Rofo 2014; 186(9): 847-859
DOI: 10.1055/s-0034-1366661
Review
© Georg Thieme Verlag KG Stuttgart · New York

High temporal and high spatial resolution MR angiography (4D-MRA)

Zeitlich und räumlich hochaufgelöste MR Angiografie („4D-MRA“)
D. R. Hadizadeh
,
C. Marx
,
J. Gieseke
,
H. H. Schild
,
W. A. Willinek
Further Information

Correspondence

Dr. Dariusch Reza Hadizadeh
Radiology, University of Bonn
Sigmund-Freud-Str. 25
53113 Bonn
Germany   
Phone: ++ 49/2 28/28 71 58 75   
Fax: ++ 49/2 28/28 71 60 93   

Publication History

06 January 2014

08 May 2014

Publication Date:
23 June 2014 (online)

 

Abstract

In the first decade of the twenty-first century, whole-body magnetic resonance scanners with high field strengths (and thus potentially better signal-to-noise ratios) were developed. At the same time, parallel imaging and “echo-sharing” techniques were refined to allow for increasingly high spatial and temporal resolution in dynamic magnetic resonance angiography (“time-resolved” = TR-MRA). This technological progress facilitated tracking the passage of intra-venously administered contrast agent boluses as well as the acquisition of volume data sets at high image refresh rates (“4D-MRA”). This opened doors for many new applications in non-invasive vascular imaging, including simultaneous anatomic and functional analysis of many vascular pathologies including arterio-venous malformations. Different methods were established to acquire 4D-MRA using various strategies to acquire k-space trajectories over time in order to optimize imaging according to clinical needs. These include “keyhole”-based techniques (e. g. 4D-TRAK), TRICKS – both with and without projection – and HYPR-reconstruction, TREAT, and TWIST. Some of these techniques were first introduced in the 1980 s and 1990 s, were later enhanced and modified, and finally implemented in the products of major vendors. In the last decade, a large number of studies on the clinical applications of TR-MRA was published. This manuscript provides an overview of the development of TR-MRA methods and the 4D-MRA techniques as they are currently used in the diagnosis, treatment and follow-up of vascular diseases in various parts of the body.

Key statements

• 4D-MRA, which differs according to manufacturer, generates high temporal and spatial resolution MRA volume data sets.

Key differences in 4D-MRA techniques concern the sequence of the acquisition of k-space portions.

• Central k-space portions define image contrast and are thus repetitively scanned with 4D-MRA.

• Numerous clinical applications of 4D-MRA are already documented in the literature.

Citation Format:

• Hadizadeh DR., Marx C, Gieseke J et al. High temporal and high spatial resolution MR angiography (4D-MRA). Fortschr Röntgenstr 2014; 186: 847 – 859


#

Zusammenfassung

Im ersten Jahrzehnt des 21. Jahrhunderts wurden Ganzkörper-Magnetresonanztomografen mit höheren Feldstärken (und damit potenziell besserem Signal-zu-Rausch-Verhältnis) entwickelt. Dies und die nahezu zeitgleiche Entwicklung bzw. Verfeinerung von Techniken wie der parallelen Bildgebung und „Echo-sharing“ erlaubten ein zunehmend höheres räumliches und zeitliches Auflösungsvermögen in der dynamischen Magnetresonanzangiografie („Time-resolved“ = TR-MRA). Somit konnten erstmals sowohl die Passage eines Kontrastmittelbolus mit einer angemessenen räumlichen Auflösung verfolgt als auch Volumendatensätze mit hohen Bildauffrischungsraten erzeugt werden („4D-MRA“). Damit eröffneten sich neue Optionen der nicht invasiven Gefäßdiagnostik, wie beispielsweise die gleichzeitige anatomische und funktionelle Analyse von Gefäßmalformationen. Es wurden zahlreiche unterschiedliche Methoden eingeführt, bei denen die Akquisitionen der k-Raum-Trajektorien strategisch unterschiedlich über den Akquisitionszeitraum verteilt wurden, um optimale klinische Ergebnisse zu erlangen. Dazu zählten beispielsweise Verfahren wie „Keyhole“-basierte Techniken (z. B. 4D-TRAK), TRICKS mit und ohne Projektions- und HYPR-Rekonstruktion, TREAT und TWIST. Einige dieser Techniken wurden bereits in den 80er- und 90er-Jahren vorgestellt, in der Folge weiterentwickelt und modifiziert und schließlich in handelsübliche MR-Tomografen der verschiedenen Hersteller implementiert. In der letzten Dekade wurden schließlich zahlreiche Studien zu klinischen Anwendungen der TR-MRA vorgestellt. Dieses Manuskript bietet eine Übersicht über die Entwicklung der TR-MRA-Verfahren und die gegenwärtig eingesetzten 4D-MRA-Techniken wie sie derzeit Ihren klinischen Einsatz bei der Diagnose, Behandlung und bei Verlaufsuntersuchungen von Gefäßerkrankungen in unterschiedlichen Körperregionen finden.

