High-resolution, non-contrast-enhanced magnetic resonance angiography of the wrist, hand and digital arteries using optimized implementation of Cartesian quiescent interval slice selective (QISS) at 1.5 T

Abstract

Purpose

Non-contrast-enhanced (CE) magnetic resonance angiography (MRA) techniques are of considerable interest for diagnosing vascular diseases in the upper extremities owing to the possibility of repeated examinations, sufficient coverage of the measurement volume, and because possible side effects of administering iodine- or gadolinium-based contrast agents and radiation exposure can be avoided. The aim of this study was to investigate the feasibility of an optimized electrocardiogram (ECG) triggered Cartesian quiescent interval slice selective (QISS) technique for MRA of hand arteries.

Material and Methods

Both hands of 20 healthy volunteers (HVs) were examined using an optimized QISS-MRA pulse sequence at 1.5 Tesla. The wrist and hand arterial trees were divided into 36 segments. Cross-sectional areas (CSA) of all arterial segments were measured. For the technical evaluation of the pulse sequence, the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were computed and six imaging artifacts were graded. Two experienced observers used an ordinal scoring system to assess the image quality of each arterial segment. Interobserver agreement was determined.

Results

The median CSA was 7.3 mm² in the ulnar and radial artery, 3.2 mm² in the four common digital arteries, and 1.5 mm² in five proper digital arteries. The median SNR and CNR of the third common proper arteries were 45.9 and 20.3, respectively. None of the arterial segments were contaminated by venous enhancement. The image quality of arterial segments for both hands was considered as diagnostic in 87.2% of all 1440 segments. An interobserver agreement of 0.67 for both hands was determined for image quality of arterial segments using a five-grade scoring system. Optimized QISS-MRA allows as the first MRA technique the classification of superficial palmar arch (SPA) and deep palmar arch (DPA) variants. 5 new SPA and 6 new DPA variants could be classified using QISS-MRA in comparison with previous studies using CE computed tomography angiography and using fixed cadaver hands.

Conclusions

By using this optimized 2D Cartesian QISS-MRA protocol, contrast agent-free angiography of the wrist and hand arteries provided a high in-plane spatial resolution and an excellent visualization of small digital arteries.

Introduction

Physical function of the hand and the quality of life associated with it can be diminished due to vascular abnormalities of the hand, such as trauma, thromboembolic disease, vasculitis, vascular malformation, and Raynaud syndrome [1].

Contrast-enhanced (CE), invasive, intra-arterial digital subtraction angiography (DSA), CE computed tomographic angiography (CTA), and CE magnetic resonance angiography (MRA) represent standard clinical techniques to visualize the hand arteries [2–4]. These methods, however, require the use of ionizing radiation and iodine- or gadolinium-based contrast agents that are associated with rare, but possibly life-threatening side effects such as renal impairment, thyrotoxicosis, allergic reactions, and gadolinium deposition in several regions of the body [5, 6]. Thus, a noninvasive and non-CE imaging approach that also avoids ionizing radiation would be beneficial for patient safety. Doppler ultrasonography (DUS) is a cost-effective and noninvasive technique for hand angiography [7]. Unfortunately, though, it is not time-efficient because coverage of the hand arteries is limited and therefore several discrete local measurements are required, as compared to a MRA examination of hand arteries. Moreover, the reproducibility of DUS results, as with most ultrasound techniques, is operator dependent.

A variety of non-CE MRA techniques, such as time-of-flight (TOF) [3], phase contrast angiography [8], 3D electrocardiogram (ECG) triggered, variable-flip-angle fast spin echo sequence [9], fresh blood imaging (FBI) [10], 3D ECG triggered flow-sensitive dephasing (FSD)-prepared balanced steady-state free precession (SSFP) [11], and a fat-suppressed multi-shot non-SSFP with balanced gradients [12] have been employed for non-CE MRA of the hand. Long measurement times, susceptibility to flow-related effects, and registration artifacts between a dark blood arterial image and bright blood have limited their clinical usage [10, 13, 14].

Edelman et al. reported in 2010 about a 2D ECG triggered MRA technique (quiescent-interval slice-selective, QISS) with Cartesian k-space sampling for visualizing the peripheral arteries [15]. This non-CE MRA technique provided promising results for evaluating arterial disease of the lower extremities [16]. The aim of the present study, therefore, was to investigate the feasibility of an optimized 2D ECG triggered Cartesian QISS technique for MRA of the wrist and hand arteries at 1.5 Tesla.

