Near-Isotropic Non-Contrast MR Angiography of the Renal and Peripheral Arteries using a Thin-Slab Stack-of-Stars Quiescent Interval Slice-Selective (QISS) Acquisition

Abstract

Purpose:

Non-contrast MR angiography avoids potential risks from gadolinium-based contrast agents. A 2D non-contrast technique, quiescent interval slice-selective (QISS), accurately evaluates the peripheral arteries but has limited spatial resolution along the slice direction. We therefore implemented a prototype thin-slab stack-of-stars version (tsSOS-QISS) with nearly isotropic spatial resolution and tested it in the renal and peripheral arteries of healthy subjects and patients with vascular disease.

Methods:

The study was approved by the hospital institutional review board. A total of 16 subjects were scanned at 1.5 Tesla, 7 for imaging of the renal arteries and 9 for imaging of the peripheral arteries. For tsSOS-QISS of the renal arteries, each slab consistent of ~16 1.3mm or 2.0mm-thick slices (interpolated to 32 0.65mm or 1.0mm-thick 3D partitions) oriented in an oblique axial or oblique coronal view along the length of the target vessel and was acquired in a breath-hold. For tsSOS-QISS of the peripheral arteries, 20 axial overlapping thin slabs were typically acquired, each with 12 1.3mm-thick slices (interpolated to 24 0.65mm-thick 3D partitions). Image quality, vessel sharpness in multiplanar reconstructions, and normalized signal-to-noise ratios (nSNR) were measured.

Results:

Image quality and nSNR in the renal and peripheral arteries were significantly better compared with 2D QISS acquired at the same spatial resolution, while vessel sharpness was improved in multiplanar reconstructions of the renal arteries.

Conclusions:

The tsSOS-QISS technique overcomes a significant limitation of 2D QISS by providing nearly isotropic spatial resolution with improved image quality, nSNR and vessel sharpness in multiplanar reconstructions.

Introduction

Magnetic resonance angiography (MRA) and CT angiography are routinely used to evaluate vascular pathology over a wide range of body regions and clinical indications (1–3). While contrast-enhanced MRA techniques are efficient and accurate, the administration of gadolinium-based contrast agents is generally avoided in patients with end stage kidney disease due to the potential risks of nephrogenic systemic fibrosis and gadolinium deposition in the brain and other tissues (4–6), with as yet undetermined long-term health consequences. In recent years, a variety of non-contrast alternatives to contrast-enhanced MRA have been proposed including time-of-flight (7,8), subtractive three-dimensional (3D) fast spin-echo (9) and balanced steady-state free precession (bSSFP) (10,11), velocity-selective magnetization-prepared 3D bSSFP (12,13), and quiescent-interval slice-selective (QISS) (14–17).

As usually implemented for vascular applications in the body, QISS is an inflow-dependent two-dimensional (2D) technique that uses a bSSFP readout. It relies on the application of an in-plane saturation or inversion radiofrequency (RF) pulse to suppress background signal, which is followed by a quiescent interval that coincides with rapid systolic flow to maximize refreshment of saturated in-plane spins before each readout. Advantages of 2D QISS with respect to 3D non-contrast alternatives include faster acquisition speed, breath-hold capability, insensitivity to patient motion, and excellent arterial conspicuity even in the setting of slow flow. However, there are disadvantages as well. For instance, unlike the case with 3D, 2D techniques like QISS do not benefit from the intrinsic signal averaging effect from collecting multiple 3D partitions. Persistent out-of-slice steady-state magnetization with a bSSFP readout can cause artifacts from rapid blood flow when very thin slices are used as well as from off-resonance effects (18). Given these factors, the image quality when acquiring very thin slices can be suboptimal. Moreover, even if nominally thin slices are acquired, the non-rectangular 2D bSSFP slice profile causes blurring in multiplanar reconstructions.

