Accuracy of non-contrast quiescent-interval single-shot and quiescent-interval single-shot arterial spin-labelled magnetic resonance angiography in assessment of peripheral arterial disease in a diabetic population

Abstract

Introduction:

Diabetic patients with peripheral arterial disease (PAD) are challenging to assess. Non-contrast magnetic resonance angiography (MRA) offers a safe alternative in patients with renal impairment. The study objective is to evaluate accuracy of lower limb quiescent-interval single-shot (QISS) MRA and pedal QISS-arterial spin-labelled (ASL) MRA for detection of significant stenosis in diabetic patients with PAD.

Methods:

Combined QISS and QISS-ASL MRA was performed in 32 diabetic PAD patients (20 male, 12 female; mean 69 years; 8 with critical ischaemia). Two readers assessed haemodynamically significant (>50%) stenosis and diagnostic confidence on MRA, against digital subtraction angiography (DSA) as the reference standard, with subgroup analysis of patients with severe renal impairment (n = 7). Inter-reader agreement of stenosis and diagnostic confidence were evaluated. Test–retest reproducibility was evaluated in 10 subjects who underwent repeat MRA on a different day.

Results:

At DSA, 262/645 segments (40.6%) had haemodynamically significant stenoses. MRA accuracy was 78.1% (478/612) and 75.6% (464/614), sensitivity 64.7% (161/249) and 77.5% (193/249), and specificity 87.3% (317/363) and 74.2% (271/365) for 2 readers. MRA accuracy was 80.9% and 80.7% for readers 1 and 2, respectively, in patients with severe renal impairment. QISS MRA but not pedal QISS-ASL MRA was considered of diagnostic image quality. Inter-reader agreement was moderate for stenosis (K = 0.60) and diagnostic confidence (K = 0.41). Test–retest reproducibility was high (K = 0.87) and moderate (K = 0.54) for individual readers.

Conclusions:

Quiescent-interval single-shot MRA has reasonable accuracy in a diabetic PAD population with high burden of disease, providing a non-contrast option in patients with renal impairment. QISS-ASL MRA requires further optimisation to be clinically feasible.

Introduction

Type 2 diabetes is a strong risk factor for peripheral arterial disease (PAD) and contributes to worse outcomes.¹ Typically in diabetics, atherosclerotic lesions are long-segment, multilevel, and involve the infragenual arteries.² Imaging is vital to inform management of critical limb ischaemia (CLI). Lesion characteristics and identification of suitable targets, including pedal vessels if no other bypass targets exist,^3,4 aid in determining endovascular or open surgical intervention.¹

Duplex ultrasonography (DUS) is non-invasive, but challenging in patients with large body habitus or heavily calcified vessels, and does not provide an easily visualised vascular roadmap. Computed tomographic angiography (CTA), gold standard digital subtraction angiography (DSA) and contrast-enhanced magnetic resonance angiography (MRA) provide excellent anatomic detail. However, CTA is limited by vessel wall calcification, DSA is invasive and both involve ionising radiation and typically iodinated contrast, while contrast-enhanced MRA requires gadolinium contrast. In the diabetic population where nephropathy is prevalent, there is a risk of iodinated contrast-induced-nephropathy (CIN),⁵ or nephrogenic systemic fibrosis (NSF) in the setting of MRA.⁶

Therefore, an unenhanced MRA technique is attractive. Time-of-flight MRA is well-established for neurovascular imaging, but difficult in the lower extremities due to long acquisition times, respiratory and bowel motion in the abdomen and pelvis, and relatively slow distal arterial flow.⁷

Quiescent-interval single-shot (QISS) MRA is an unenhanced electrocardiograph (ECG)-gated two-dimensional (2D) balanced steady-state free precession (bSSFP) technique⁸ that demonstrates high accuracy against contrast-enhanced MRA⁹ including in diabetes.¹⁰ It has performed well in CLI¹¹ and compared favourably to CTA in a prior study,¹² particularly for the below knee vessels. QISS-arterial spin-labelled (ASL) MRA is a related subtraction-based technique that has potential for pedal imaging due to theoretically improved background suppression.¹³

The primary objective of this study was to assess accuracy, sensitivity and specificity of combined QISS MRA and QISS-ASL MRA for lower extremity artery stenosis in a diabetic population with symptomatic PAD, including patients with renal failure, against DSA as the reference standard. A combined protocol has not previously been assessed to our knowledge. Secondary objectives were to assess diagnostic confidence of QISS MRA and QISS-ASL MRA, inter-reader agreement and test–retest reproducibility.