Material and methods

Healthy volunteers

In total, 20 healthy volunteers (HVs) were included in this study. The status “healthy” was based on the following criteria: 1) absence of acute or chronic wrist and hand artery embolism in the medical history, and 2) no history of inflammatory, malignant, or fibrotic disease in the upper extremities or any hand and digital arterial stenosis before the examination. The general exclusion criteria for MRI safety were large-sized ferromagnetic materials or non-MRI compatible implants in the body and further contraindications to MRI [17].

The study was performed according to the protocol (No. D 508/18) approved by the ethic committee at the university medical center in Kiel and in accordance with the ethical standards laid down in the 1964 declaration of Helsinki and its later amendments. All study participants gave their informed consent in a written form.

Demographic data of study population

Demographic information (age, weight, body mass index “BMI” at examination, and gender) of the study population was recorded.

Preparation of the hands

Modeling clay (Kids efa plast classic, Eberhard Faber GmbH, Neumarkt, Germany) was used to reduce the magnetic susceptibility effects due to the air-filled spaces between the fingers. In order to have the same initial hand temperature for all examinations, the hand being examined and the modeling clay wrapped in a plastic glove were placed in a warm water bath at 50°C for 10 minutes [9]. The warmed modeling clay was rolled out with a wooden rolling pin to have a thickness of 1–2 mm. Depending on the size of the fingers, a piece of the rolled modeling clay was cut out and wrapped around the finger. The HVs lay prone with an outstretched hand (“superman position”). To reduce possible movements during the examinations and to make the examination convenient for the subjects, the cavities between the patient table and the forehead, chest, and arm of the HVs were padded with suitable fixation and foam materials/cushions.

Protocol optimization

To obtain high-quality 2D MRA images of wrist and hand arteries, the product QISS-MRA pulse sequence for visualizing peripheral arteries was optimized with respect to the following 12 points (Table 1):

1. Field of view (FOV) was changed from 400 to 258 mm².

2. Radiofrequency (RF) pulse type: The system offers a choice of RF pulses as “fast”, “normal”, and “low specific absorption ratio (SAR)” with a pulse length of 1.28, 2.56, and 3.84 ms, respectively. The type of RF pulse was changed from “normal” to “low SAR”.

3. Number of measurement slabs was reduced from 10 to 3 (proper digital arteries, hand, metacarpal and carpal, and wrist, Figure 1).

4. Breath-hold function and its associated voice command were deactivated.

5. Venous saturation slab was set as a tracking slab before all three measurement slabs.

6. Number of slices per measurement slab was increased from 50 to 80.

7. Thickness of venous saturation slab was reduced from 75 to 60 mm.

8. ECG electrodes were placed on the back of the HVs on the left side. All HVs had a normal cardiac anatomy (no situs inversus).

9. Trigger delay for the data acquisition was set from 100 to 0 ms.

10. Geometry of shim volume was varied between all three measurement slabs.

11. Adjustment tolerance of B₀-shim was deactivated. It was initially set to “auto”.

12. Asymmetric echo technique: The echo can be asymmetrically acquired with two strengths, “weak” and “strong”, with a relative echo position of 36% and 23% if this function is activated. Otherwise, the echo is sampled symmetrically. A symmetric readout refers to a relative echo position of 50%, whereas a fully asymmetric readout refers to 0%. Asymmetric echo acquisition was deactivated in the optimized protocol compared to “weak” asymmetric echo in the product protocol.

Table 1:

Angiographic images protocols

Parameters	ECG triggered Cartesian QISS-MRA
	Product	Optimized
Imaging mode	2D
TR (ms)	667.6	800.8
TE (ms)	1.7	2.8
Field of view (mm2)	400 × 260	258 × 167.7
Acquisition matrix (Px)	400 × 260	400 × 260
Reconstructed pixel (mm2)	0.5 × 0.5	0.3 × 0.3
In-plane interpolation	On
Slice thickness (mm)	3	1.5
Number of slices per slab	50	80
Slice distance factor (%)	−20
Number of averages	1
Receiver bandwidth (Hz/Px)	658
Flip angle (°)	120
Slice orientation	Transversal
Phase oversampling (%)	0
Filter	Distortion correction (2D); prescan normalizer
B0 shim mode	Standard
Asymmetrical echo	Weak	Off
RF pulse type	Normal	Low SAR
pulse length (ms)	2.56	3.84
Gradient mode	Fast
Maximum amplitude (mT/m)	24
Maximum rise time (μs/mT/s)	5.55
Maximum slew rate (mT/m/ms)	180.2
iPAT modus (acceleration factor/number of reference lines)	GRAPPA (2/24)
Partial Fourier (phase and slice)	5/8
Thickness of venous saturation slab (mm)	75	60
TI (ms)	345
TD (ms)	100	0
Acquisition time per slice (s)	<1

