Black-blood MRI at 7T using 2D RARE MRI: In vitro testing and in vivo demonstration

Abstract

Vessel walls play a crucial role in many inflammatory vascular diseases. Vessel wall imaging (VWI) using mangnetic resonance imaging (MRI) is one of the few methods by which vessel walls and inflammation can be visualized noninvasively, in vivo, and without ionizing radiation. VWI is based on black-blood (BB) MRI, where the signal from flowing blood is suppressed and contrast agent accumulation in the (inflamed) vessel wall is highlighted. Here, high resolution, T1 weighting, suppression of fat and flowing spins is essential. Whereas VWI is often applied in humans, only very few reports describe its use in small animals.

Here, we investigated whether BB MRI for rodents can be implemented using a state-of-the-art, but commercially available, preclinical MRI system and imaging sequence. We identified 2D spin-echo (RARE)-based BB-MRI as a promising sequence that is widely available and not vendor dependent. First, we investigated the properties of the sequence in vitro with respect to image contrast, resolution, the suppression of signal of flowing spins and fat using a newly developed, 3D-printed model setup (cylindrical model with exchangeable nuclear magnetic resonance tubes and flow tube in agarose, printed with stereolithography). For example, good signal-to-noise ratio, BB and T1 contrast were obtained for TE = 5 ms for slice thickness equal or below 0.352 mm or slice thickness = 0.8 mm with TE at least 25 ms. In vivo, we obtained a pronounced BB effect for both intracranial and abdominal vessels of healthy rats down to a 0.25 mm diameter in no more than 1:36 min with TE = 12 ms, TR = 750 ms, voxel 156 × 156 × 800 µm³, and 11 slices. Compared to in vitro, we were able to reduce TE without apparent artifacts likely because the flow was faster in vivo than in vitro. Additionally, we needed to increase the resolution to image small vessels.

Thus, we found that BB-MRI with 2D spin-echo sequences is feasible on rodents with state-of-the-art, commercially available preclinical MRI systems. We believe that these results will facilitate the development and application of rodent VWI in longitudinal studies, which, in comparison to histology, may reduce the number of needed animals and intersubject variability at the same time.

Introduction

Vascular disease is one of the leading causes of death worldwide [1]. Early diagnosis, stratification, and therapy control are therefore crucial factors for successful treatment in these patients. Impressive diagnostic results have been obtained using imaging techniques such as computed tomographic angiography, digital subtraction angiography, or magnetic resonance angiography. With these methods the vessel lumen can be assessed, for example, to visualize stenosis or aneurysms [2]. Here, bright-blood methods play an important role (the blood appears brighter than the surrounding static tissue). While these methods were developed to image the inner lumen of the vessel, they are not well suited to image the vessel wall itself. For many diseases, however, the vessel wall is of great interest, as this is where the disease develops.

For example, in atherosclerotic diseases, the arterial vessel responds to plaque growth. As a result, the vessel either expands (positive remodeling) or shrinks (negative remodeling). In cases of positive remodeling, the plaque was found to be more vulnerable and required more aggressive treatment [3].

To image the vessel wall and its inflammatory alterations, vessel-wall imaging (VWI) using magnetic resonance has emerged. Here, the signal of the blood is suppressed, such that the wall can be evaluated in detail. VWI is the only method by which inflammation or damage to the wall of intracranial arteries can be visualized directly. It can depict intracranial vasculopathy and help identify and improve the diagnosis and treatment of cerebrovascular diseases [4], [5]. VWI is also used to identify the underlying pathological process in aneurysms, vasculitis, dissection, or atherosclerosis [6].

VWI comes with several challenges, however. For one, it requires high spatial resolution, as intracranial vessel walls are thin (e.g., 0.3 mm) [7]. In fact, wall thickness can be used as a biomarker for atherosclerosis and other diseases [8]. At the same time, large spatial coverage is often needed, as the location of the pathology is not known beforehand. Furthermore, fat suppression and T1 weighting (T1w) are needed to show contrast agent accumulation in the wall [2]. Finally, the sequence must suppress the signal of flowing spins and be short, e.g., < 10 min, so that it can be applied routinely in clinical settings.

Thus, a VWI sequence typically must 1) suppress the signal in the vessel lumen (black-blood magnetic resonance imaging, BB, MRI), 2) have high resolution < 0.7 mm, 3) provide T1 contrast, and 4) suppress the signal from fat.

BB-MRI was first mentioned in 1992 by Chien et al, [9] and several approaches have been proposed since then. To suppress the signal in the lumen, either the motion (flow) of the spins or the difference in relaxation times with respect to the surrounding tissue were used. In the latter case, a slice-selective saturation pulse or double-inversion recovery pulse is applied to suppress the signal of blood. 2D spin-echo sequences make use of the fact that spins flowing out of the slice are not refocused by the 180° pulse and thus do not contribute to the signal [10]. In addition, BB-MRI takes advantage of intravoxel dephasing due to complex flow patterns, which further suppresses the signal in the vessel lumen [11].

