had an inappropriate time interval between the index test and the reference standard, did not include all its subjects in the analysis or did not use the same reference standard for all its subjects. Studies considered to have an unclear risk of bias in the index test and reference standard did not state whether they interpreted their results without knowledge of the reference standard and did not state whether they interpreted the results of the reference standard without knowledge of the index test. Studies with an unclear risk of bias in flow and timing did not state the time interval of their data acquisition.

Discussion
In summary, the results from the included studies show that 4D Flow MRI of the abdominal veins is feasible, and that hemodynamic parameters based on 4D Flow MRI can be used to discern between healthy subjects and patients with cirrhosis and/or portal hypertension. The visual quality of the acquired flow images was rated good to very good and the included studies indicate that 4D Flow MRI has a low inter-observer variability.
The studies included in this systematic review were considered to have low bias. The acquisition time of spiral sampling combined with compressed sensing was remarkably faster (18 to 25 s) than the cartesian and radial sampling methods (6 to 15 min and 3 to 12 min, respectively) [1,10,14,19,21,22,[26][27][28][29]32]. The acceleration method known as k-t GRAPPA was successfully applied in four studies, though the acquisition times of these studies were still slower than the studies that applied spiral sampling and compressed sensing [10,18,22,23,29,31].
There is no significant loss of visual quality when applying spiral sampling to 4D Flow MRI in the aorta, despite the major reduction in scan time it provides, according to the included studies [10,22]. Compressed sensing has been stated to have no significant effect on visual quality when not overused; however, it does increase computational complexity, thereby potentially increasing the reconstruction time of the acquired image data [35]. The two included studies which combined spiral sampling and compressed sensing demonstrated that 4D Flow MRI with spiral sampling and compressed sensing can be applied in the abdominal veins, with greatly reduced scan time compared to radial sampling with or without k-t GRAPPA and cartesian sampling, and still maintain a good vascular conspicuity and strong agreement of the acquired quantitative parameters with those from established techniques [10,22].
Doppler sonography's high inter-observer variability and limited ability to visualize complex and variable anatomy in the abdominal vasculature compared to 4D Flow MRI were demonstrated by several studies included in this systematic review [10,13,18,20,23,24,27,[29][30][31]. This assessment seems to agree with the established literature. According to the consensus statement of P. Dyverfeldt et al. [36], volumetric flow imaging of 4D Flow MRI may provide a more accurate flow quantification in the presence of complex vessel geometry. In addition to this, included studies showed that the ability of 4D Flow MRI to assess volumetric flow in multiple veins simultaneously could potentially provide a better overview of blood flow parameters in the abdominal veins compared to Doppler sonography, in which each vein would have to be examined individually. Therefore, 4D Flow MRI may have potential as an alternative to Doppler sonography for the non-invasive measurements of hemodynamic parameters in the abdominal venous vasculature.
While the included studies indicate that 4D Flow MRI with spiral sampling and compressed sensing had a higher inter-observer agreement than Doppler sonography, one should also consider the comparatively high cost that is associated with MRI [37,38]. In addition to this, while 4D Flow MRI with spiral sampling and compressed sensing has a remarkably short acquisition time, Doppler sonography does not require the same amount of time for patient preparation and MRI safety as 4D Flow MRI. This means that the full examination time of 4D Flow MRI with spiral sampling and compressed sensing may still be notably higher than the examination time of Doppler sonography.
This study had limitations. Since 4D Flow MRI did not have a MeSH term at the time of the literature search, and the scan technique has several different proposed names, it is possible that some studies on 4D Flow MRI were not found during the literature search. The evaluation on how the choice of sampling method affects the scan time and visual quality of 4D Flow MRI was limited due to nine studies not stating their applied sampling method [13,18,20,23,24,30,31,33,34]. The limited number of participants (from 3 to 61) and the different patient groups were heterogenous and, therefore, it was not possible to make a collective conclusion of the expected hemodynamic parameters derived from 4D Flow MRI. For example, the 11 studies that investigated 4D Flow MRI's application in patients with cirrhosis had variations in patient age and severity of cirrhosis [1,10,13,18,20,25,27,28,30,31,33]. This also limited the possibility of making an estimate of the recommended venc value for future studies.
Repetition of the included studies with larger subject/patient groups is recommended to assess the possibility of hemodynamic data acquired by 4D Flow MRI as a diagnostic marker for cirrhosis or portal hypertension. In addition, this systematic review suggests that future research on 4D Flow MRI should use spiral sampling with compressed sensing, as it appears to be the most clinically viable path for 4D Flow MRI, and should further assess the reliability of this sampling method.

Conclusions
In conclusion, 4D Flow MRI examination of abdominal veins for the purpose of visual assessment and quantification of hemodynamic parameters is feasible. The hemodynamic parameters derived from 4D Flow MRI can be used to discern between healthy subjects and patients with cirrhosis and/or portal hypertension. Four-Dimensional Flow MRI of the abdominal veins has a higher inter-observer agreement than Doppler sonography, which is a currently used non-invasive method, and permits the acquisition of flow data in several vessels simultaneously. Recent developments of MRI sampling methods have allowed 4D Flow MRI scans of the abdominal veins to be acquired in approximately 20 s, while still maintaining good vessel conspicuity and reliability of hemodynamic data, thus potentially greatly facilitating routine clinical use.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.

Conflicts of Interest:
The authors declare no conflict of interest.