In contrast to 2D QISS, 3D bSSFP-based techniques allow the acquisition of very thin slices with nearly isotropic voxels for non-contrast MRA of the renal arteries (19,20), coronary arteries (21), pulmonary veins (22), and other vascular territories. A drawback of such techniques is that they use a thick-slab RF excitation that limits inflow refreshment and may result in poor conspicuity of arteries situated deep within the 3D volume. Saturation effects are further exacerbated when a saturation or inversion RF pulse is applied over the large 3D volume to suppress background tissue signal, as in the case of free-breathing non-contrast renal artery MRA. Moreover, scan times are lengthy (on the order of several minutes or longer), requiring the use of respiratory gating for chest and abdominal applications. While scan times can be substantially shortened using a non-Cartesian k-space trajectory in conjunction with acceleration techniques such as compressed sensing (23), respiratory gating is still required.

Instead of a single thick slab, a multiple overlapping thin-slab acquisition (MOTSA) (24,25) has proved advantageous for 3D time-of-flight MRA of the head and neck. MOTSA uses a low flip angle excitation in conjunction with a tilted optimized non-saturating excitation (TONE) RF pulse (26). However, MOTSA is not optimal for body MRA applications in part due to the use of a fast-low angle shot (FLASH) readout, which predisposes the technique to signal loss from in-plane flow saturation and provides only a modest signal-to-noise ratio (SNR).

In order to obtain the SNR and spatial resolution benefits of a volumetric acquisition, while maintaining the high arterial conspicuity, motion insensitivity and breath-hold capability of 2D QISS, we propose using a thin-slab stack-of-stars implementation of the QISS MRA technique (tsSOS-QISS). The feasibility of this technique was tested in two vascular territories: (1) lower-extremity peripheral arteries (where arterial flow is primarily directed through-plane) and (2) renal arteries (where arterial flow is primarily directed in-plane).

Methods

The study was approved by the hospital institutional review board (IRB). All subjects provided written informed consent or were scanned for clinical purposes with consent waived retrospectively by the IRB. Imaging was performed using a 1.5 Tesla MRI system (MAGNETOM Avanto, Siemens Healthcare, Erlangen, Germany) with phased array peripheral, body and flex coils. A total of 16 subjects (ages 28 – 76 years old, 14 male) were scanned, 7 for imaging of the renal arteries and 9 for imaging of the peripheral arteries. The 16 subjects included 10 healthy volunteers and 6 patients with suspected renal artery or peripheral arterial disease.

Thin-slab stack-of-stars technique:

Two prototype versions of the tsSOS-QISS pulse sequence were implemented. The initial version, used in all 7 renal artery studies and 7 of 9 peripheral artery studies, is similar in structure to an ECG-gated 2D radial QISS pulse sequence (27), except that it includes a phase-encoding loop along the 3D partition-encoding direction. Separate RF pulses are applied sequentially for in-plane and venous signal suppression followed by a quiescent interval spanning the period of rapid systolic flow to facilitate inflow refreshment. Whereas venous signal suppression in the 2D QISS sequence is obtained using a sinc saturation RF pulse, with tsSOS-QISS it is obtained using a frequency offset corrected inversion (FOCI) RF pulse (28). The venous inversion is 50mm thick, applied with a 5mm gap inferior to the imaging slab.

As with 2D QISS, in-plane signal suppression with tsSOS-QISS is obtained using a sinc-shaped saturation RF pulse with a thickness twice that of the imaging slab. The slab thickness used for peripheral artery MRA is 15.6mm with in-plane saturation thickness of 31.2mm, which may lead to inadvertent saturation of inflowing spins. To address this limitation, a second sequence version, used in 2 subjects for imaging of the peripheral arteries, was developed to reduce inflow-related saturation effects. In this version, the in-plane saturation RF pulse is omitted, and both in-plane and venous signal suppression are obtained via the FOCI RF pulse, with its thickness increased to 65mm and location shifted so that the top edge of the inversion region sits just below the top edge of the 15.6mm-thick imaging slab (Figure 1).

1.