While BB-MRI is often used in humans, only few examples have been described for rodents, none of which is based on 2D spin-echo sequences.

The challenges of translating VWI from humans to mice and rats include the much smaller anatomy and higher heart rate in these animals, and a higher field strength is required. The matter is complicated further by the fact that preclinical MRI scanners (designed for small animals) are not produced by the same vendors that manufacture those for humans; thus, sequences cannot be transferred easily (this problem may improve in the future with vendor-independent pulse programming such as Pulseq [12] or GammaStar [13]).

Still, some animal BB-MRI experiments have been reported in the literature: two in mice, and once each in rats, rabbits, and swine. An interesting work, Lefrancois et al. performed FLASH-based CINE MRI in mice at 9.4T with a T1w and BB effect. Flow suppression was achieved by using a bipolar gradient and additional saturation slices [14]. Ito et al. showed that BB-MRI at 7T with slices perpendicular to the vessel axis is a feasible method for evaluating delayed cerebral vasospasm in rats and is less invasive than histopathological examinations and measurements can be readily repeated [15]. Mulder et al. used T1w contrast-enhanced MRI as well as liposome-enhanced MRI in mice at 6.3T, and Hur et al. applied BB-MRI with saturation recovery in rabbits at 3T to identify inflammation in atherosclerosis [16], [17]. Worthley et al. performed BB-MRI in swine at 1.5T using a double-inversion recovery pulse sequence [18]. As pointed out before, however, none of these approaches used the current standard technique for human BB-MRI and VWI, namely, slice-selective spin-echo sequences with spectral fat suppression.

Thus, the goal of this work was to determine whether the spin-echo sequences in current state-of-the-art preclinical scanners are suited for high-resolution, cranial and abdominal BB-MRI in rodents. More specifically, we evaluated whether the sequence provided a) flow suppression (in the lumen), b) T1w (but not T2w) contrast, c) sufficient resolution for cranial vessels, and d) fat suppression in e) a reasonable time period. To achieve this goal with the least number of animals (following the 3R principle [19]), we developed a 3D-printed, in vitro model for measuring these properties and analyzing the sequence under standardized conditions (T1w, T2w, and flow and fat suppression). After gaining insights on the sequence performance by varying the parameters in vitro, we optimized the sequence in vivo and obtained strong BB-MRI contrast in the abdominal and intracranial vessels of healthy rats.

Materials and methods

In vitro model for VWI/T1w BB-MRI

To test and optimize VWI in vitro, we aimed to develop a setup that has 1) at least four model solutions (with different T1 and T2 contrasts) that can be exchanged easily, 2) a tube with a flowing medium where the velocity of the flow can be adjusted, 3) homogeneous, non–flowing “tissue” for evaluating artifacts, 4) a size similar to that of a rat which fits into a rat body coil, 5) is MR compatible, and 6) can be manufactured with reasonable effort.

We accommodated these requirements by designing a 3D-printed, cylindrical model, where a central flow tube was embedded in agarose gel and surrounded by four holders for 5 mm nuclear magnetic resonance (NMR) tubes (Autodesk Inventor 2021, USA).

The model was printed with stereolithography using a mixture of methacrylic acid esters and a photoinitiator (Form 3, Clear Resin (urethane dimethacrylate), Formlabs, USA).

The model consisted of a cylindrical container with an inner diameter of 43 mm (x-y axis) and a length of 106 mm (z-axis). This large cylinder contained five parallel, cylindrical holes 1–5 of equal length along the z-axis: holes 1-4 had an inner diameter of 5.575 mm to fit standard NMR tubes (outer diameter 5 mm, length 177 mm) and were arranged equidistantly around the center; another cylindrical hole, no. 5, with an inner diameter of 15 mm, was placed in the center. One to four indentations were added along cylinders 1–4 to facilitate identification on MRI. Another cylinder, no. 6, was placed in the center of cylinder 5, creating a lumen between the two cylinders and inside cylinder 6 (inner diameter 6.25 mm, outer diameter 14.25 mm).

A silicone tube (4-mm inner diameter, 1-mm wall strength) was fed through cylinder 6 and connected to a peristaltic flow pump (≈5 cm/s, Masterflex, Computerized Drive, Cole Parmer, UK). An inline transonic flow sensor was put on the silicone tube before performing MRI to measure the flow velocities (ME6PXN325, Transonic System Inc.).

Model solutions

Four NMR tubes were filled with either oil, tap water, or aqueous mixtures of gadolinium (Gd)-based contrast agent: isotonic saline solution 1:10000 and 1:8000 vol% and placed in tubes 1–4 (Fig. 1, Gadovis 1,0 mmol/ml, Bayer, Germany). The lumen around cylinders 1-4 was filled with a non–viscous model solution containing 1% agarose gel in tap water, and the lumen between cylinders 5 and 6 was filled with 1:4000 Gd:isotonic saline solution. Tap water was used as a flow medium.

Fig. 1.