Es wurden zahlreiche unterschiedliche Methoden eingeführt, bei denen die Akquisitionen der k-Raum-Trajektorien strategisch unterschiedlich über den Akquisitionszeitraum verteilt wurden, um optimale klinische Ergebnisse zu erlangen. Dazu zählten beispielsweise Verfahren wie „Keyhole“-basierte Techniken (z. B. 4D-TRAK), TRICKS mit und ohne Projektions- und HYPR-Rekonstruktion, TREAT und TWIST. Einige dieser Techniken wurden bereits in den 80er und 90er Jahren vorgestellt, in der Folge weiterentwickelt und modifiziert und schließlich in handelsübliche MR-Tomografen der verschiedenen Hersteller implementiert. In der letzten Dekade wurden schließlich zahlreiche Studien zu klinischen Anwendungen der TR-MRA vorgestellt. Dieses Manuskript bietet eine Übersicht über die Entwicklung der TR-MRA-Verfahren und die gegenwärtig eingesetzten 4D-MRA Techniken wie sie derzeit Ihren klinischen Einsatz bei der Diagnose, Behandlung und bei Verlaufsuntersuchungen von Gefäßerkrankungen in unterschiedlichen Körperregionen finden.


#

Introduction

MR angiography strategies have undergone constant refinement since the introduction of contrast-enhanced MR angiography by Prince et al. in 1993 and dynamic imaging using the “keyhole” method by van Vaals et al. and Jones et al. that same year. In particular, parallel imaging had a large impact on increasing the temporal efficiency of data collection sufficiently dynamically showing the contrast medium bolus passage in the blood vessel with high spatial resolution [1] [2] [3] [4] [5]. For the simultaneous imaging of flow dynamics and vascular anatomy, the invasive method of catheter angiography in the form of digital subtraction angiography (DSA) is the reference standard against which all new methods have to compete.

Dynamic information about blood flow in vessels is essential for functional analysis of arterial inflow and venous outflow of vascular malformations, for example. However, TR-MRA can also be helpful when it is necessary to image multiple vascular territories simultaneously while administering low doses of contrast medium, e. g. in children and patients who can tolerate only short examination times [6] [7]. 4D-MRA is generally useful for every clinical problem involving a short arteriovenous transit time, for example in the pulmonary artery flow path, where it has ben used to investigate arteriovenous fistulae and shunts [8] [9] [10] [11] [12] [13]. This review provides an overview of the temporal resolution MR angiography techniques currently used routinely and their development, while presenting various clinical applications in which these techniques have already been employed successfully. Diagnosing cerebral arteriovenous malformations (cAVM) is an example of a key clinical application for comparing individual methods, since these malformations can be examined using all of the techniques described in the article, and data from prospective studies have already been published and in turn cited in this article.


#

The development of temporal resolution MR angiography techniques

The first temporal resolution MR sequences were developed to roughly image the bolus passage of contrast agents and thereby ascertain bolus arrival time. For this purpose, T1-weighted gradient echo techniques single-thick slice-(known as 2D-MRA) were created. At that time, however, significant compromises had to be made in terms of spatial resolution to facilitate achieving the requisite very short time intervals of 1 – 2 seconds [14] [15] [16].

This idea of dynamically tracking a contrast medium bolus (integrated with a mask subtraction of non-contrast images) is indispensable for planning and facilitating precise temporal initiation of static sequences such as high spatial resolution volume data sets (3D-CEMRA), which absolutely require precise “timing”, i. e. a precise starting time (fluoroscopic triggering).