Illustration of acquisition schemes using 2D QISS and tsSOS-QISS for peripheral MRA (showing three 2D slices or 3D slabs). A) With 2D QISS, a tracking venous saturation region is applied below a thin slice, followed immediately by application of an in-plane saturation region having twice the slice thickness. After a quiescent interval of several hundred msec, data are acquired in a single shot using a bSSFP readout. The slice (along with the in-plane and tracking saturation regions) is then shifted superiorly and the process is repeated. B) With the initial implementation of tsSOS-QISS, a tracking venous inversion is applied just below the 3D slab, followed immediately by application of an in-plane saturation region having twice the slab thickness. Data are subsequently acquired as with 2D QISS, except that a stack-of-stars k-space trajectory is used. The 3D slab (along with the in-plane and tracking saturation regions) is then shifted superiorly by one-half the slab thickness and the process is repeated. C) With the optimized implementation of tsSOS-QISS, the in-plane saturation region is not applied. Instead, suppression of venous signal as well as in-plane background signal is obtained using a single inversion region that tracks with the imaging slab.

The time from the application of the in-plane saturation pulse to the center of k-space (TI) for both 2D QISS and tsSOS-QISS typically ranged from approximately 500ms to 700ms depending on heart rate. A bSSFP readout was used to image the peripheral arteries, whereas a variant of bSSFP, fast interrupted steady-state (FISS) using 5 to 8 sequence repetitions per FISS module (29), was preferred for the renal arteries due to the excellent fat suppression and resultant improvement in arterial conspicuity and reduction of radial streak artifacts. The number of radial views for 2D and tsSOS-QISS was typically ~98 compared with ~92 views for Cartesian QISS. Typical bSSFP echo time was ~2ms with repetition time of ~4ms. The excitation flip angle ranged from 70 to 90 degrees. Scan time for tsSOS-QISS was reduced using a partial Fourier factor of 6/8 or 7/8 along the 3D partition direction. For instance, assuming an RR interval of 1 sec, 12 3D partitions, 50% slice oversampling, and 6/8 partial Fourier along the 3D partition direction, the scan time for each tsSOS-QISS slab will be 14 sec.

For imaging of the renal arteries, 2D radial QISS and tsSOS-QISS were each acquired in a breath-hold of ~14 to 20 sec (depending on heart rate), typically in tilted axial and coronal views along the length of the target vessel. With 2D radial QISS, ~16 slices using slice thickness of 1.3mm or 2.0mm were acquired in each breath-hold. With tsSOS-QISS, ~16 3D partitions were acquired in each breath-hold, yielding 32 0.65mm- or 1.0mm-thick partitions after interpolation. Free-breathing inversion-prepared inflow-dependent 3D non-contrast MRA was acquired in an axial orientation using a standard clinical protocol with a cross-pair navigator placed over the right hemi-diaphragm. The inversion slab was 150mm thick and was positioned so that the top edge extended just above the highest extent of the renal arteries.

Unlike the renal arteries, where only a single 3D slab is acquired for each tsSOS-QISS scan, peripheral artery imaging involves the acquisition of multiple overlapping 3D slabs. Consequently, venetian blind artifact (VBA) at the slab junctions, which is visible in multiplanar reconstructions (MPRs), becomes a significant concern. Based on initial phantom and volunteer studies, VBA was minimized by increasing the excited tsSOS-QISS slab thickness by a factor of 1.5 from 15.6mm to 23.4mm, overlapping adjacent 3D slabs by 50%, and using slice oversampling of 50% or 66.7%. Only the top 12 (of 24 interpolated) slices of each slab were used to create coronal MPRs (to ameliorate flow saturation effects that worsen as the arterial spins progress caudally through the slab). In all cases, MPRs were created using a maximum intensity projection (MIP) algorithm.