Schematic view of the cross-section (a) and photo (b) of the VWI model we developed. The model consisted of a 3D-printed cylindrical container filled with 1% agarose gel, four nuclear magnetic resonance tube holders, and a hollow, double-walled cylinder in the center. A silicone tube with flowing water was fed through the central cylinder and the space between its double walls filled with a model solution. In the agarose, the regions of interest (ROI) left and right (blue, in read-out direction, ROI_ro) and above and below (red, in phase-encoding direction, ROI_pe) were used to quantify flow artifacts, while the flow tube (ROI_flow) was used to quantify flow suppression. The regions used to measure the noise are indicated by dashed circles. The indicators (bumps) printed on the tube holders are meant to facilitate their identification.

Magnetic resonance imaging

A 7-T preclinical MRI equipped with an 86-mm quadrature volume transmit-receive coil was used for MRI (BioSpec 70/30, Paravision 360, Avance NEO, Bruker, Germany). For in vivo brain MRI, a 2 × 2 channel receive coil was added.

In vitro MRI

The longitudinal and transverse relaxation times were quantified with T1 and T2 maps provided by the system's manufacturer. To acquire a T1 map, a rapid acquisition with relaxation enhancement (RARE) sequence with variable repetition times (TR) ranging from 867 ms to 7500 ms was used (TE = 7 ms). For the T2 map, a multi-slice, multi-echo (MSME) sequence with variable echo times (TE) ranging from 75 ms to 1050 ms with an echo spacing of 75 ms was employed (TR = 2200 ms).

For BB-MRI, a slice-selective RARE sequence with spectral fat suppression was used [20]. Here, a spectral fat suppression pulse was followed by a 90° excitation pulse and multiple 180° refocusing pulses separated by the echo spacing. Two echoes (RARE factor) were acquired after each excitation to reduce T2w.

To investigate the BB properties of the sequence, we kept TR, scan time, and in-plane resolution constant and varied TE (5 ms to 30 ms), slice thickness (0.8 mm to 0.35 mm), and flow velocity (2.87 cm/s to 32.05 cm/s, Table 1).

Table 1.

MRI acquisition parameters used for in vitro experiments

	T1w RARE	2D PC	T1 map RARE	T2 map MSME
(effective) TE [ms]	5 (5–30)	5	7	(75–1050)
TR [ms]	1200	50	(867-7500)	2200
FOV [mm2]	45 × 45	58 × 58	45 × 45	45 × 45
Number of slices	15	1	1	1
Pixel size [mm2]	0.35 × 0.35	0.45 × 0.45	0.35 × 0.35	0.35 × 0.35
Slice thickness [mm]	0.8 (0.8–0.3)	1	1	1
Bandwidth [Hz]	87719	100000	78125	62500
RARE factor	2		2
Echo spacing [ms]	5		7	75
Venc [cm/s]		80
Flip angle	90°/180°	30°	90°/180°	90°/180°
Scan time	1 min 16 s	12 s	17 min 43 s	4 min 41 s

Additionally, 2D phase-contrast (2D PC) MRI was used to measure the velocity of the flow in one slice and one direction. Here, the bipolar gradient provided a flow-dependent signal phase. From the difference between two acquisitions with opposite polarity of the flow-encoding gradient, the velocity component was calculated by the manufacturer’s software [21].

In vivo MRI

For in vivo MRI, we examined two male Wistar rats (RjHan:WI from Janvier Lab) . The animals were purchased at 6 weeks of age (Saint Berthevin Cedex, France) and housed in a specific-pathogen-free (SPF) facility (Central Animal Facility of the University Medical Center Schleswig–Holstein, Kiel, Germany). They were ∼10 weeks of age and weighed ∼380 g at the time of measurement.

This study was conducted in compliance with the German Animal Protection Law. The animal protection committee of the local authorities (Ministry of Energy, Agriculture, the Environment, Nature and Digitalization Schleswig-Holstein (MELUND)) approved all experiments (V242-75981/2020(2-1/21)). The experimental procedures conformed with those guidelines and with the principles outlined in the European Directive 2010/63/EU. This study is reported in accordance with ARRIVE guidelines.

The rats were anesthetized by intraperitoneal injection of 75 mg/kg ketamine and 0.5 mg/kg medetomidine and placed on a heated animal bed. A pressure pad was used to monitor their breathing and trigger abdominal MRI (SA Instruments, USA). For imaging the brain, a 2 × 2 channel receive-only surface coil (Bruker, Germany) was added to the body resonator. BB-MRI was acquired in the brain and abdomen with different parameter sets (Table 2). After the MRI measurements, the rats were euthanized without being awakened.

Table 2.

Acquisition parameters for in vivo T1w RARE sequences.