T1-weighted 2D-multi-slice and 3D-gradient echo sequences are generally suited for showing the arrival of a contrast medium bolus with a high signal-to-noise ratio. However, to achieve a high image refresh rate in this process, it is necessary to limit the k-space portions that are actually scanned. First, image acquisition was accelerated using above all the symmetry characteristics of the k-space. Major advancements in the acceleration of dynamic sequences were then achieved through the introduction of parallel imaging [17] [18] [19], the use of higher field strengths with a more favorable signal-to-noise ratio and the use of complex, strategic k-space acquisition schemes. One such scheme is “echo sharing”, which is the practice of strategically distributing k-space acquisition over the sequence duration and using a certain temporal interpolation to reconstruct k-space. What is critical here is that the portion of k-space (central portion) that essentially determines image contrast is scanned at a higher frequency than k-space portions (peripheral portions) that are less significant to image contrast. This allows contrast change to be mapped over time even if only a small portion of k-space is actually acquired per dynamic phase [20] [21]. Through the use of higher field strengths and improved coils [11] [22] [23] [24] sequences were established in the last decade that simultaneously facilitate both high temporal and spatial resolution in dynamic vascular imaging. The result was a steady increase in publications on the field of 4D-MRA which continues to this day ([Fig. 1a]). Because the development of these complex techniques required close cooperation between clinical institutions and commercial partners, the various “echo-sharing” techniques were used mostly in a manufacturer-specific manner ([Table 1]). These techniques are listed in this review according to method.

Zoom Image
Fig. 1 Original publications on time-resolved MRA between 2000 and 2012; a Number of publications related to field strength; b Publications broken down by TR-MRA technique.
Table 1

Time-resolved contrast-enhanced TR-MRA techniques.[1]

TR-MRA technique

year

field strength

manufacturer

keyhole

1993

1.5 T

Philips Healthcare, Best, Netherlands

4D-TRAK

2008

1.5 T

Philips Healthcare, Best, Netherlands

2008

3.0 T

Philips Healthcare, Best, Netherlands

4D-TRAK+

2011

3.0 T

Philips Healthcare, Best, Netherlands

TRICKS

1996

1.5 T

General Electric Healthcare, Milwaukee, WI, USA

PR-TRICKS

2002

1.5 T

General Electric Healthcare, Milwaukee, WI, USA

HYPR PR-TRICKS

2006

1.5 T

General Electric Healthcare, Milwaukee, WI, USA

2007

3.0 T

General Electric Healthcare, Milwaukee, WI, USA

TREAT

2005

1.5 T

Siemens Healthcare, Erlangen, Deutschland

2006

3.0 T

Siemens Healthcare, Erlangen, Deutschland

TWIST

2008

1.5 T

Siemens Healthcare, Erlangen, Deutschland

2010

3.0 T

Siemens Healthcare, Erlangen, Deutschland

1 Year = year first described in technical literature; Manufacturer = manufacturer of the MR systems on which the indicated technique was implemented at the time


As mentioned above, that was used to accelarate dynamic imaging cAVM are especially well suited for testing the effectiveness of 4D-MRA. This clinical picture was therefore examined in clinical studies using each of the methods introduced, resulting in data that facilitates good comparison. The currently employed “echo-sharing” techniques are introduced below and compared using the example of cAVM imaging in view of the good body of data:


#

“Keyhole”-based techniques

In 1993, the same year contrast-enhanced 3D-CEMRA was introduced by Prince et al., Van Vaals et al. and Jones et al. described separate techniques which involved repetitively scanning the central k-space, while acquiring the peripheral k-space only once. Afterwards, missing information from the one-time scanning of the peripheral k-space is added to each of the incomplete, central, dynamic data sets to finally generate complete k-space data sets for each point in time [2] [3].

The complete data set from which the missing k-space portions are “borrowed” is identified as the reference data set and is usually acquired after the dynamic phases upon completion of the sequence. In only 1993 Van Vaals declared that the “keyhole" technique should be combined with other acceleration methods “based on different principles” to achieve even higher acceleration.

This predicted development was realized in the subsequent two decades as a result of different techniques facilitating accelerated data acquisition actually being developed. “Temporal resolution MR angiography with CENTRA-Keyhole” (4D-TRAK, Philips Healthcare, Best, Netherlands) represents the combination of multiple acceleration techniques with keyhole imaging and already enjoys routine clinical use. This technique combines the keyhole principle with a randomized acquisition of central k-space data (CENTRA-Keyhole), parallel imaging (SENSE) with high acceleration factors and half-Fourier acquisition [25] [26] [27]. Further refinement of the technique involves subdividing the central k-space partition (the “keyhole window”) into three smaller fractions, which are likewise acquired alternatingly (temporal resolution MR angiography with keyhole and view sharing [4D-TRAK+]; [Fig. 2a]) [28].