For peripheral artery imaging, an initial survey examination was performed using single-shot 2D QISS (1 slice acquired per RR interval) with contiguous 3.0mm-thick axial slices. This was followed by axial thin-slice 2D Cartesian QISS and tsSOS-QISS through the popliteal trifurcation vessels, with each sequence spanning a region approximately 164mm in cranio-caudal extent. 2D QISS acquisition used 1.3mm-thick slices with either 20% overlap (158 slices) or 50% overlap (252 slices). For tsSOS-QISS of the popliteal trifurcation vessels (6 subjects), 20 axial overlapping thin slabs were acquired, each with 12 1.3mm-thick slices (interpolated to 24 0.65mm-thick 3D partitions). Scan time was dependent on heart rate. For instance, assuming an RR interval of 1 sec, equal head-to-foot coverage and slice thickness, total scan time for 2D QISS with 20% overlap and 158 slices would require 158 sec, compared with 280 sec for tsSOS-QISS using 20 slabs. For tsSOS-QISS of the aorto-iliac bifurcation and pelvic vessels (2 subjects), 36 axial overlapping thin slabs were acquired.

Objective Image Quality Analysis

Measurements were performed in 7 renal artery studies and 6 peripheral artery studies (popliteal trifurcation vessels only). Our approach for measuring SNR was an approximation of Firbank et al. (30), modified to avoid the time-consuming requirement to repeat each 3D acquisition. The SNR was estimated by computing the mean signal of the vessel in each slice divided by the standard deviation of signal in the difference image of the given slice and an immediately adjacent slice in a region near the vessel, multiplied by √2. Normalized SNR (nSNR) was calculated by correcting for voxel dimensions according to the relation nSNR = SNR / V, where V denotes the acquired voxel size in mm³.

Subjective Image Quality Analysis:

Image quality for all normal and diseased cases was assessed individually by two radiologists (E.A. and R.R.E.) based on anonymized MIP images. All acquisitions were de-identified and evaluated by the observers in random order. Observers were not aware of the type of acquisition presented. Per-segment image quality was subjectively rated independently by each observer according to a 4-point scale: (1) vascular anatomy not assessable due to severe image artifacts and/or poor vascular signal, image quality inadequate for diagnosis; (2) vascular anatomy assessable despite moderate image artifacts and/or decreased vascular signal, marginally acceptable image quality for diagnosis; (3) good image quality with minor artifacts and/or relatively homogenous vascular signal, adequate for confident diagnosis; and (4) excellent image quality without artifacts and homogenous vascular signal, highly confident diagnosis. In addition, an orthogonal MPR was paired with a MIP of the source images and the two images were viewed simultaneously to assess the impact of acquisition technique on blurring in the MPR. In comparison with the source MIP, which was considered the reference, image sharpness for the renal MPR data was rated on a 4-point scale ranging from (1) vessel margins severely blurred to (4) vessel margins sharply delineated.

Statistical Analysis:

Data were analyzed using PSPP software (The GNU Project, version 1.2.0). Continuous SNR data were compared using paired t-tests. Diagnostic image quality ratings for the 2D and SOS-QISS techniques were compared using Wilcoxon signed-rank tests.

Results

In the lower-extremity peripheral arteries, tsSOS-QISS provided superior image quality to 2D QISS (Figures 2 and 3) acquired with either thin (≤1.5mm) or standard survey (3.0mm) slice thickness. Median image quality scores for tsSOS-QISS versus 2D QISS were significantly different for rater 1 (4 versus 3, P<0.05), and trended higher for rater 2 (3.5 versus 3, P=0.16). tsSOS-QISS provided a 4.2-fold improvement of nSNR (24.7 versus 5.9, P<0.01) with respect to 2D QISS. Both VBA and flow saturation effects with tsSOS-QISS were minimized by proper selection of imaging parameters (Figure 2A). However, using an inadequate amount of slice oversampling or too thin a 3D slab increased the prominence of VBA (Figures 2C and D). Flow saturation effects were more prominent with a thicker 3D slab (Figure 2E). Without any RF pulses applied for background signal suppression, flow saturation was ameliorated but at the cost of increased background signal from soft tissues and superficial veins (Figure 2F).

2.