Methods

Study design

Single-centre prospective study with recruitment over an 18-month period, with institutional ethics approval and informed consent obtained from all patients, all of whom were adult (>18 years) diabetics with PAD referred for DSA. Exclusion criteria were contraindications to magnetic resonance imaging (MRI). MRA was performed within 5 weeks prior to DSA (mean 5 days, range 0–35 days). Based on Hodnett et al., recruitment of 30 diabetic patients was estimated to have 80% power to detect haemodynamically significant stenosis with minimum 85% sensitivity and 89% specificity, assuming equivalent performance in patients with renal failure.⁹ Two patients represented with new symptoms (Patient 3, 71-year-old female first presented with left leg claudication, and six months later with right leg claudication; Patient 22, 66-year-old female first presented with non-healing left 3rd toe ulcer, underwent anterior tibial artery and dorsalis pedis angioplasty and 5 months later represented with a gangrenous toe), were re-enrolled and considered as new subjects. Therefore, the final cohort consisted of 32 subjects (male n = 20, female n = 12, mean 69 years, range 46–91 years, median body mass index (BMI) 31.7 kg/m²). This included patients with stage 4 or 5 chronic kidney disease (CKD) (eGFR < 30 mL/min/1.73 m²) n = 7, CLI n = 8 and non-sinus cardiac rhythm n = 8 (Table 1). The first 10 consecutive patients underwent repeat MRA within 58 days of initial MRA (mean 24 days, range 6–58 days), and within 56 days of DSA (n = 3 performed pre-DSA, mean 20 days, range 17 days pre-DSA – 56 days post-DSA). Figure 1 summarises the study design.

Table 1.

Patient demographics

Age (years)
Range	46–91
Average	69.2
Median	71
Gender (n)
Male: Female	20: 12
BMI† (kg/m2)
Range	21–43
Average	31.4
Median	31.7
Diabetes type (n)
Type 1	1
Type 2	31
Renal function (eGFR, mL/min/1.73 m2)
Range	7–91
Average	52
Median	51.5
Stage 4 or 5 CKD‡(<30)	n = 7
Reason for presentation (n)
Claudication	21
Rest pain	2
Tissue loss (ulceration)	6
Cellulitis	1
Asymptomatic stenotic graft	1
Enlarging popliteal aneurysm	1
Endovascular treatment at DSA (n)
Angioplasty	5
Stent insertion	1
Heart rate (beats/min)
Range	43–108
Average	67
Cardiac rhythm (n)
Sinus rhythm	24
Ectopics	5
Tachycardia	1
Atrial fibrillation	1
Arrhythmia not specified	1

^†Body mass index.

^‡Chronic kidney disease.

Fig. 1.

Study design. *Includes second clinical presentations for patients 3 and 22 following treatment for the first presentation. **Repeat MRA was performed in a subset of patients (n = 10); was performed prior to DSA in n = 3 and following DSA in n = 7 for logistic reasons, with exclusion of any intervened upon segments and segments downstream of these at statistical analysis.

Imaging

MRA was performed at 1.5T (Avanto, Siemens Healthineers, Erlangen, Germany).

QISS-ASL MRA of the feet was performed first. Two data sets are acquired with this technique: the first employs an in-plane 90-degree saturation RF pulse to suppress all non-arterial signal and the second a B1-insensitive non-selective RF pulse to suppress all tissue signal. Subtraction of the two data sets leaves only arterial signal.¹³ A 12-channel head coil was used for signal reception, with foam cushion knee support and vacuum cushion around the feet for immobilisation.