Zoom Image
Fig. 2 K-space acquisition using various 4D-MRA techniques; K-space acquisition schemes in phase encoded (ky), slice encoded (kz) and frequency encoded direction (kx). A “zero filling” non-acquired k-space regions is not illustrated in graphics a through d. a Keyhole acquisition with view-sharing (4D-TRAK+): During dynamic imaging phase, either segment A or one of the segments B and C is scanned on an alternating basis. Area D is scanned following the dynamic acquisition phase and added to all dynamic acquisitions. b 3D-TRICKS: Initial concept with four equally sized Ky-segments (containing the same number of (ky, kz) points) with A, B, C and D. Either segment A or one of the segments B, C and D is scanned on an alternating basis. Complete k-spaces are yielded by putting together all temporally neighboring segments A, B, C and D. c TREAT: Concept with n possible regions. The regions A, B, C and D contain the same number of (ky, kz) points and are divided according to their distance from the center of the k-space into the innermost segment A and the additional peripherally located segments B, C and D. Either segment A or one of the segments B, C and D are scanned on an alternating basis, and complete K-spaces are yielded in turn by putting together the temporally neighboring segments A, B, C and D. d TWIST: In each dynamic phase the central portion of k-space is scanned, while only a fraction of the peripheral region B is scanned. This region is comprised of elliptical trajectories interwoven (“twisted”) with one another. Missing data needed for assembling a complete k-space are taken from the respective temporally neighboring dynamic phases.

Keyhole techniques were used for examining the abdomen, thorax and extremities as well as the head and neck at 1.5 T [29] [30] [31] [32] [33] [34] [35] [36] and the pelvis, thorax and extremities as well as the head and neck at 3.0 T [26] [28] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47].

For example, these techniques allowed the imaging of surgically created shunt connections between the superior vena cava and the pulmonary artery during Fontan procedures. This technique was likewise used to image dialysis shunts or to acquire purely arterial images of the lower leg region in cases of asymmetrical contrast perfusion.

In principle limitations in the complete suppression of the venous signal can appear with keyhole-based techniques if too small a fraction of the central k-space is selected. Caution is advised when using very high compression factors with 4D-TRAK+, since a flickering artefact can appear in cinematic view [28]. This is based on the fact that 100 % symmetrical boundary conditions are never present in k-space due to noise and that when high compression factors are present only positive or negative central k-space portions can be used for image reconstruction on an alternating basis. This results in deviations, albeit minor, in image contrast (“flickering”). Taking these limitations into account, a temporal resolution of 572 ms at a simultaneous spatial resolution of (1.1 × 1.1 × 1.1) mm³ was achieved when 4D-TRAK+ was used for examining cAVM, for example, while imaging all cranial blood vessels [41].


#

TRICKS

In 1996 Korosec et al. described a technique named 3D-TRICKS (time-resolved imaging of contrast kinetics) [42] in which different k-space portions are acquired over a period of time, and missing portions are borrowed from prior or subsequent data sets ([Fig. 2b]) in the sense of a temporal interpolation [43].

The peripheral (higher frequency) portions of k-space (B, C and D) are scanned three times less frequently than the central (lower frequency) portions (A). Using these k-space fractions, data collection is repeated according to the following sequence, for example: D, A, C, A, B, A. In addition, the entire k-space is scanned at the beginning and end of the TR-MRA sequence with all of its four portions.

Various enhancements to this acquisition scheme were subsequently developed. For example, (PR)-TRICKS uses radial projection reconstructions on the kxky plane combined with a variable k-space scanning rate for accelerated dynamic data acquisition [44]. Cartesian coding is then employed in the slice encoding direction. The next step was “HYPR TRICKS”, which added “highly-constrained back-projection reconstruction“ (HYPR) to improve the dynamic low-frequency data of the TRICKS-algorithm as well as (by means of high-frequency data) the signal-to-noise ratio and thus both temporal as well as spatial resolution with the aid of an additional high-resolution data set following venous filling [45] [46]. Combining all of these methods (HYPR PR-TRICKS), however, results in high sensitivity to patient movements, which can in turn compromise image quality [47]. It is therefore necessary to precisely weigh the advantages and disadvantages of these highly complex acquisition schemes to find the ideal compromise for the particular clinical application. Applications for the abdomen, thorax, extremities and head and neck at 1.5 T [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] and for the head and neck as well as the extremities at 3.0 T [65] [66] [67] [68] [69] [70] [71] [72] have been published.