Coronal MPRs of the popliteal trifurcation arteries in a healthy subject showing impact of imaging parameters on venetian blind artifact (VBA) using axial tsSOS-QISS with overlapping slabs and 1.3mm-thick slices (interpolated to 0.65mm). (A) Using optimized imaging parameters (50% slab overlap, 66.7% slice oversampling, slab thickness = 23.4mm), there is excellent delineation of the vessels with only minimal VBA. SNR is improved compared with 2D QISS (B) acquired with the same slice thickness. (C) With either a reduction in slice oversampling to 33.3% (C) or a reduction in slab thickness to 15.6mm (D), VBA becomes more pronounced. (E) With doubling of the slab thickness and corresponding increase in slice thickness to 2.6mm, there is loss of signal due to flow saturation in the distal segments of the peroneal and posterior tibial arteries. (F) Using the same technique as (E) without in-plane saturation, signal loss from flow saturation is avoided but background signal from soft tissues and superficial veins is substantially increased.

3.

Sequence comparisons in the peripheral arteries. (A) Healthy subject, coronal MPRs. Top left: Axial 2D QISS using 3mm-thick slices. Top right: Thin-slab 3D Cartesian QISS using 0.65mm-thick slices. Bottom left: tsSOS-QISS using 0.65mm-thick slices. Both 3D Cartesian and tsSOS-QISS provide more vascular detail than 2D QISS. However, compared with 3D Cartesian, tsSOS-QISS shows more uniform intravascular signal and less venetian blind artifact. Bottom right: tsSOS-QISS without in-plane saturation, using overlap of the FOCI inversion region with the imaging slab for combined venous and background signal suppression. Compared with using a thick in-plane saturation for background suppression, this approach demonstrates more uniform intravascular signal intensity with improved branch vessel detail (arrows).

B) Same subject as (A). Oblique coronal MPR from tsSOS-QISS of the aorto-iliac vessels with 0.65mm-thick slices and 36 slabs (right, scan time ~8min 45s) shows markedly improved branch vessel detail compared with 2D QISS using 3mm-thick slices (left, scan time ~2min 30s). Axial tsSOS-QISS source image (inset) shows negligible streak artifact from radial undersampling due to the high degree of background signal suppression with the QISS technique.

C) 76-year-old male with left leg pain, coronal MPRs. Left: Axial 2D QISS using 3mm-thick slices. Right: Axial ts-SOS-QISS MRA using 0.65mm-thick slices shows improved SNR and better delineation of small branch vessels such as a genicular artery (arrow).

In 2 subjects in whom tsSOS-QISS and thin-slab 3D Cartesian QISS were compared, image quality was overall similar. However, image quality was slightly worse with the Cartesian acquisition due to greater intravascular signal heterogeneity and increased venetian blind artifact (Figure 3A). Eliminating the in-plane saturation RF pulse and instead using a single inversion region that simultaneously suppressed signal from venous flow and stationary tissues within the imaging slab was found to improve intravascular signal homogeneity and conspicuity of small branch vessels (Figures 3A and B). tsSOS-QISS provided excellent depiction of small calf arteries in patients with severe multi-vessel PAD (Figure 3C).

For imaging of the renal arteries, both the thin-slice 2D QISS and tsSOS-QISS scans generally provided good-to-excellent image quality. However, branch vessel detail was superior with tsSOS-QISS, while MPRs were significantly less blurred compared with MPRs from 2D QISS acquired at the same slice thickness (Figures 4 and 5). Wilcoxon signed-rank tests revealed that median image quality provided by tsSOS-QISS was superior to that of 2D QISS (4 versus 3 for rater 1, P<0.05; 4 versus 3 for rater 2, P<0.05) and improved nSNR by a factor of 2.1 (19.9 for tsSOS-QISS versus 9.3 for 2D QISS, P<0.05). Median scores for MPR image sharpness for tsSOS-QISS and 2D QISS were 4 versus 3 (P<0.05) for rater 1, and 4 versus 2 for rater 2 (P<0.05). Although not formally assessed, vessel depiction using breath-hold tsSOS-QISS appeared comparable to much lengthier free-breathing inflow-dependent 3D bSSFP scans.

4.