ECG = electrocardiogram, TR = repetition time, TE = echo time, Px = pixel, Hz = Hertz, bSSFP = balanced steady-state free precession, RF = radiofrequency, iPAT = integrated parallel imaging technique, GRAPPA = GeneRalized Autocalibrating Partial Parallel Acquisition, TI = time from in-plane and venous saturation to the acquisition of central k-space (ky=0), TD = trigger delay.

Figure 1:

Illustration of measured slabs. A dataset for each hand consists of three slabs: proper digital arteries (first slab), metacarpal and carpal (second slab), and wrist (third slab).

MRA imaging

Imaging was performed on a 1.5-Tesla MRI system (Magnetom Aera, XQ gradients, syngo MR VE 11C software, Siemens Healthcare GmbH, Erlangen, Germany) with a maximum gradient strength of 45 mT/m and a maximum slew rate of 200 T/m/s. The MR signal was received using a 16-element hand/wrist coil from the same manufacturer. Data was acquired using an optimized 2D ECG triggered Cartesian QISS-MRA pulse sequence (Table 1) with continuously moving table technique [18]. Both hands of right-handed HVs were examined in transversal slice orientation to the fingertips. Imaging of the right hand was performed first, which was followed by the left hand. This order remained the same for all examinations in HVs.

Image analysis

The wrist and hand arterial tree was divided into 36 segments [19] (Figure 2). The technical and clinical aspects in a total of 1440 arterial segments (20 HVs with two hands each) were evaluated.

Figure 2:

Schematic segmentation of hand arteries based on a maximum intensity projection image of a healthy volunteer. Each proper digital artery consists of three segments (orange marked numbers/lines) and thumb consists of 2 segments.

The technical quality of MRA images was graded by consensus decision of two board-certified radiologists (A. L. and J. H.) and a physicist (M. SR.) using a scoring scale of 0 (no artifact) to 1 (artifact present) with respect to various imaging artifacts (Figure 3). If an artifact was found, it was classified as either 1a, 1b, 1c, 1d, 1e, or 1f as follows:

• Grade 0: Absence of imaging artifacts

• Grade 1: Presence of

  1. venous contamination of arterial vessels
  2. signal void or interruption of signal intensity due to fracture points in the modeling clay or due to air bubbles under the modeling clay
  3. signal void or reduction in signal intensity along a vessel due to compressed vessel parts
  4. signal void along a vessel due to its parallel course to the measured slice orientation
  5. signal intensity changes within a measurement slab due to insufficient local magnetic field homogeneity
  6. stair-step artifacts along a vessel if the reconstruction algorithm of the angiogram failed.

Figure 3:

Exemplary views for evaluation of imaging artifacts. None of the hand datasets was affected by venous contamination (artifact 1a). Therefore, this artifact could not be demonstrated in this Figure. All other mentioned artifact types (1b – 1f) in the section “image analysis” are presented in different colors. The images for artifacts belong to two different healthy volunteers. For better visibility of the artifacts, the images were zoomed in or out in different ways.

The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were computed for technical evaluation of the MRA imaging quality. Three regions of interest (ROIs) with the same area (2.0–6.8 mm²) were placed in the middle of the third common proper artery (vessel 5.3, Figure 2), in the background without any visual imaging artifacts (ROI_Air), and in the radial artery. The SNR and CNR were determined using the extracted mean value and standard deviation (sd) of the signal intensities (SI) from these ROIs as follows [20]:

$$\text{SNR}=0.655\frac{\text{mean}\left(S{I}_{\mathit{\text{Vessel }}5.3}\right)}{\text{sd}\left(S{I}_{\mathit{\text{Air}}}\right)},\text{CNR}=0.655\frac{\text{mean}\left(S{I}_{\mathit{\text{Radialis}}}-S{I}_{\mathit{\text{Vessel }}5.3}\right)}{\text{sd}\left(S{I}_{\mathit{\text{Air}}}\right)}$$

There are nine variants of superficial and seven of the deep palmar arch (SPA and DPA) in the hands (Figures 4 and 5). Five of these have already been described in the literature [21–23]. We determined to which of these variants an angiogram of the right and left hand of each HV could be assigned.