Literature contains examples of clinical applications of TRICKS-4D-MRA for pulmonary angiographies, for diagnosing carotid-cavernous sinus fistulae as well as for improved diagnosing of diseases in the arteries of the lower legs, particularly in patients with diabetes mellitus. As with all 4D-MRA techniques, limitations of the TRICKS technique can appear through temporal interpolation.

In the case of HYPR-TRICKS techniques, low data density (“sparsity”) can result in signal fluctuations. It was shown that very high image refresh rates can be achieved with HYPR-TRICKS while employing to some extent very complex reconstruction algorithms and extremely low portions of actually acquired data. Based on the clinical application example of examining cAVM, an image refresh rate of 0.8 s at a voxel size of 0.5 × 0.75 × 4mm³ was achieved using a TRICKS algorithm [69].


#

TREAT

TREAT (time-resolved echo-shared angiography technique) was first described in 2005 by Fink et al. and divides the k-space into n regions (for example, regions A, B, C and D when n = 4) [73]. Each region covers an equal portion of k-space and thus the same number of (ky, kz) points ([Fig. 2c]).

Following complete k-space acquisition, the scanning of the central segment A and one of the peripheral segments B, C or D alternates similar to the scheme employed in the TRICKS technique. To reconstruct the images, complete k-space data is generated by summarizing the k-space data of the next consecutive acquisitions.

For a segmentation pattern with n segments, a complete data set can then be reconstructed in intervals of 2 • TA/n (TA being the time needed for obtaining a complete data set). However, as the number of segments increases, the signal intensity is distributed further over the image plane, resulting in increasingly less defined smaller blood vessels.

Numerous studies use this technique on various regions of the body at 1.5 T [12] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] and in the head and neck at 3.0 T [85] [86] [87] [88]. Among the published indications for these examinations are the imaging of pathological flow conditions with the occurrence of endoleaks in cases of aortic prostheses, [76] as well as diagnosing pulmonary embolisms [83] or subclavian steal syndromes [84]. Possible limitations arise from the fact that each data set is reconstructed from similarly sized portions of k-space partitions from various points in time and is thus interpolated over time. This particularly effects the imaging of small blood vessels to the extent that neighboring small arteries and veins can no longer be distinguished from one another. Using the TREAT algorithm, a temporal resolution of 1.5 s per 3 D data set with a simultaneous spatial resolution 1.2 × 1.0 × 4 mm3 was achieved when imaging cAVM [88].


#

TWIST

The details of the TWIST technique (time-resolved imaging with stochastic trajectories) were reported by Vogt, Lim and Song over the period of 2007 to 2009 [89] [90] [91]. In this technique, all points in the k-space are sorted according to their radial distance from the center of the k-space. A critical radial distance can be defined around the k-space and subdivided into two subareas, a central region A (low-frequency portions) and a peripheral region B (high-frequency portions) ([Fig. 2 d]).

During data acquisition, the data of the entire k-space is gathered only once, either at the beginning or end of the sequence. For the dynamic phase of image acquisition, the entire region A within each time window is acquired, while one of n portions is scanned from region B, and missing data points are taken from temporally adjacent acquisitions of region B for complete k-space reconstruction.

In this process, the k-space trajectories of region B follow a spiral pattern on the ky-kz plane, with the trajectories from region B being intertwined with one another, giving the sequence its name. With TWIST, acceleration is based on both the size of region A and the density of trajectories in region B. This technique has been used on the abdomen, thorax, extremities as well as on the head and neck at both 1.5 T [92] [93] [94] [95] [96] and 3.0 T [92] [97] [98] [99] [100] [101] [102] [103] [104]. TWIST has likewise been used to image ovarian vein reflux [92] as well as thoracic venous outflow obstructions [94], examine pathologies of the abdominal aorta [100] and successfully diagnose changes in flow conditions in cases of peripheral occlusive arterial disease [97]. However, the complex acquisition pattern of TWIST data recording with its data collection interwoven over the acquisition time poses special challenges for assigning artefacts to certain k-space portions, which is made even more difficult by the fact that artefacts are always present in multiple consecutive data sets because of the temporal interpolation reconstructed of the data. Regardless of these limitations, patients with cAVM, for example, have been successfully examined at a temporal resolution of 0.58 s and a voxel size of (1.6 × 1.6 × 1.6)mm3 [103]