Quiescent-interval single-shot MRA of the lower extremities from renal arteries to ankles was then performed with peripheral, spine and two 6-channel body phased array coils. Briefly, a slice-selective RF pulse is applied to set longitudinal magnetisation of tissues within the slice to zero. A tracking saturation RF pulse is applied caudally to saturate venous inflow. A quiescent interval follows that includes systole to maximise inflow. A single-shot 2D bSSFP readout is performed in diastole when flow is minimal or absent [9]. Additional thin-slice QISS MRA of the calf images was obtained. Assuming a typical heart rate of 60 beats per minute, total scan time was 20 min (QISS-ASL MRA 8.5 min and QISS MRA 11.5 min). Imaging parameters are summarised in Table 2.

Table 2.

QISS MRA and QISS-ASL MRA parameters

Parameter	QISS MRA (Full lower limbs)	QISS MRA (Calf)	QISS-ASL MRA (Feet)
Acquisition time	432 RR intervals	256 RR intervals	256 RR intervals × 2
TR† (ms)	3.5	3.7	3.7
TE‡ (ms)	1.4	1.6	1.6
Flip angle	90	90	90
Field of view (mm)	400 × 260	400 × 240	400 × 240
In-plane resolution (mm2)	1 × 1	1 × 1	1 × 1
Slice thickness (mm)	3	1.2	1.2
Number of slices	48	128	128
Stations	9	2	2
GRAPPA§ imaging factor	2	2	2
Quiescent interval (ms)	350	350	228
Bandwidth (Hz/pixel)	658	658	658

^†Repetition time.

^‡Echo time.

^§Generalised Autocalibrating Partial Parallel Acquisition.

Digital subtraction angiography was performed by an experienced interventional radiologist or vascular surgeon. Access was obtained via common femoral arterial puncture and 5- to 7-French sheath. Depending on clinical indication, combinations of distal aortic and lower limb angiograms were performed with rate and volume of injections according to institutional protocol, and endovascular intervention performed (n = 5 angioplasty, n = 1 stent insertion). Iodixanol (Visipaque, GE Healthcare, Rydalmere, Australia) was routinely used (mean 88 mL, range 20–235 mL, n = 28). Four patients underwent carbon dioxide (CO₂) DSA (iodine allergy n = 1, eGFR < 30 mL/min/1.73 m² n = 3). At least one standard view (anteroposterior, lateral or oblique) of the pelvis, lower limb or foot was obtained.

Image analysis

Two radiologists with ten and four years’ MR vascular experience independently analysed anonymised MRA images on a PACS workstation (IMPAX Version 6.5.2, AGFA Healthcare, NV, Belgium), blinded to clinical information. Subtraction data sets for QISS-ASL MRA and maximum intensity projections (MIPs) in sagittal and coronal views were also provided for reference, generated on an independent workstation by an experienced MRI technologist (Multimodality Workplace, Siemens, Erlangen, Germany).

A maximum of 47 segments were evaluated in four regions per subject: pelvis (infrarenal aorta, common iliac artery, internal iliac artery origin and external iliac artery), thigh (common femoral artery, profunda femoris origin, proximal superficial femoral artery (SFA), distal SFA, above-knee popliteal artery), calf (below-knee popliteal artery, horizontal segment of anterior tibial artery (ATA), proximal vertical segment of ATA, distal vertical segment of ATA, tibioperoneal trunk, proximal posterior tibial artery (PTA), distal PTA, proximal peroneal artery, distal peroneal artery) and foot (dorsalis pedis artery, medial plantar artery, lateral plantar artery, pedal arch, metatarsal arteries). Proximal and distal SFA and tibioperoneal arteries were divided with calipers by a non-reader radiologist (1 year of experience) to ensure these segments were consistent at MRA and DSA. Grafts, if present, were assessed separately. In the foot, only dorsalis pedis and lateral plantar arteries were assessed for stenosis. Assessment was based on The American College of Radiology grading system¹⁴: 0, normal; 1, <50% stenosis; 2, 1 lesion with ≥50%; 3, >1 lesion with >50% stenosis; 4, occluded. Diagnostic confidence was assessed on a 5-point Likert scale⁹: 1, non-diagnostic, poor delineation globally; 2, poor visualisation, not confident; 3, sufficient for diagnosis, reduced anatomic detail; 4, good visualisation, confident; 5, excellent visualisation, diagnosis certain.