Figure 4:

The classification of superficial palmar arch (SPA). Type A) Open SPA. The ulnar artery supplies the ulnar side of the 2^nd finger and fingers III-V completely, the radial artery the thumb, and the radial side of the 2^nd finger. Type B) Open SPA. The ulnar artery supplies the ulnar side of the 3^rd finger, the radial side of the 4^th finger, and the entire 5^th finger. The radial artery supplies the 1^st, the 2^nd, and the radial side of the 3^rd finger. Type C) Classical, completely closed SPA. It is formed by a strong ulnar artery and the superficial palmar branch of the radial artery. Five common palmar digital arteries emerge from the arch (corresponds to type A according to Coleman and Anson [23]). Type D) Completely closed, but not classically round SPA. Five common palmar digital arteries emerge from the arch (a variant of type A according to Coleman and Anson [23]). Type E) Open SPA. The radial artery supplies the thumb and the radial side of the 2^nd finger, the remaining fingers III-V, and the ulnar side of the 2^nd finger are fed from the ulnar artery. Type F) Open SPA. The radial artery supplies only the radial side of the thumb, the remaining fingers II-V and the ulnar side of the thumb are fed from the ulnar artery. (a variant of type B according to Coleman and Anson [23]). Type G) Open SPA. The radial artery supplies the thumb, the 2^nd finger, and the radial side of the 3^rd finger. The remaining fingers IV-V and the ulnar side of the 3^rd finger are fed from the ulnar artery. Type H) Mediano-ulnar SPA. In this case the arch is formed on the ulnar artery and a strong median artery. The radial artery does not contribute to the arch. Five common palmar digital arteries emerge from the arch. (corresponds to type C according to Coleman and Anson [23]). Type I) Radio-mediano SPA. The arch is formed on the radial artery and the median artery. The ulnar artery does not contribute to the arch. Two common palmar digital arteries emerge from the arch, three from the ulnar artery.

Figure 5:

The classification of deep palmar arch (DPA). Type A) Closed DPA. The radial artery and the deep palmar branch of the ulnar artery unite to form a closed arch. Four palmar metacarpal arteries emerge from the arch, the first of which completely supplies the radial side of the 2^nd finger. Type B) Open DPA. Two palmar metacarpal arteries emerge from the radial artery, the first of which completely supplies the radial side of the 2^nd finger. The two palmar metacarpal arteries branch out further. Type C) Closed DPA. The radial artery and the deep palmar branch of the ulnar artery unite to form a closed arch. Four palmar metacarpal arteries emerge from the arch, of which the first three branch out even further. Type D) Open DPA. Four palmar metacarpal arteries emerge from a strong radial artery. Type E) Open DPA. Four palmar metacarpal arteries emerge from a strong radial artery (with a slightly different course than type D). Type F) Open DPA. Two palmar metacarpal arteries emerge from the radial artery. Type G) Open deep palmar arch. Three palmar metacarpal arteries emerge from the radial artery. (corresponds to radial variant according to Mezzogiorno et. al. [22])

The cross-sectional area (CSA) of 1360 (34 arterial segments × 20 HVs × 2 hands) arterial segments were measured depending on their anatomical position either just before the bifurcation or at the beginning of a phalanx. Due to the variation in SPA and DPA within HVs, the CSAs of these arterial segments were not measured. The CSA evaluation was used to assess the accuracy of the optimized QISS-MRA pulse sequence for visualizing the wrist and hand arteries with a different length, size, and geometry.

The diagnostic quality of MRA images was evaluated by two board-certified radiologists (A. L. and J. H.) independently during separate sessions using a scoring scale of 0 to 4 with respect to visualization of the arterial segments.

• Grade 0: The arterial segment is not visible.

• Grade 1: Up to 25% of the length of the arterial segment is visible.

• Grade 2: Up to 50% of the length of the arterial segment is visible.

• Grade 3: Up to 75% of the length of the arterial segment is visible.

• Grade 4: The arterial segment is completely visible.

The interobserver agreement was calculated based on the diagnostic evaluation results of the two radiologists. Image analysis was performed on a workstation (IMPAX EF, Agfa, Healthcare, Bonn, Germany).