# Summary and supplemental technical observations

While implementing TR-MRA demands high standards for software and hardware, current tomography machines from major manufacturers implement high-quality manufacturer-specific TR-MRA sequences. Each of these techniques has its pros and cons that would make particular methods appear to be optimal for particular cases.

It must be emphasized that images reconstructed from complex temporally intertwined data generally do not correspond to any exact visualization of a single time point (temporal interpolation). However, the central k-space portions used exclusively at a particular point in time are essential for the image contrast of the corresponding image so that the contrast curve can be realistically mapped.

Because of the potential cons of this temporal interpolation, Riederer et al. have introduced the concept of “temporal footprint" of high temporal resolution sequences. This concept describes the period of time that is needed to acquire a complete k-space data set, i. e. the entire k-space portions. Dividing the k-space into multiple portions in the process of time-resolved imaging would produce the following "temporal footprint" in which the exemplary constellation has 4 equal-sized portions A-D (acquisition duration 1 s each, A = central k-space portion):

Example 1 ([Fig. 3a]): Acquisition in the sequence A, B, C, D...; central k-space portions are acquired every 4s; the temporal footprint is the product of the summation of individual k-space portions A-D, with the acquisition time between each being omitted, resulting in a temporal footprint = 4 s.

Zoom Image
Fig. 3 Image refresh rate (contrast refresh) scheme compared to "temporal footprint" in the case of two different acquisition schemes a, b.

Example 2 ([Fig. 3b]): Acquisition in the sequence A, B, A, C, A, D…; central k-space portions are acquired alternatingly with peripheral k-space portions B, C or D every 2 s. The temporal footprint is, in contrast, the product of the summation of individual k-space portions A-D plus the intermediate acquisition times, resulting in this case in a temporal footprint = 6 s.

In Example 2, contrast information is accordingly refreshed every 2 s, while only every 4 s in Example 1. However, Example 1 is the temporally “cleaner” representation, since the temporal footprint is shorter. For TRICKS, TREAT and TWIST the “temporal footprint” is in each case accordingly a multiple of the time duration of an individual dynamic phase and reflects the period of time over which the acquisition of a complete k-space data set extends.

On the other hand, it would not be wise to apply this concept to keyhole-based methods, since the entire peripheral k-space in this instance is acquired only once, which would thus yield periods of differing length for the “temporal footprint”, each depending on what point in time the acquisition of the central k-space is observed [28].

Regardless of the technique employed, a high degree of temporal accuracy in the collected dynamic data is desired in clinical applications. Future studies are needed to more accurately examine which technique constitutes the optimal compromise between temporal interpolation, “temporal footprint” and image for a particular clinical problem [105] [106].

For precisely imaging smaller structures, in particular, the signal-time-curve (“point spread function”) is ultimately of special importance. Temporal interpolation can result in inferior definition that can manifest itself in, for example, limitations in imaging small arterial supply vessels of arteriovenous malformations [26] [107]. With increasing segmentation, the inferior definition of the peak of the signal-to-time curve becomes more pronounced, which also limits the degree of the possible segmentation steps in the increasingly complex acquisition schemes [73].


#

Clinical perspectives

The technical methods for acquiring 3D-CEMRA image series, some of which are highly complex, are already facilitating detailed observation of numerous pathological changes in flow that were previously the domain of invasive DSA. [Fig. 1b] provides a quantitative overview of which methods have come into use in the past few years. [Fig. 3], [4], [5], [6] provide several examples of images, while [Table 2] offers an overview of clinical applications to date and shows the multifaceted application possibilities of 4D-MRA. Established routine clinical applications in the meantime include the diagnosis of arteriovenous malformations or fistulae. Unlike conventional static sequences, these applications allow the imaging of premature filling of veins or sinuses by shunt mechanisms. Pathological flow conditions that would otherwise be possible to reconstruct only with invasive methods can likewise be shown in cases of aortic dissections, endoleaks from aortic prostheses or subclavian steal syndrome. Finally, the method sees frequent routine clinical use in generating purely arterial images of the lower leg in cases of asymmetrical arterial filling due to upstream stenoses (in peripheral occlusive arterial disease) or arteriovenous shunts (often in cases of diabetic microangiopathy). In these cases, precise timing for adjusting arterial perfusion and thus a pure arterial 3D-CEMRA is oftentimes not possible.