A blinded interventional radiologist with five years’ experience analysed DSA images for stenosis and diagnostic confidence. Treated segments and segments downstream of these were excluded from statistical analysis.

Statistical analysis

Accuracy, sensitivity and specificity of MRA for haemodynamically significant stenosis (≥50%) was compared against DSA for each reader, overall and per-region, with all segments considered diagnostic at DSA (scores ≥ 3) included for analysis. Subset analysis of 7 subjects with eGFR < 30 mL/min/1.73 m² was also performed. Mean and standard deviation of overall, regional and segmental diagnostic confidence scores were calculated, with regional scores compared with the Wilcoxon signed-rank test. Inter-reader agreement for stenosis and diagnostic confidence and test reproducibility for haemodynamically significant stenosis in patients with repeat MRA were assessed with Kappa coefficients (≤0.20 poor; 0.21–0.40 fair; 0.41–0.60 moderate; 0.61–0.80 good; 0.81–1.00 very good). For MRA test reproducibility, all segments treated at DSA (and segments distal to this) were excluded from analysis if repeat MRA followed DSA. Analyses were performed with Stata Version 14 (Stata Statistical Software: Release 14, College Station, TX: StatCorp LP), with P-values < 0.05 considered statistically significant.

Results

MRA was completed in 28 of 32 presentations. Incomplete studies were due to patient girth incompatible with the magnet bore (BMI 40 kg/m², n = 1), discomfort (n = 2) and anxiety (n = 1).

Eight hundred and one segments were imaged at DSA, of which 664 (82.9%) were considered of diagnostic image quality, with 645 segments available for stenosis assessment. 55/140 (39.3%) segments of the foot imaged at DSA were considered diagnostic, largely due to poor opacification in heavily diseased pedal arteries. CO₂ assessment was performed in 72 segments, with only 40 segments (55.6%) considered diagnostic. Haemodynamically significant stenosis was identified at DSA in 262/645 segments (40.6%).

Accuracy, sensitivity and specificity

Overall combined QISS MRA and QISS-ASL MRA accuracy, sensitivity and specificity for haemodynamically significant stenosis for reader 1 (R1) were 78.1% (478/612 segments), 64.7% (161/249) and 87.3% (317/363), respectively, and for reader 2 (R2) was 75.6% (464/614), 77.5% (193/249) and 74.2% (271/365), respectively. Overall and per-region accuracy, sensitivity and specificity are summarised in Table 3, with representative examples shown in Figure 2 and Figure 3. Of note, per-region sensitivity was highest in the foot at 75% (9/12) for R1 and 84.6% (11/13) for R2.

Table 3.

Accuracy, sensitivity and specificity by region and overall, with segments reported in parentheses () and 95% confidence intervals in brackets []

	Accuracy (%)		Sensitivity (%)		Specificity (%)
	R1	R2	R1	R2	R1	R2
Pelvis	84.02* (63/75) [72.6–92.9]	72.7^ (56/77) [58.5–85.3]	33.3 (4/12) [9.9–65.1]	41.7 (5/12) [15.2–72.3]	93.7 (59/63) [84.5–98.2]	78.5 (51/65) [66.5–87.7]
Thigh	76.0** (137/181) [65.7–84]	76.0^^ (136/179) [65.9–84.2]	64.8 (46/71) [52.5–75.8]	81.4 (57/70) [70.3–89.7]	82.7 (91/110) [74.3–89.3]	72.5 (79/109) [63.1–80.6]
Calf	79.8*** (256/321) [73.2–85.1]	77.6^^^ (249/321) [70.3–83.8]	66.2 (102/154) [58.2–73.6]	77.9 (120/154) [70.5–84.2]	92.2 (154/167) [87.1 −95.8]	77.2 (129/167) [70.1–83.4]
Foot	62.9 (22/35) [37.3–82.9]	62.2 (23/37) [38.1–80.5]	75.0 (9/12) [42.8–94.5]	84.6 (11/13) [54.6–98.1]	56.5 (13/23) [34.5–76.8]	50.0 (12/24) [29.1–70.9]
Overall	78.1 (478/612) [73.3–82.5]	75.6 (464/614) [70.4–80.2]	64.7 (161/249) [58.4–70.6]	77.5 (193/249) [71.8–82.5]	87.3 (317/363) [83.5–90.6]	74.2 (271/365) [69.4–78.7]