Statistical analysis

Median and range of variables are reported as summary statistics for all variables due to the severe skewness of the respective distributions. Normality (Gaussian distribution) was tested for all variables using the Shapiro-Wilk test. The p-values for comparison of two groups were obtained by applying a Wilcoxon signed-rank test, as appropriate. All significant p-values were reported with a precision of 10⁻⁵. A p-value <0.05 was considered statistically significant. Concordance between the two readers was calculated with the rater’s package (CRAN: raters). Agreement was interpreted as follows: <0.00, poor agreement; 0.00–0.20, slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81–1.00, almost perfect agreement [24]. The data was statistically analyzed with R (R Statistic package, version 3.5.1, R Foundation for Statistical Computing, Vienna, Austria). Diagrams were also plotted with the R Statistic package.

Results

All MRI examinations were performed successfully. All datasets were included in the study.

Characteristics of the study population

Our HVs consisted of 9 women and 11 men (median [min, max] age of 29.8 [19.7, 55.1] years, BMI of 25.0 [19.0, 34.0] kg/m²).

Technical optimization

1. FOV: A high in-plane resolution of 0.3×0.3 instead of 0.5×0.5 mm² could be achieved by decreasing the size of the FOV from 400 to 258 mm², depending on the size of the hand compared to the legs, while the base resolution remained constant.

2. RF pulse type: By changing the type of RF pulse from “normal” to “low SAR” the slice thickness could be reduced from 3 to 1.5 mm to obtain a sharp in-plane visualization of the vessel edges.

3. Number of measurement slabs was reduced, depending on the length of the hand compared to the legs, from 10 as in the product protocol to 3.

4. Breath-hold function and its associated voice command were initially intended to reduce motion during MRA imaging in the abdominal part of the body, which were not appropriate for hand MRA.

5. Venous saturation slab: Anatomically, the fingertips are not at the same height. Thus, it is not possible to set a spatial venous saturation slab before the first measurement slab for all fingertips. We decided, therefore, to set the spatial venous saturation slab before the tip of the middle finger (longest finger) to avoid differing venous saturation effects due to the position of the saturation slab for the finger segments.

6. Number of slices per measurement slab: The slab in the optimized protocol is thinner than that of the product pulse sequence (80×1.5 mm= 96 mm versus 50×3 mm = 120, a distance factor of −20% was set in both protocols).

7. Thickness of a venous saturation slab was set as half of the measurement slab in optimized protocol without considering the distance factor between slices.

8. ECG electrodes were placed on the backs of the HVs to make the examination more comfortable for them.

9. Trigger delay was set to 0 ms because the hands are anatomically positioned distal to the heart. Thus, this anatomical distance provides the required trigger delay for data acquisition.

10. Geometry of shim volume was adapted to the hand size for each measurement slab.

11. Adjustment tolerance of B₀-shim was deactivated, in order to have the best possible B₀-shim without variance for the measurement.

12. Asymmetric echo technique: Partial echo acquisition (asymmetric echo) allows reducing the echo time compared to imaging with full echo acquisition. However, acquisition of only a fraction of the echo (data) reduces the SNR in an image because the high frequency noise variance is increased while the low frequency signal remains the same. This fact represents a disadvantage for visualizing the arterial segment of the hand and for distinguishing signal intensities from the background [25].

Variations in the superficial and deep palmar arch

There were differences between the right and left hands of the HVs regarding the SPA and DPA variants. The SPAs were visible in all HVs. These SPA variants (A/B/C/D/E/F/G/H/I) were not equally present on both hands: right hands: 1/0/6/5/6/0/0/1/1; left hands: 1/1/6/5/4/1/2/0/0. The same SPA variant was observed on the both hands in 0/0/4/3/1/0/0/0/0 cases. The DPA variants could not be classified in 7 right hands and 8 left hands due to insufficient signal intensity. The classified DPA variants (A/B/C/D/E/F/G) were found 0/1/3/5/0/0/4 times in right hands and 1/0/2/2/1/1/5 times in left hands. The same DPA variant could be observed in both hands in 0/0/0/2/0/0/1 cases.

Cross-sectional areas of hand arterial segments

The CSAs in 1335 of 1360 (98.2%) arterial segments were measured. In total, 8 arterial segments in the right hands of HVs and 17 segments in left hands could not be included in the CSA evaluation. The invisible segments belong to a different part of the proper digital arteries. Of these segments, 15 (60.0%) were in the distal phalanx.

The CSAs of arterial segments were continuously reduced from the wrist to the fingertips on both hands (Figure 6). The ulnar and radial arteries had a median CSA [interquartile range, IQR] of 7.3 [6.4, 8.2] mm², four common digital arteries had a median CSA of 3.2 [2.5, 4.0] mm², and five proper digital arteries had a median CSA of 1.5 [1.0, 2.3] mm².