Zoom Image
Fig. 4 75-year old patient with stage III peripheral arterial occlusive disease, ambulant < 50 m, complaints primarily on the right side. Total volume maximum intensity projections of high temporal resolution 4D-MRA show the lower legs prior to arrival of contrast agent (a), arterial bolus passage (bd), and premature venous enhancement in the right lower leg resulting from a shunt between the right peroneal artery (arrows in d, e) and the great saphenous vein, as well as venous filling in a later phase f.
Zoom Image
Fig. 5 69-year old female patient with acute swelling of the left arm presenting septic clinical picture, imaging performed to exclude thrombosis. Representative total volume maximum intensity projections per one phase of 4D-MRA prior to arrival of contrast agent bolus a, during arterial bolus passage b and during venous phase of contrast agent passage c.
Zoom Image
Fig. 6 25-year old male patient before a – f and after g – l surgical resection of a right frontal cerebral arteriovenous malformation (arrow in d) in a non-eloquent area of the brain. Lateral total volume maximum intensity projections of temporally corresponding phases of contrast-enhanced 4D-MRA.
Table 2

Clinical applications of TR-MRA.

bodily region

clinical problem

references

head

carotids

[95] [114] [115] [116]

cerebral arteriovenous fistulae and malformations

[33] [47] [65] [117] [118] [119]

cavernous hemangioma

[117] [120]

sinus thrombosis

[33] [121]

multiple sclerosis

[122] [123] [124]

orbital lesions

[65] [125]

spinal column

spinal arteries

[30] [71] [126]

arteriovenous fistulae and malformations

[54] [72] [76]

thorax/abdomen

aortic dissection

[17] [127]

aortic isthmus stenosis

[128]

pulmonary embolisms

[63]

pulmonary perfusion

[9] [10] [12] [13] [29]

pulmonary hypertension

[8] [11]

arteriovenous malformations

[36] [129]

endoleaks from vascular prostheses

[82] [130] [131]

subclavian steal syndrome

[90]

coronary vessels

[132] [133] [134]

renal artery stenoses

[135] [136] [137]

pelvic congestion

[32] [57] [98] [138] [139]

arteries of the extremities

asymmetrical contrast perfusion

[56] [99] [140] [141] [142] [143]

arteriovenous malformations

[144] [145] [146]

diabetic microangiopathy

[31] [147] [148] [149] [150]

That cAVM was examined notably using 4D-MRA techniques [23] [26] [61] [69] [72] [103] [150] [151] [152] [153], can be explained by the fact that on the one hand intracranial vessels are particularly suited given their fundamentally low susceptibility to artefacts in view of the circumscribed examination volume without significant patient movements including breathing. On the other hand, the dynamic information offers especially high potential benefits, and the reference standard of DSA is frequently on hand for comparative purposes.

TR-MRA, however, also opens numerous new diagnostic avenues. For example it is conceivable when using an intravascular contrast agent to first acquire detailed functional information with the aid of TR-MRA (e. g. for diagnosing a pulmonary embolism) and then to image an underlying leg vein thrombosis with the aid of a high-resolution spatial sequence [154] [155]. Primarily intravascularly dwelling contrast agents of this type with reversible protein binding and thereafter considerably delayed pervasion of the interstitium are currently available only in North America.

In the studies on clinical applications of TR-MRA some of which involved high image refresh rates and high spatial resolution, the majority of temporal information was displayed and at the same time the contrast agent dose administered was oftentimes significantly reduced compared to the static, high spatial resolution 3D-CEMRA. In the future, combining echo-sharing techniques with special data reconstruction methods using only a fraction of the actually required data (HYPR, compressed sensing) may possibly allow even significantly higher image refresh rates, thereby facilitating the generation of real-time sequences and detailed examination of flow conditions [156] [157].