^*Inferior to thigh (P = 0.59), inferior to calf (P = 0.10) and superior to foot (P = 0.70)

^**Inferior to calf (P = 0.08), superior to foot (P = 0.22)

^***Significantly superior to foot (P = 0.03)

^{^}Significantly inferior to thigh (P = 0.004), significantly inferior to calf (P = 0.0008) and inferior to foot (P = 0.29)

^{^^}Inferior to calf (P = 0.64), superior to foot (P = 0.26)

^{^^^}Superior to foot (P = 0.15)

Fig. 2.

A 75-year-old male presenting with bilateral claudication, (a) DSA, (b) coronal QISS MRA MIP, (c) feet QISS-ASL MRA MIP. Good correlation of right tibioperoneal trunk bifurcation stenosis is depicted (open arrows). Horizontal portion of left anterior tibial artery is not well seen on QISS MRA MIP where in-plane saturation of the vessel has occurred (arrows), and may not be recognised as an artefact, but was visible on thin slice high-resolution calf MRA, which is more sensitive to in-plane and slow flow. Good correlation of right medial plantar artery (solid arrows) and right lateral plantar artery (open dashed arrows). QISS-ASL MRA of the feet better shows the left lateral plantar artery than DSA, possibly related to timing of contrast administration on DSA (dashed arrows)

Fig. 3.

A 66-year-old female with stage 5 CKD (eGFR 10 mL/min/1.73 m²) on haemodialysis presenting with non-healing left foot ulcer, (a) DSA and coronal QISS MRA MIP of the left calf, and (b) DSA and lateral QISS-ASL MRA of the left foot. Good correlation of occluded distal left ATA shown on DSA (solid black arrows) and QISS MRA (solid white arrows), and patent dorsalis pedis on DSA (open black arrow) and QISS-ASL MRA (open white arrow). Successful angioplasty of the left ATA was subsequently performed

In subset analysis of 7 patients with stage 4 or 5 CKD, R1 assessed 115 segments and R2 assessed 114 segments for stenosis with available DSA correlation. Accuracy was 80.9% (93/115) and 80.7% (92/114), sensitivity 80% (44/55) and 76.3% (45/59), and specificity 81.7% (49/60) and 85.5% (47/55) for readers 1 and 2, respectively.

MRA diagnostic confidence

Each reader scored 1362 segments for diagnostic confidence. For R1, 87.5% (1192/1362) and for R2, 93.2% (1270/1362) segments were considered diagnostic. Overall mean diagnostic confidence was considered sufficient for diagnosis (score ≥ 3) for both readers (R1 3.89 ± 1.15; R2 3.25 ± 0.69, Table 4). Per-region mean diagnostic confidence scores of 3 or greater were reported in the pelvis (R1 3.86 ± 0.99; R2 3.26 ± 0.59), thigh (R1 4.04 ± 0.99; R2 3.25 ± 0.69) and calf (R1 4.42 ± 0.73; R2 3.51 ± 0.53). Pedal mean diagnostic confidence scores were less than sufficient for diagnosis, 2.8 ± 1.25 and 2.76 ± 0.74 for readers 1 and 2, respectively. For both readers, calf mean diagnostic confidence was significantly superior compared with other regions (P < 0.0001), and for the foot was significantly inferior compared with other regions (P < 0.0001). Although pedal region mean diagnostic score was non-diagnostic for both readers, segmental dorsalis pedis (R1 3.62 ± 1.10; R2 3.16 ± 0.41) and lateral plantar (R1 3.26 ± 1.15; R2: 3.02 ± 0.39) artery scores were diagnostic.

Table 4.