Figure 6:

Comparison between the cross-sectional areas of right and left hand of all 20 Healthy volunteers. In the middle and lower row, the proximal phalanx is abbreviated to ”p” intermediate to “i”, and distal to “d”.

There were no significant differences between the CSAs of the ulnar and radial arteries of the same hand (right hand: p=0.23, left hand: p=0.70) or those of both hands (ulnar artery: p=0.45, radial artery: p=0.16). CSAs of the common digital arteries of both hands were in the same range (p=0.41). The CSAs of the ulnar and radial side of the proper digital arteries of the same hand were in the same range (right hand: p=1.0, left hand: p=0.40), whereas the ulnar and radial side of the proper digital arteries in right hands were significantly lower than those in left hands (radial artery: 1.4 [0.9, 2.1] vs. 1.7 [1.1, 2.5], p=0.0004; ulnar artery: 1.4 [0.9, 2.1] vs. 1.7 [1.1, 2.3], p=0.006).

SNR and CNR

There were no significant differences between the SNR of the common proper artery (artery 5.3 in Figure 2) of the right hand and that of the left hand (Right hand vs. left hand: Median [IQR], 61.2 [38.2, 82.8] vs. 40.1 [32.4, 55.2], p=0.05). The results of CNR are in line with those of SNR (Right hand vs. left hand: Median [IQR], 25.9 [10.1, 62.7] vs. 19.6 [11.3, 37.5], p=0.56). The median SNRs and CNRs of the third common proper arteries of the right and left hands were 45.9 [33.6, 66.0] and 20.3 [10.1, 54.0], respectively.

Presence of imaging artifacts

None of the arterial segments 0.0% (0/1400) was contaminated by venous enhancement. A signal void or interruption of signal intensity due to fracture points in the modeling clay or due to air bubbles under the modeling clay was observed in 38.0% (547/1440) of segments. In all, 13.0% (187/1440) of segments were affected by a signal void or reduction in signal intensity along a vessel due to compressed vessel parts. A signal void along a vessel due to its parallel course to the measured slice orientation was observed in 19.0% (274/1440) of segments. Signal intensity changes within a measurement slab due to insufficient local magnetic field homogeneity were present in 35.0% (504/1440) of segments. Stair-step artifacts along a vessel were observed in 10.0% (144/1440) of segments.

Diagnostic quality of hand arterial segments

In the right hands, both readers scored 65.7% (946/1440) of branches as grade 4 (excellent), 19.9% (287/1440) as grade 3 (good), 8.7% (125/1440) as grade 2 (non-diagnostic), 4.4% (63/1440) as grade 1 (limited visible), and 1.3% (19/1440) as grade 0 (invisible). In the left hands, both readers identified 62.0% (893/1440) of branches as grade 4, 26.8% (386/1440) as grade 3, 6.0% (87/1440) as grade 2, 3.1% (44/1440) as grade 1, and 2.1% (30/1440) as grade 0 (Table 2).

Table 2:

Comparison between two readers for grading the visibility of arterial segments in both hands. The p-values refer to Wilcoxon signed-rank test.

	Right hand			Left hand
Visibility of arterial segments (%)	Reader 1	Reader 2	p-value	Reader 1	Reader 2	p-value
0	1.2% (9/720)	1.4% (10/720)	0.83	1.9% (14/720)	2.2% (16/720)	1.0
25	5.0% (36/720)	3.8% (27/720)	0.28	3.5% (25/720)	2.6% (19/720)	0.62
50	7.9% (57/720)	9.4% (68/720)	0.43	4.7% (34/720)	7.4% (53/720)	0.05
75	19.9% (143/720)	20.0% (144/720)	0.89	28.6% (206/720)	25.0% (180/720)	0.16
100	66.0% (475/720)	65.4% (471/720)	0.86	61.3% (441/720)	62.8% (452/720)	0.66

The image quality of both hands was considered as diagnostically “good to excellent, grade≥3” in 87.2% (2512/2880) of arterial segments. In total, 12.8% (368/2880) of segments had a nondiagnostic imaging quality. Of the nondiagnostic segments, 17.4% (64/368) belonged to SPA and DPA, 11.4% (42/368) to common digital arteries, and 71.2% (262/368) to all branches of proper digital arteries.

Interobserver agreement

A substantial interobserver agreement was determined for image quality of arterial segments for both hands (Table 3).

Table 3:

Interobserver agreement for the evaluation of Cartesian QISS-MRA based on the 5-point scale scoring system introduced in the section “image analysis”.