Disadvantages compared to DSA resulting from the simultaneous contrast agent perfusion in all vascular segments (a not insignificant advantage of DSA is the possibility of selective contrast agent administration) are compensated partly with a vascular selective excitement using what is known as arterial spin labelling (ASL) [34] [158]. In the future, further insight on flow dynamics may possibly be gained through techniques such as high temporal and spatial resolution and at the same time quantitative phase contrast angiography (4D-PC-MRA), which has lately shown considerable promise [159] [160] [161] [162] [163] [164] [165]. Finally, real-time imaging, for example during catheter interventions with susceptibility markers, is a promising area of application for 4D-MRA. However, this is the subject of predominantly preclinical studies [165] [167].


#

Summary

TR-MRA was initially used only for facilitating fluoroscopic bolus triggering and gaining exact timing for performing a high spatial resolution static, pure arterial 3D-CEMRA, for instance, of the supra-aortic arteries. With the aid of complex k-space acquisition algorithms, parallel imaging and new echo-sharing techniques while at the same time at higher spatial resolution. This is opening up new areas of application that had mainly eluded non-invasive testing until now. It will be exciting to see what benefits further technological advancements in both coil and scanner technology, such as for example fully digitalized signal transmission and reception, will have for TR-MRA in the future and what importance this innovative technology will have in routine clinical use.


#
#

Deutscher Artikel/German Article


Correspondence

Dr. Dariusch Reza Hadizadeh
Radiology, University of Bonn
Sigmund-Freud-Str. 25
53113 Bonn
Germany   
Phone: ++ 49/2 28/28 71 58 75   
Fax: ++ 49/2 28/28 71 60 93   


Zoom Image
Fig. 1 Original publications on time-resolved MRA between 2000 and 2012; a Number of publications related to field strength; b Publications broken down by TR-MRA technique.
Zoom Image
Fig. 2 K-space acquisition using various 4D-MRA techniques; K-space acquisition schemes in phase encoded (ky), slice encoded (kz) and frequency encoded direction (kx). A “zero filling” non-acquired k-space regions is not illustrated in graphics a through d. a Keyhole acquisition with view-sharing (4D-TRAK+): During dynamic imaging phase, either segment A or one of the segments B and C is scanned on an alternating basis. Area D is scanned following the dynamic acquisition phase and added to all dynamic acquisitions. b 3D-TRICKS: Initial concept with four equally sized Ky-segments (containing the same number of (ky, kz) points) with A, B, C and D. Either segment A or one of the segments B, C and D is scanned on an alternating basis. Complete k-spaces are yielded by putting together all temporally neighboring segments A, B, C and D. c TREAT: Concept with n possible regions. The regions A, B, C and D contain the same number of (ky, kz) points and are divided according to their distance from the center of the k-space into the innermost segment A and the additional peripherally located segments B, C and D. Either segment A or one of the segments B, C and D are scanned on an alternating basis, and complete K-spaces are yielded in turn by putting together the temporally neighboring segments A, B, C and D. d TWIST: In each dynamic phase the central portion of k-space is scanned, while only a fraction of the peripheral region B is scanned. This region is comprised of elliptical trajectories interwoven (“twisted”) with one another. Missing data needed for assembling a complete k-space are taken from the respective temporally neighboring dynamic phases.
Zoom Image
Fig. 3 Image refresh rate (contrast refresh) scheme compared to "temporal footprint" in the case of two different acquisition schemes a, b.
Zoom Image
Fig. 4 75-year old patient with stage III peripheral arterial occlusive disease, ambulant < 50 m, complaints primarily on the right side. Total volume maximum intensity projections of high temporal resolution 4D-MRA show the lower legs prior to arrival of contrast agent (a), arterial bolus passage (bd), and premature venous enhancement in the right lower leg resulting from a shunt between the right peroneal artery (arrows in d, e) and the great saphenous vein, as well as venous filling in a later phase f.
Zoom Image
Fig. 5 69-year old female patient with acute swelling of the left arm presenting septic clinical picture, imaging performed to exclude thrombosis. Representative total volume maximum intensity projections per one phase of 4D-MRA prior to arrival of contrast agent bolus a, during arterial bolus passage b and during venous phase of contrast agent passage c.
Zoom Image
Fig. 6 25-year old male patient before a – f and after g – l surgical resection of a right frontal cerebral arteriovenous malformation (arrow in d) in a non-eloquent area of the brain. Lateral total volume maximum intensity projections of temporally corresponding phases of contrast-enhanced 4D-MRA.