	Brain - model parameters	Brain optimized	Brain high resolution	Abdomen
TE [ms]	5	12	6	12
TR [ms]	1200	750	750	750
FOV [mm2]	45 × 45	40 × 20	35 × 35	60 × 60
Number of slices	15	11	9	12
Pixel size [mm2]	0.35 × 0.35	0.156 × 0.156	0.137 × 0.137	0.234 × 0.234
Slice thickness [mm]	0.8	0.8	0.8	0.8
Bandwidth [Hz]	87719	87719	87719	88487
RARE factor	2	8	2	8
Echo spacing [ms]	5	6	6	75
Scan time	1 min 16 s	1 min 36 s	3 min 12 s	3 min 12 s

Image analysis

To quantify the suppression of flow artifacts and signal of flowing spins, the mean and standard deviation (SD) of the signal were measured on the images in several ROIs (Fig. 1, Paravision 360, Bruker, Germany).

To quantify the artifacts induced by moving spins, four rectangular ROIs were placed in the agarose gel between the NMR tubes: two in phase-encoding (ROI_pe), where the artifacts occurred, and two in readout direction (ROI_ro), where no artifacts were apparent (Fig. 1). To quantify the artifacts, we compared the standard deviation of the signal in the ROIs in phase encoding direction, where artifacts are expected, to those in readout direction, where we expect none.

To quantify the flow suppression, a circular ROI was placed in the silicone tube in the center of the model (ROI_flow, Fig. 1) and compared to the mean value of two circular ROIs outside of the model solution to estimate the noise. The BB effect was quantified by comparing the signal intensity of the flow (ROI_flow) to the noise outside the model.

To calculate the signal-to-noise ratio (SNR), the signal intensity for a region without apparent artifacts (ROI_ro) was divided by the standard deviation of the noise outside of the model.

Results

In vitro setup

The newly developed, cylindrical model (Fig. 1) was readily set up in the preclinical MRI in a rat body coil and we were able to investigate different contrasts and flow properties.

T1 and T2 maps were used to quantify the relaxation properties of the model solutions (Fig. 2, Table 3). The longitudinal relaxation times varied between T1(fat, S_oil) = (476 ± 28) ms and T1(agarose, S_agarose) = (2805 ± 85) ms for agarose gel. The transverse relaxation times varied from (61 ± 3) ms for fat (S_oil) up to (1097 ± 50) ms for water (S_water, mean plus standard deviation).

Fig. 2.

T1w, fat-suppressed RARE MRI (a), T1 map (b) and T2 map (c) of the model. The model was suitable for evaluating fat suppression (S_oil), T1w (compare S_Gd1:8k and S_Gd1:10k), T2w (compare S_water, S_agarose), flow suppression (center), and flow artifacts. Note the printed markers to label the four NMR tubes. Sequence parameters: T1w RARE: TE = 5 ms, RARE factor = 2, TR = 750 ms, voxel size = 0.35 × 0.35 × 0.8 mm³, FOV = 45 × 45 mm², v≈5 cm/s; T1 map: TE = 7 ms, TR = (867-7500) ms, voxel size = 0.35 × 0.35 × 1 mm³, FOV = 45 × 45 mm²; T2 map: TE = (40-560) ms, TR = 2200 ms, voxel size = 0.35 × 0.35 × 1 mm³, FOV = 45 × 45 mm².

Table 3.

Longitudinal and transverse ¹H relaxation times of different solutions acquired with a RARE-based method (TR variation) and MSME-based method (TE variation).

Solution		T1 [ms]	T2 [ms]
Soil	Food-grade olive oil	476 ± 28	61 ± 3
Swater	Tap water	2741 ± 80	1097 ± 50
SGd1:10k	Gd:Isotonic saline solution 1:10000	1200 ± 29	670 ± 15
SGd1:8k	Gd:Isotonic saline solution 1:8000	842 ± 23	530 ± 7
SGd1:4k	Gd:Isotonic saline solution 1:4000	710 ± 15	450 ± 5
Sagarose	Agarose 1%	2805 ± 85	171 ± 5

The contrast between water and agarose was well suited to evaluate T2 contrast (similar T1, but very different T2), and the contrast between Gd 8000 and Gd 10000 was used to evaluate T1 contrast (similar T2, different T1). Fat suppression was evaluated using the signal of the tube filled with oil.

The velocity of water flowing in the center tube was varied between 2.8 and 32.1 cm/s by adapting the rotation of a flow pump (10–120 RPM), calibrated using the ultrasound sensor.

Investigation of BB effect

Satisfied with the contrasts of the model, we set out to investigate the BB properties of a T1w 2D RARE sequence with fat suppression as described above.

Strong fat suppression (low signal of S_oil), prominent T1w, and little T2w was observed (similar intensities of S_water and S_agarose, different intensities for S_Gd1:4k, S_Gd1:8k and S_Gd1:10k).

For low velocities, severe flow artifacts were observed on the agarose gel in phase-encoding direction but none were apparent in readout direction (Fig. 3). We quantified these artifacts by comparing the SD of the agarose signal in the phase-encoding (with artifacts) and readout directions (no apparent artifacts). The SD in phase-encoding direction (SD_pe) first decreased slowly (for v= 2 cm/s to 5 cm/s), then rapidly and converged asymptotically to the SD in readout direction (SD_ro). An increase in the velocity beyond 15 cm/s had little effect on the artifacts, and no flow artifacts were apparent for a velocity of 21 cm/s.