MRA mean (SD) diagnostic confidence scores. Scores of 3 or higher were considered sufficient for diagnosis

Vessel	Reader 1	Reader 2
Per-segment
Infrarenal aorta	3.79 (0.73)	3.34 (0.55)
Common iliac artery	4.02 (1.00)	3.31 (0.54)
Internal iliac artery (origin)	3.67 (0.93)	3.22 (0.56)
External iliac artery	3.93 (1.14)	3.21 (0.69)
Common femoral artery	4.28 (1.17)	3.10 (0.77)
Profunda femoris (origin)	3.98 (1.10)	3.05 (0.76)
Superficial femoral artery (proximal)	3.79 (1.02)	3.19 (0.78)
Superficial femoral artery (distal)	4.13 (0.77)	3.38 (0.49)
Popliteal artery (above-knee)	4.08 (0.84)	3.48 (0.54)
Popliteal artery (below-knee)	4.62 (0.66)	3.49 (0.54)
Anterior tibial artery (horizontal)	4.21 (0.88)	3.51 (0.54)
Anterior tibial artery (proximal vertical)	4.41 (0.64)	3.51 (0.54)
Anterior tibial artery (distal vertical)	4.48 (0.72)	3.54(0.50)
Tibioperoneal trunk	4.56 (0.72)	3.67 (0.54)
Posterior tibial artery (proximal)	4.30 (0.67)	3.41 (0.53)
Posterior tibial artery (distal)	4.31 (0.74)	3.39 (0.49)
Peroneal artery (proximal)	4.41 (0.72)	3.56 (0.53)
Peroneal artery (distal)	4.46 (0.79)	3.51 (0.50)
Dorsalis pedis	3.62 (1.10)	3.16 (0.41)
Medial plantar artery	2.93 (1.14)	2.97 (0.31)
Lateral plantar artery	3.26 (1.15)	3.02 (0.39)
Pedal arch	2.08 (1.14)	2.26 (0.93)
Metatarsal arteries	2.07 (0.89)	2.39 (0.90)
Graft	3.43 (0.79)	3.71 (0.76)
Per-region
Pelvis	3.86 (0.99)*	3.26 (0.59)^
Thigh	4.04 (0.99)**	3.25 (0.69)^^
Calf	4.42 (0.73)***	3.51 (0.53)^^^^
Foot	2.80 (1.25)	2.76 (0.74)
Overall	3.89 (1.15)	3.25 (0.69)

^*Significantly inferior to thigh (P = 0.02), significantly inferior to calf (P < 0.0001) and significantly superior to foot (P < 0.0001).

^**Significantly inferior to calf (P < 0.0001) and significantly superior to foot (P < 0.0001).

^***Significantly superior to foot (P < 0.0001).

^{^}Superior to thigh (P = 0.79), significantly inferior to calf (P < 0.0001) and significantly superior to foot (P < 0.0001).

^{^^}Significantly inferior to calf (P < 0.0001) and significantly superior to foot (P < 0.0001)

^{^^^}Significantly superior to foot (P < 0001).

Reasons provided for reduced diagnostic confidence (score ≤ 2) included susceptibility artefact (R1, 17 segments; R2, 22 segments), motion (R1, 5 segments; R2, 14 segments), noise (R1, 4 segments; R2, 21 segments) and other artefacts (R1, 3 segments; R2, 1 segment) including wrap, inhomogeneous fat suppression, spatial resolution limitations and flow saturation of vessels coursing within the acquisition plane (Figs 2,4). Susceptibility artefact from orthopaedic metalware was observed in 6 subjects: total hip replacement (n = 4), lumbar spinal fixation (n = 1) and tibial intramedullary nail (n = 1).

Fig. 4.

A 91-year-old male presenting with left leg claudication, (a) lateral left foot DSA and (b) lateral left foot QISS-ASL MRA. In-plane saturation of the posterior tibial artery shown on DSA (black arrow) along the plane of acquisition at QISS-ASL MRA (dashed white line) results in loss of signal (white arrow). Medial plantar artery is better shown on MRA than DSA (open arrows)

Inter-reader agreement

There was moderate overall inter-reader agreement for stenosis (K = 0.60), with good agreement in the calf region (K = 0.65), and moderate agreement in the pelvis, thigh and foot (K = 0.55, 0.51 and 0.36, respectively). Inter-reader agreement for diagnostic confidence was moderate overall (K = 0.41), with regional agreement in pelvis, thigh, calf and foot of 0.5, 0.48, 0.22 and 0.27, respectively.