Healthy volunteer. (A) MIP from breath-hold oblique axial 2D QISS using 2mm-thick slices. (B) Oblique coronal MPR shows a moderate degree of vessel blurring. (C) MIP from breath-hold oblique coronal 2D QISS for comparison with (B). (D) MIP from breath-hold oblique axial ts-SOS-QISS using 2mm-thick slices (interpolated to 1mm). (E) Oblique coronal MPR shows improved vessel detail compared with (B). (F) MIP from breath-hold oblique coronal tsSOS-QISS provides excellent depiction of the intrarenal arterial branches. (G) MIP from axial navigator-gated inversion recovery 3D bSSFP. Scan time was 3min 38s compared with 16.5s for breath-hold 2D QISS and 18.8s for tsSOS-QISS. (H) Oblique coronal MPR provides similar arterial depiction to (F).

5.

67-year-old female with peripheral arterial disease. (A) CTA depicts a moderately severe proximal right renal artery stenosis (arrow). (B) MIP from axial navigator-gated inversion recovery 3D bSSFP (scan time = 2min 49s). (C) Corresponding coronal MPR. (D) MIP from breath-hold oblique axial 2D QISS MRA. (E) Oblique coronal MPR. (F) MIP from breath-hold oblique coronal 2D QISS MRA. (G) MIP from breath-hold oblique axial ts-SOS-QISS MRA. (H) Corresponding oblique coronal MPR. (I) MIP from breath-hold oblique coronal tsSOS-QISS. The renal artery stenosis is accurately depicted by tsSOS-QISS with improved detail compared with 2D QISS, despite the acquired slice thickness (2mm) being the same for both techniques.

Discussion

For the non-contrast evaluation of the lower-extremity peripheral arteries, 2D QISS MRA has proven to be an efficient and accurate technique (31). Nonetheless, a recognized limitation is the limited spatial resolution along the slice direction (32). To overcome this limitation, a prototype thin-slab stack-of-stars QISS MRA technique with nearly isotropic spatial resolution was implemented, providing consistently better image quality than 2D QISS for the renal and peripheral arteries. In the renal arteries, branch vessels were more conspicuous, the sharpness of MPRs was significantly improved compared with 2D QISS at the same spatial resolution, while normalized SNR was significantly higher. In the peripheral arteries, vascular detail was also improved, even in the small calf arteries of patients with multi-vessel PAD.

In the standard implementation of the QISS technique, an inferior saturation or inversion RF pulse is applied to suppress signal from venous spins while an in-plane sinc-shaped RF pulse with twice the thickness of the slice is applied to ensure uniform background suppression and maximize arterial conspicuity. For 2D QISS, the in-plane saturation only affects a narrow 6mm-thick region (assuming a 3mm-thick slice) so that inadvertent saturation of inflowing spins is negligible. Our initial implementation of tsSOS-QISS used this same pulse sequence structure, overall with excellent image quality. However, with a 3D acquisition, the thick in-plane saturation region extends well above the top margin of the imaging slab, an undesirable phenomenon that is exaggerated by the poor slab profile produced by the sinc-shaped saturation RF pulse. As a result, there can be inadvertent signal suppression of inflowing arterial spins to the detriment of MRA image quality. We therefore created an improved version of the tsSOS-QISS sequence that repurposes the FOCI RF pulse, originally only used for venous suppression, to suppress both venous and in-plane background signal. Since the suppression region from the FOCI RF pulse does not extend above the top of the imaging slab and has a much sharper profile than that of the original sinc in-plane suppression RF pulse, inadvertent saturation of inflowing arterial spins is minimized, and arterial conspicuity enhanced.

A major consideration when using a multiple overlapping thin-slab 3D MRA technique is the need to minimize VBA, which is caused by the non-rectangular slab profile (33). For MOTSA MRA of the head and neck vessels using a FLASH readout, a long echo spacing does not worsen image artifacts and can in fact be advantageous with respect to flow saturation effects. Moreover, the low flip angle ensures that specific absorption rate is not a concern. These features provide considerable flexibility with respect to optimizing the length and time-bandwidth product of the RF excitation pulse to achieve a more rectangular slab profile, while also facilitating the use of a TONE RF pulse that minimizes flow-related saturation effects.