Fig. 3.

Effect of flow velocity on T1w 2D RARE MRI. T1w RARE MRI of the VWI model (a-g), schematic view of the model with ROIs (h), quantified flow artifacts (i), and signal in the flow tube (j) as a function of flow velocity. Both artifacts and signal in the flow tube were found to converge asymptotically to the reference level with increasing velocity (artifact-free agarose or apparently signal-free region, respectively). All MRI parameters were kept constant (TE = 5 ms, RARE factor = 2, TR = 1200 ms, voxel size = 0.35 × 0.35 × 0.8 mm³, FOV = 45 × 45 mm²), and velocities were quantified with 2D PC MRI.

The signal in the flow tube (ROI_flow) was found to decrease asymptotically with increasing flow, too. For velocities > 25 cm/s, the signal reached the noise level (as measured in apparently signal-free regions outside of the model). These results confirmed our expectations that higher velocities improve the flow suppression and reduce flow artifacts.

Being able to vary the flow velocity in a controlled manner is one of the advantages of using a model setup (and not possible in vivo). The maximum velocity chosen was similar to that of blood in the rat carotid artery, which is about 35 cm/s [22]. Still, higher velocities will only improve the BB effect, and thus allow a shorter echo time. To evaluate the effect of the other sequence parameters, we chose a slow flow producing strong artifacts, namely, 5 cm/s, for subsequent experiments. It should be noted, though, that the flow tube used here has a diameter much larger than many vessels in vivo.

The flow artifacts, the flow signal, and SNR of static tissue decreased approximately linearly with decreasing slice thickness, while the contrasts were not affected (as expected) (Fig. 4).

Fig. 4.

Effect of the slice thickness on T1w RARE of the VWI model (a-g), schematic view of the model and ROIs (h), quantified flow artifacts (i), and signal in the flow tube (j) as a function of slice thickness. Both artifacts and flow suppression decreased about linearly towards the reference level. As expected, the noise increased for thinner slices. All other parameters (apart from the slice thickness) were kept constant (TE = 5 ms, RARE factor = 2, TR = 1200 ms, pixel size = 0.35 × 0.35 mm², FOV = 45 × 45 mm², v≈5 cm/s).

For the thinnest slice, 300 µm, the artifacts and the signal in the flow tube approached the noise level but did not reach it. A good BB effect was observed for a slice thickness of 0.3 mm, while some flow artifacts were still visible. Again, the observed trends were expected, as thinner slices reduce the time needed for spins to leave the imaging slice.

Increasing the echo time resulted in an asymptotic decline in flow artifacts, while visually, SNR, contrast, and fat suppression were hardly affected (Fig. 5).

Fig. 5.

Effect of echo time on T1w 2D RARE. T1w RARE MRI of the VWI model (a-f), schematic view of the model and ROIs (g), quantified flow artifacts (h), and signal in flow tube (i) as a function of echo time. Both artifacts and signal in the flow tube were found to decrease and converge asymptotically to the reference level. A prolongation of the echo time reduced the flow artifacts and increased the BB effect. All parameters (apart from the echo time) were kept constant at RARE factor = 2, TR = 1200 ms, voxel size = 0.35 × 0.35 × 0.8 mm³, FOV = 45 × 45 mm², v≈5 cm/s.

Echo times larger than 15 ms reduced the flow artifacts only marginally. Similarly, an increased echo time also resulted in an asymptotically increased suppression of the flow signal. For an echo time of 25 ms, the signal in the flow tube was close to the noise level.

The contrasts did not change notably for any of the echo times tested. Strong T1 contrast and little T2 contrast (compare S_water and S_agarose) were maintained. The SNR was only slightly decreased for increasing echo time, by about 10 % from TE 5 ms to TE 25 ms (Fig. 5i). An increase in noise was not visually apparent.

Visually, for an echo time of 25 ms, only slight flow artifacts are visible, while no flow signal was detected.

Again, these results are in line with expectations that longer echo times provide more time for the flowing spins to leave the imaging slice.

These in vitro results showed that contrast, flow artifacts, and flow suppression depended on the interplay of flow velocity, echo time, and slice thickness. The results suggest, too, that a), the model system is suited for investigating BB MRI under well-controlled conditions, and that b), a 2D RARE sequence is well suited for BB imaging at 7T. Good suppression was obtained by the appropriate choice of TE and slice thickness. For example, for a velocity of 5 cm/s, which is comparably slow (and thus difficult to suppress) with respect to the circulation in rodents, the artifacts were eliminated for slice thicknesses of less than or equal to 0.352 mm for TE = 5 ms, and for echo times larger than 25 ms at a slice thickness of 0.8 mm. T1 contrast was maintained in all cases.

In vivo BB-MRI

Using the insights gained from the in vitro experiments, we set out to optimize the sequence in vivo, using these results as a starting point.

Without apparent deterioration of the flow suppression, we were able to reduce the echo time in vivo further (e.g., to 5 ms in an 800-µm slice, Fig. 6). This effect likely occurred because the blood flow velocity in vivo was larger than the velocities used in vitro (5 cm/s). A shorter TE is beneficial as it resulted in a stronger signal and improved T1 contrast.