Test–retest reproducibility

Test–retest reproducibility for stenosis assessment was very good for reader 1 in 302 segments (K = 0.87) and moderate for reader 2 in 305 segments (K = 0.54).

Discussion

In our experience of symptomatic diabetic patients with high burden of disease and comorbidities including renal failure and morbid obesity, results of combined QISS MRA and QISS-ASL MRA are promising.

Accuracy, sensitivity and specificity reported in this study are lower than previous 1.5T studies. Hodnett et al.’s initial experience in 25 diabetic patients with PAD reported high sensitivity (87.4%) and specificity (92.1%) of QISS MRA compared with contrast-enhanced MRA, and high sensitivity (96.2%) and specificity (96.1%) compared with DSA in 115 segments.¹⁰ A more recent study reported high sensitivity (84.9%) and specificity (97.2%) of QISS MRA compared with DSA, and importantly showed similar accuracy to CTA and better evaluation of heavily calcified arterial segments.¹² Another study in a general population of 53 patients with known or suspected PAD also reported high sensitivity (85.0–89.7%) and specificity (94.6–96.8%) of QISS MRA compared with contrast-enhanced MRA, and in a subset of 15 patients, high sensitivity (91.0%) and specificity (96.6%) compared with DSA.⁹ Two smaller studies comparing QISS MRA to contrast-enhanced MRA reported even higher sensitivity and specificity; Klasen et al.’s¹⁵ study of 27 patients, sensitivity (98.6%) and specificity (96%), and Ward et al.’s¹⁶ study of 20 patients referred for chronic lower limb ischaemia, mean combined regional sensitivity (90.8%) and specificity (94.5%). Furthermore, Altaha et al.’s¹¹ study of 19 patients with CLI comparing QISS MRA against DSA showed sensitivities of 92% and 81% for 2 readers and specificities of 95% and 97%.

We believe this reflects differences in our and most previous study populations and the reference standard, as well as differences in scan protocol. In our study, only patients with known PAD in whom DSA was planned for clinical reasons were recruited, including patients with stage 4 or 5 CKD. This resulted in a population with a high prevalence (40.6%) of haemodynamically significant stenosis. This is in comparison to the majority of prior studies, where patients with suspected as well as known PAD were recruited, leading to a lower prevalence of haemodynamically significant disease ranging from 23.8%¹⁵ to 32.8%⁹. Also, in prior studies, contrast-enhanced MRA was primarily the reference standard, necessitating exclusion of patients with severe renal impairment due to concerns regarding NSF.^9,10,15,16 In our study, DSA was used as the reference standard in all 32 subjects, including 4 subjects in whom CO₂ DSA was performed, where iodinated contrast was contraindicated. Encouragingly, in our subset of 7 patients with stage 4 or 5 CKD, the population where non-contrast MRA is most likely to be helpful who have not been previously systematically assessed good accuracy, sensitivity and specificity results were achieved.

Our study’s total scan time of 20 min, assuming a typical heart rate of 60 bpm, combining two techniques to cover the arteries from the infrarenal aorta to the feet, was longer than similar studies assessing QISS MRA alone. Hodnett et al.⁹ reported scan times of 6.2 min for imaging from infrarenal aorta to calf, without pedal assessment. Altaha et al.¹¹ reported a scan time of 106 ± 17s (82–137s) in assessment of 19 patients with CLI (14 with diabetes mellitus), but did not include high-resolution imaging (repeat calf imaging and QISS-ASL of the feet) that our study did. Our relatively long scan time may have contributed to incomplete studies in patients with discomfort (n = 1) and anxiety (n = 2), and led to greater proportion of motion affected segments. In unwell patients, faster imaging would therefore potentially be most important in ensuring reliable results. In clinical practice, it is likely that QISS-ASL MRA would not be required in all patients, and could be implemented additionally after QISS MRA only where no suitable distal bypass target is identified proximal to the ankle. It could also be argued that the additional higher resolution calf QISS MRA imaging may not need to be performed, with QISS MRA playing a preliminary assessment role prior to determining whether DSA is required.