Variable	Interobserver agreement
Image quality
Right hand	0.68 (0.63, 0.73)
Left hand	0.67 (0.62, 0.72)
Both hands	0.67 (0.64, 0.71)

Data presented in the form: agreement (95% confidence interval).

Discussion

In contrast to the lower extremity vessels, only few reports about MRA of the hands and digital arteries have been published because atherosclerosis, the most frequent indication for MRA, is uncommon in the peripheral arteries of the upper extremities [2, 3, 9, 11, 12]. The filigree and more complex vasculature of the hands with marked individual, superficial and deep vessel structure may be another reason for the limited number of studies in this field [26, 27]. Moreover, the diameter of the hand arteries varies over a short distance between wrist and fingertips from 2.5–3.5 mm to 1.0–1.2 mm [27–29]. Thus, obtaining a reliable and high spatial resolution angiogram of the upper extremities has clinical impact, particularly on diagnosis and treatment of nonatherosclerotic diseases, such as systemic sclerosis and Buerger’s disease.

In these entities, invasive intra-arterial CE-DSA is often used as the standard imaging modality [30, 31]. To avoid invasive vascular access, radiation exposure, and administration of iodine-based contrast agents, DUS of hand and finger arteries can be used as an alternative method in connective tissue diseases and Raynaud’s disease [32]. As DUS of finger arteries is sophisticated and time consuming, however, a less challenging and examiner-dependent imaging modality would be helpful in assessing the peripheral arteries of the upper extremities. CE-CTA represents an alternative minimally invasive method with the ability to visualize the hand arteries, for example, the superficial palmar arch [27]. However, this method is not used in clinical routine due to the technical limitations for imaging parts of the upper extremity arteries that are more peripheral than the wrist [33]. CE-MRA of the hand and the finger arteries is an established clinical method which can be applied in rheumatic diseases, such as systemic sclerosis [34].

Non-CE MRA techniques are of considerable interest for diagnosing vascular diseases in the upper extremities owing to their possibility for repeated examinations, sufficient coverage of the measurement volume, and because possible side effects of administering iodine- or gadolinium-based contrast agents and radiation exposure can be avoided, as compared to invasive intra-arterial CE-DSA, DUS, CE-CTA, and CE-MRA. Since non-CE 2D ECG triggered Cartesian QISS-MRA was introduced in 2010 for evaluating the lower extremities [15], this technique and its variants have been used and clinically evaluated in a variety of vascular territories.

Main results of the study

In our prospective study, an optimized implementation of 2D ECG triggered Cartesian QISS-MRA was technically evaluated for angiography of the wrist and hand for the first time in 20 HVs. The main findings of our study are as follows: 1) QISS-MRA allows complete suppression of venous signal intensity, 2) QISS-MRA provides good visualization of the hand and digital arteries without contrast agent, 3) based on the evaluation of CSAs, QISS-MRA provides high in-plane spatial resolution for a reliable detection of wrist, hand, and digital arteries of various diameters, 4) QISS-MRA allows as the first MRA technique the classification of SPA and DPA variants, and 5) 5 new SPA and 6 new DPA variants could be classified using QISS-MRA in comparison with previous studies using CE-CTA and using fixed cadaver hands. Optimized 2D ECG triggered Cartesian QISS-MRA has, therefore, the potential to be used in a non-CE MRI-based approach for diagnosing vascular diseases in the upper extremities in clinical routine.

Technical aspects

In the first step of our study, we optimized the product 2D ECG triggered Cartesian QISS-MRA pulse sequence for hand MRA.

Temperature-dependency of the signal intensity

The MR signal characteristics are temperature-dependent [25]. Low temperature minimizes the vessel lumen (vasoconstriction) and the perfusion [9]. This phenomenon can be observed in patients with Raynaud syndrome [35]. Therefore, the hands were warmed up before an examination. Preparing the hands was a time-consuming step in this project. We speculate that this step could be avoided if 1) an ultra-short or a zero echo time technique would be available for MRA imaging instead of the commonly used readout technique in bSSFP, or 2) the hand could be warmed up using an integrated warm air pump while the hand is positioned in the coil, like with a commonly used warm air pump for a small animal examination.