Fig. 6.

In vivo T1w BB 2D RARE MRI of a healthy rat brain. T1w and good BB effect (red arrows) were observed without apparent flow artifacts. The resolution, similar to that described above for the in vitro studies, however, was not sufficient for visualizing smaller vessels (arrows, TE = 5 ms, echo train length = 2, TR = 1200 ms, FOV = 45 × 45 mm², voxel size = 0.35 × 0.35 × 0.8 mm³, 15 slices, scan time = 01:16 min).

Imaging the rat brain using a resolution similar to that used in vitro showed a good BB effect. However, the resolution of vessels was limited; this was not apparent in vitro as the flow tube was much larger than the in vivo vessels (Fig. 6, TE=5 ms, echo train length=2, TR=1200 ms, FOV=45 × 45 mm², voxel size=0.35 × 0.35 × 0.8 mm³, 15 slices, scan time=01:16 min). A larger matrix of 256 × 128 and smaller FOV of 40 mm × 20 mm led to a voxel size of 0.156 × 0.156 × 0.8 mm³, good BB effect, improved vessel resolution, good SNR, and a still short scan time of 1:36 min (Fig. 7).

Fig. 7.

In vivo T1w BB 2D RARE MRI of two healthy rat brains. Both animals (a, b) showed T1w and good SNR and BB effect for through-plane (red arrows) and in-plane (blue arrows) vessels without apparent flow artifacts. Scan parameters: TE = 12 ms, echo train length = 8, TR = 750 ms, FOV = 40 × 20 mm², voxel size = 0.156 × 0.156 × 0.8 mm³, scan time = 01:36 min; surface receive coil.

By further adapting the sequence parameters, we could increase the resolution to a pixel size of 0.137 mm × 0.137 mm, while still having a good SNR, BB effect, and no flow artifacts (RARE factor = 8, Fig. 8). Here, vessels of 0.2 mm or more were well resolved. A BB effect was not only observed in through-plane vessels (red arrow), but also for in-plane vessels (not shown).

Fig. 8.

In vivo T1w BB 2D RARE MRI of a healthy rat brain with increased resolution. Good SNR, T1w, and BB effect for through-plane (red) vessels was observed in 3:12 min for 9 slices. Flow artifacts were not apparent. Resolution was good for vessels with a diameter of 0.2 mm (magnification bottom left). Scan parameters: TE = 6 ms, echo train length = 2, TR = 750 ms, FOV = 35 × 35 mm², voxel size = 0.137 × 0.137 × 0.8 mm³, scan time = 03:12 min, RARE factor  = 8; surface receive coil.

In the abdomen, which requires a larger FOV, a strong BB effect and strong T1w were obtained with an echo time of 12 ms (Fig. 9). The T1 contrast was maintained despite a RARE factor of 8. No flow artifacts were apparent, but some chemical shift displacement was found, and motion artifacts were observed close to the heart (not shown). A breathing trigger was successfully used to reduce motion artifacts in the abdomen.

Fig. 9.

In vivo, T1w BB 2D RARE MRI at 7T of the abdomen of a healthy rat. Good SNR, T1w, and BB effect were observed for through-plane (a, red arrow) and in-plane vessels (b, yellow arrow) in 3:12 min. No flow artifacts were apparent, although some chemical shift displacement was found. A breathing trigger was used to reduce respiratory motion artifacts, and cardiac motion artifacts were apparent closer to the heart (not shown, TE = 12 ms, echo train length = 8, TR = 750 ms, FOV = 60 × 60 mm², voxel size = 0.234 × 0.234 × 0.8 mm³, scan time = 3:12 min, 12 slices, RARE factor = 8, volume transmit-receive coil)

The chosen matrix size of 256 × 256 yielded an in-plane resolution of 0.234 × 0.234 mm² and was sufficient to depict major vessels and their walls with reasonable values for scan time (3:12 min), SNR, and scan volume (lateral coverage 5 cm, 12 slices). Vessel-wall enhancement (VWE) was not observed, as no contrast agent was given, and no pathological model was used.

A strong BB effect was also found in a cross-section of a vessel with diameter of 0.5 mm (Fig. 9b)

Discussion

In this work, we explored the properties of a preclinical, T1w, 2D RARE sequence with spectral fat suppression for BB VWI in vitro and in vivo. To this end, we analyzed the flow suppression, flow artifacts, T1 contrast, and SNR of the sequence on a novel model system in vitro and, after optimization, in living rats in vivo.

In vitro setup

Using an in vitro model, we investigated the essential parameters, contrast, artifacts, and flow suppression under well-controlled conditions. In addition, and following the 3R principle, we were able to gain many insights without using live animals [23].