It is also noted that DSA is an imperfect gold standard. In our population, 17.1% of segments were considered non-diagnostic, with a higher proportion (44.4%) in the 4 patients who underwent CO₂ angiography. Poor opacification in heavily diseased segments contributed to poor diagnostic confidence. Bowel-related peristalsis can degrade evaluation of pelvic segments, and the planar nature of standard DSA could lead to non-visualisation of en-face stenosis that is more easily identified with volumetric MRA.

With regards to image quality, despite including clinically obese patients, patients with arrhythmias, implanted hardware and severe renal impairment, readers still judged imaging of overall diagnostic quality in the pelvis, thigh and calf, speaking to the robustness of QISS MRA. This was shown in initial work by Edelman et al.⁸, in which two patients had cardiac arrhythmias, with excellent QISS MRA image quality, and previous work demonstrating superior performance of QISS MRA compared with turbo spin echo-based subtraction MRA.^11,16 Despite non-diagnostic image quality reported in the foot for per-region analysis, diagnostic image quality was achieved in the dorsalis pedis and lateral plantar artery, the main pedal targets for bypass surgery.

Overall, despite a combined training session prior to independent image analysis, there was only moderate inter-rater agreement for both diagnostic confidence and stenosis assessment. Our method of assessment with a 5-point Likert scale may also have contributed to the inter-rater agreement for diagnostic confidence as it is a broad scale of measurement. This highlights the subjective nature of assessment of peripheral arterial stenosis in clinical practice. Test reproducibility for stenosis assessment was very good for reader 1 (10 years’ experience) and moderate for reader 2 (4 years’ experience), suggesting experience contributes towards reproducibility of results for these techniques.

Study limitations included relatively small sample size in a single-centre, and a relatively lower number of diagnostic DSA segments available for comparison. Patients were usually referred for DSA for management of symptomatic disease, so only targeted essential runs were obtained. However, to capture symptomatic patients with severe renal impairment where QISS MRA has not been systematically assessed, we chose to use DSA as the reference standard.

Further refinement of QISS MRA and QISS-ASL MRA could lead to improved performance. For example, centre-out partial Fourier acquisition could enable better fat suppression⁸ and non-Cartesian readouts with accelerated techniques¹⁷ could shorten scan times and reduce patient discomfort. 3T imaging has also become more prevalent in clinical practice, offering potential for improved signal-to-noise ratio, but also challenges of increased field inhomogeneity and off-resonance artefacts. Several studies have reported favourable initial results of QISS MRA at 3T,^18–21 including a study of 32 patients compared with contrast-enhanced MRA and DSA¹⁸ with 14.4% disease prevalence, excluding patients with eGFR less than 30mL/min/1.73m², with reported accuracy of 92.4% and 93.8% in two readers.

In conclusion, QISS MRA is a feasible unenhanced technique for assessment of lower limb arteries in a symptomatic diabetic population, performing well in patients with significant renal impairment and critical ischaemia. QISS-ASL MRA is a potential add-on technique for assessment of pedal arteries where pedal bypass is under consideration, but requires further improvement if it is to be used clinically.

Abbreviations

2D

2 dimensional

ASL

arterial spin labelling

ATA

anterior tibial artery

BMI

body mass index

bSSFP

balanced steady-state free precession

CIN

contrast-induced nephropathy

CKD

chronic kidney disease

CLI

critical limb ischaemia

CO₂

carbon dioxide

CTA

computed tomographic angiography

DSA

digital subtraction angiography

DUS

duplex ultrasonography

ECG

electrocardiograph

eGFR

estimated glomerular filtration rate

MIPs

maximum intensity projections

MRA

magnetic resonance angiography

MRI

magnetic resonance imaging

NSF

nephrogenic systemic fibrosis

PAD

peripheral arterial disease

PTA

posterior tibial artery

QISS

quiescent-interval single-shot

RF

radiofrequency

SFA

superficial femoral artery