In contrast, the bSSFP readout used for QISS allows much less flexibility in RF pulse design. For instance, RF pulse duration must be minimized to maintain a sufficiently short echo spacing to avoid flow and off-resonance artifacts, while the large flip angle of the RF excitation increases the specific absorption rate enough to become a limiting factor. Other problems with the use of a bSSFP readout are that TONE RF pulses are incompatible and that steady-state and off-resonance effects worsen the slab profile, especially with the use of a large flip angle (34,35).

To reduce VBA to an acceptable level for tsSOS-QISS in the peripheral arteries, we used a substantial degree of slice oversampling and overlap of adjacent slabs. The use of these imaging parameters substantially decreased scan efficiency compared with 2D QISS. For instance, comparing tsSOS-QISS acquired with 50% slice oversampling and 50% slab overlap to 2D QISS acquired with 20% slice overlap, scan time is approximately 77% longer. Nonetheless, even correcting for the impact of scan time, normalized SNR for tsSOS-QISS in the peripheral arteries remains more than three-fold higher than for 2D QISS. Moreover, with further optimization of the slab profile and other imaging parameters, we anticipate that scan efficiency can be improved. For instance, recent work suggests that scan efficiency can be substantially increased by reducing the percentages of both the slice oversampling and slab overlap from 50% to 25%, in conjunction with reducing the slab thickness multiplier from 1.5 to 1.2.

For renal artery imaging, only a single tsSOS-QISS slab is acquired in each breath-hold, so that VBA is not a concern. Although respiratory-gated inflow-dependent 3D bSSFP techniques have already proven accurate for evaluation of the renal arteries (20), tsSOS-QISS may have potential utility as a more rapid breath-hold adjunctive or alternative approach. In our cases where breath-hold tsSOS-QISS and respiratory-gated 3D bSSFP were both obtained, the renal arteries were well depicted by both techniques. The breath-hold tsSOS-QISS technique could be helpful in situations where the patient’s respiratory pattern is too irregular to permit adequate respiratory gating. Moreover, a breath-hold acquisition provides the flexibility to directly image a renal artery stenosis in cross-section, or to rapidly evaluate accessory arteries that are outside of the region encompassed within the axial thick slab of the free-breathing acquisition. Another potential benefit of the tsSOS-QISS technique could be in patients where flow is too slow (e.g. distal to a hemodynamically critical stenosis) to permit visualization of the more distal portions of the renal arteries using a free-breathing thick-slab 3D acquisition. With tsSOS-QISS, unsaturated spins from the aorta typically need to travel a much shorter distance through the renal artery than with the free-breathing technique where a thick inversion slab has been applied, so saturation effects should be reduced. One also has the option of imaging the various renal artery segments in separate breath-holds to minimize flow saturation effects. However, further study is needed to determine the benefits and drawbacks of the breath-hold approach.

Limitations:

Potential drawbacks of the thin-slab stack-of-stars QISS technique include: (1) reduced scan efficiency compared with 2D QISS for peripheral MRA. Consequently, the technique is best suited for the detailed evaluation of a limited number of vessels segments rather than for a complete survey examination, where 2D QISS using thicker (e.g. 3mm) slices excels. Potentially, scan efficiency can be improved using parallel imaging along the slice direction, a continuous moving table acquisition (36), or by adapting techniques specifically directed to reducing VBA such as SLINKY (37). (2) In the setting of very slow flow (as in a small caliber arterial branch or distal to a flow-limiting stenosis), signal loss from flow saturation would be expected to be more severe than with 2D QISS, particularly when flow is directed obliquely.

Conclusions:

The tsSOS-QISS technique overcomes a significant limitation of 2D QISS for evaluation of the renal and peripheral arteries by providing nearly isotropic spatial resolution with improved image quality, nSNR and vessel sharpness in multiplanar reconstructions. Moreover, it may have potential utility as a rapid breath-hold adjunctive or alternative approach to free-breathing non-contrast renal MRA. Further investigation is needed to determine accuracy and clinical utility.