Of course, a model does not fully replace an in vivo experiment, for example, with respect to tissue heterogeneity, motion, vessel size and susceptibility. Still, we were able to analyze and identify parameters that provided good flow suppression, sufficient T1 contrast, and few flow artifacts. While these properties are very important for VWI, we could not assess actual VWE, which is observed in vivo for inflamed vessels after administration and accumulation of contrast agents. To integrate this feature, permeable walls or some tissue and Gd injections could be used in the future.

To investigate the fat suppression, food-grade olive oil was used. Low signal intensity (S_oil) and dark appearance on T1w MRI images indicated that suppression was successful. In vivo, fat suppression may need to be adapted to the scientific question being investigated as it could suppress some lesions. For VWI (and other applications), fat suppression is needed to distinguish fat from areas of contrast agent uptake. Spectral fat suppression is well known to be affected by field inhomogeneities.

To investigate T2w, the signal of agarose gel and water were compared. Both solutions had similar T1, but very different T2 relaxation times. As they appear with similar intensity on all images, strong T1w and little T2w can be assumed. Of course, the non–viscous agarose has different diffusion properties, too, which is not likely to matter here, however. T1w was assessed by comparing a NaCl solution with different concentrations of Gd, a paramagnetic substance that shortens the T1 time of a solution with little effect on T2. A shorter T1 relaxation time enhances the signal in case of T1w MRI [24]. Future model solutions might be found where the T1s are even more similar (while maintaining a difference in T2).

The signal in the flow tube was found to decrease with flow velocity, echo time, and the inverse of the slice thickness. Thus, shorter echo times could be applied for thinner slices, which is beneficial for the desired T1 contrast. In turn, shorter echo times provided more signal, allowing for higher resolution or less averaging. For the parameters investigated, strong flow suppression was already obtained for 5 cm/s. The faster flow velocities in vivo are expected to improve the matters further.

For simplicity, we chose water as flow medium. However, the flow characteristics of water and blood are not the same, which could lead to a more difficult suppression of slow flow. The T1 relaxation times at 7T are similar for water and blood, but T2 is different [25], [26]. Indeed, relaxation times and blood viscosity are also dependant on temperature, oxygenation, pH, and hematocrit concentration [27]. Still, for analyzing the properties of 2D RARE BB MRI, water seems to be suitable.

Using 3D printing for manufacturing, we were able to easily build and modify the MR-compatible model with a size comparable to that of rats. Indeed, contrasts could be exchanged and flow adjusted easily. Further adapting the model to mouse size appears feasible but would require some new design decisions. By adding a “curved” vessel, multidirectional flow could be analyzed in a future implementation of the model.

BB effect (flow suppression) of 2D RARE

The BB effect of a 2D RARE sequence is based on the principle that spins do not contribute to the signal if they move out of the imaging slice before the refocusing pulse is applied. Spins that moved into the slice after the 90° excitation, on the other hand, only experience a 180° pulse, and thus do not provide signal either.

However, in reality, several matters will affect this principle. For example, some blood will remain in the imaging slice during the 90° excitation, refocusing, and readout (e.g., slow or turbulent flow). In this case, the blood in the lumen results in high signal intensity. This effect may occur at the walls of the vessels, where the flow approaches zero. Other spins may be excited and refocused but leave the slice before the readout gradient is applied, so that they still provide signal.

If the excited spins move out of the slice between the excitation and refocusing pulse, no echo will be produced, resulting in a BB effect [28]. Therefore, a BB effect is good if

$$\mathit{velocity}>\frac{\mathit{slice}thickness}{{\mathit{TE}}_{\mathit{eff}}-t_{\mathit{esp}}/2},$$ (1)

where TE_eff is the effective echo time and t_esp is the echo spacing [11]. TE_eff is the time at which the center of k-space is acquired. Nevertheless, the flow will often not be perfectly perpendicular to the slice. This effect maybe approximated by assuming a larger (effective) slice thickness. Thus, three main parameters affect the BB effect here: velocity, slice thickness, and echo time.

Velocity

An increase in velocity (in vitro) improved the BB effect and reduced flow artifacts at constant sequence parameters (of course, in vivo, the velocity cannot be easily modified and is dependent on the vessel diameter and other factors). For a slice thickness of 0.8 mm, an effective echo time TE_eff = 5 ms and an echo spacing t_esp = 5 ms, a velocity of 32 cm/s is needed to clear the slice (perpendicular flow, Eq. (1)). This finding was well in line with our in vitro results, where the signal in the flow tube approached background level at ≈ 30 cm/s. As the echo times are usually longer, flow velocities faster, and the slices thinner, this condition should be met often in vivo. For example, higher blood flow velocities were reported for rats at 35 cm/s [22]. We did not observe apparent flow artifacts while achieving a good BB effect in vivo (Fig. 6, Fig. 7, Fig. 8, Fig. 9).

Slice thickness

Reducing the slice thickness improved the BB effect and reduced the flow artifacts. The intensity of flow artifacts decreased with a reduction in slice thickness, too.

However, reducing the slice thickness also lowers the SNR [29]. Still, thin slices are preferred if scan time and SNR permit. Additionally, for isotropic voxels, thin slices are needed. To achieve a flow suppression for an effective echo time of 5 ms, an echo spacing of 5 ms, and a velocity of 5 cm/s, the slice thickness needs to be smaller than 0.125 mm (Eq. (1)). Experimentally, the smallest slice tested was 0.3 mm, where the artifacts were already greatly reduced (to about twice the noise level, Fig. 4). For in vivo velocities (35 cm/s for rats), the required thickness would be 0.875 mm. Thus, larger slices are possible, which increases the SNR. This estimation was confirmed, as no flow artifacts but a good BB effect was observed in the in vivo images for a slice thickness of 0.8 mm (Fig. 6, Fig. 7, Fig. 8, Fig. 9).

Echo time

Increasing the echo time leads to less signal, more flow suppression, but also more T2w. An increase in T2w is not desirable as VWE requires T1w to show accumulation of Gd contrast agent [24] and the overall loss of SNR (by long echo time) [30]. Thus, we chose short-to-moderate TE to maintain a low T2w in vivo and in vitro. Good flow suppression is expected for echo times longer than 32 ms, for a slice thickness of 0.8 mm, and a velocity of 5 cm/s (Eq. (1)). Experimentally, for the longest echo time tested, 30 ms, we had sufficient flow suppression and only slight flow artifacts (Fig. 5). For in vivo flow velocities (35 cm/s), a good flow suppression is reached for echo times longer than 4.5 ms. Again, this estimation was confirmed as no flow artifacts and good flow suppression were achieved with echo times as low as 5 ms.

Overall, compared to the calculated values for the flow suppression (Eq. (1)). sufficient suppression in vitro was achieved already for thicker slices, smaller velocities, and shorter echo times. This effect may be due to flow-related intravoxel dephasing, which leads to additional flow suppression [11].

Flow artifacts

Ghosting due to periodic motion result in artifacts along the phase-encoding direction regardless of the direction of flow or motion. Depending on the period T of flow and TR there will be multiple ghosts visible along the phase-encoding direction. In case of periodic motion, different lines in k-space will have different phases. The variation in phase is shown in the image as aliasing of the vessel [31].

For some protocols (e.g., TE = 5 ms, slice thickness = 0.8 mm, velocity = 5 cm/s), flow artifacts were prominent in the phase-encoding direction.

In the tube as well as in a vessel, the flow is assumed to be periodic and laminar. This means that the flow consists of multiple layers which move parallel to each other. The velocity is dependent on the position of the layer and increases from the outer layer – friction with the wall decreases the velocity – to the center layer.

Here, the effect of different velocities for moving spins is apparent: the faster-moving spins in the center of the tube produce less signal for the ghosting artifacts than the slower-moving spins at the walls.

In vivo application

The insights gained with the model setup provided a good starting point for in vivo BB experiments. Here, a good BB effect and no flow artifacts were apparent in vessels ranging from 0.25 to 2.52 mm. The high in-plane resolution was sufficient to show brain vessels in a short time (01:36 min, 7 slices), but with thick slices. As the vessel wall is much thinner than the diameter, increased resolution would be necessary for VWI. To obtain a similar image quality for higher resolution, the number of averages should be increased, resulting in longer scan times, however. By decreasing the slice thickness to an isotropic acquisition multiplanar reconstruction could be performed while keeping the in-plane resolution for all three directions. Compared to the results obtained in vitro, faster flow allowed choosing a lower TE without artifacts appearing, and higher resolution was needed to resolve smaller vessels.

The study rats were healthy and did not receive contrast agent; therefore, no Gd-based VWE was observed. Future studies may focus on covering an appropriate portion of the brain while using thinner slices, improving the resolution further, as well as using a pathological model to observe VWE after administering a contrast agent.

Conclusion

BB MRI for rodents at 7 T was found to be feasible using a 2D RARE sequence with optimized parameters with high resolution and short scan times. Using an in vitro model, we were able to investigate key parameters (T1w, BB, artifacts, and fat suppression) under well-defined conditions and without the need to sacrifice animals.

Flow velocity, slice thickness, and echo time affected the flow artifacts, BB effect and T1w. An echo time of 25 ms and a slice thickness of 0.8 mm yielded a good BB effect for a velocity of 5 cm/s without apparent flow artifacts while maintaining T1w in vitro.

In vivo, where flow velocities up to 35 cm/s are expected, good black blood effect without (apparent) flow artifacts were observed for an echo time of 6 ms, slice thickness of 0.8 mm, and an in-plane resolution of 0.137 × 0.137 mm. Longer echo times and thinner slices are expected to improve both BB effect and flow artifacts, but to reduce SNR.

Thus, 2D RARE is a robust, viable method to perform black blood vessel imaging in rodents that is available on virtually all small animal MRI systems.

CRediT authorship contribution statement

Eva Peschke: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Data curation. Mariya S. Pravdivtseva: Writing – review & editing, Software. Olav Jansen: Writing – review & editing, Supervision, Resources, Funding acquisition. Naomi Larsen: Writing – review & editing. Jan-Bernd Hövener: Writing – review & editing, Supervision, Resources, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.