Shape of the hand/wrist coil

The shape of the inside of the hand/wrist coil corresponds to a flat and narrow box with two openings on the fingertips and the wrist side. This shape does not match the anatomical shape of the hand. The hand/wrist coil is equipped with two flat pads for fixating the hands. Men’s fingers and, in particular, fingers prepared with modeling clay are usually slightly thicker than women’s and unprepared fingers. Due to the small space inside the coil, the prepared fingers are automatically pressed together, even without the two pads. The arterial signal intensity of the thumb and the little finger are therefore reduced or a signal void is observed in these fingers and their segments. Depending on the size of hands, we could partly overcome this limitation. We unrolled a sterile gauze bandage and formed an oval pad such that an anatomically arched position of the hand in the coil could be achieved. It would be desirable to use a glove-shaped coil to measure the hand [36].

Imaging artifacts

In the literature, various materials and high-permittivity dielectric pads have been presented for reducing susceptibility artifacts, such as barium titanate (BaTiO₃) and rice pads [37–40]. In our experience, the fine structure, moisture, and flexibility of the modeling clay used in our study provide for a suitable substance that can be wrapped around the fingers for about ten minutes without crumbling or drying out. The use of a modeling clay or a similar material for filling the empty spaces between the fingers and thus eliminating the susceptibility artifacts could be omitted if a broadband excitation pulse could be implemented in the sequence [41]. The observed changes in the signal intensity between arterial segments within a slab can be caused by the different shim status of the individual vessels. This artifact could be omitted by employing high-order shimming for more accurate magnetic field correction.

The appearance of stair-step artifacts in 10% of arteries is due to geometric mismatch between adjacent slices in the reconstruction algorithm. The artifact does not degrade the diagnostic quality of the hand arteries.

QISS-MRA using ECG triggering

QISS was originally described as a technique to leverage cardiac gating and optimally synchronize the quiescent intervals and readouts to rapid systolic and slow diastolic arterial flow, respectively. In our study, the angiograms were acquired using an optimized 2D ECG triggered Cartesian QISS-MRA. Since the hands are positioned distal to the heart, peripheral pulse (PPU) triggering instead of ECG triggering might be more appropriate from a physiological point of view. However, there is not enough space for a PPU clip in the hand/wrist coil. Moreover, a PPU clip locally compresses the small digital arteries together and causes additionally local magnet field inhomogeneities, which could degrade the quality of the angiogram. However, all MR imaging protocols for imaging in the hand are performed without cardiac or PPU triggering. Indeed, it is most convenient from a clinical perspective to image without cardiac synchronization and ECG electrodes. Two potential strategies for making the acquisition more comfortable for clinical routine may include: 1) using the novel beat sensor technique for ECG triggering without ECG electrodes [42], or 2) using an ungated implementation of the QISS-MRA as presented for MRA imaging in head and neck [43].

Dependency of the perfusion status on the handedness

In the literature, some studies [44, 45] reported that the perfusion status of hand arteries may be positively affected, depending on the handedness, if a certain hand side is dominant. We included only right-handers in our study to avoid this possible physiological effect on our results.

Classification of SPA and DPA

We classified SPA and DPA variants using a MRA technique for the first time. To our knowledge, no CE or non-CE MRA technique has been reported in the literature showing the classification and visibility of these variants. In comparison with previous studies using CE-CTA and using fixed cadaver hands, we classified 5 new SPA and 6 new DPA variants [21–23]. This highlights the diagnostic value of QISS-MRA in the assessment of hand arteries.

In some hands of HVs, the DPAs could not be clearly observed due to insufficient signal intensity. In the classified DPA variants, the entire vessel courses along only some of the fingers were visible. Based on the source QISS-MRA images, the caliber of DPA branches decreases shortly after branching from the arch faster than those of SPA to supply the hand muscles. The reduced perfusion in the small-caliber DPA branches and also their nonlinear course may make it difficult to distinguish them from surrounding tissue.

Portions of SPAs and DPAs as well as the roots of some common proper digital arteries and thumb arteries lie anatomically parallel to the measured slice and were not visualized due to technical limitations. Such signal void segments are undesirable for an accurate diagnosis. The use of flow-indepedent MRA, such as bSSFP or related techniques could be helpful to depict such segments [46, 47].

CSA of wrist and hand arterial segments

In the second part of our study, we measured the CSA of wrist and hand arterial segments in our HVs. The CSA could not be assessed in 0.8% of arterial segments, where signal intensity was too low to reliably distinguish the vascular lumen from surrounding tissue. Based on CSA results, the wrist and hand arteries with median diameters ranging from 3.1 to 1.4 mm (CSA of 7.5 to1.5 mm²) can be distinguished from each other. We found only a significant median difference of 0.3 mm between the ulnar and radial sides of proper digital arteries in the right and left hands of HVs. This significant difference is due to missing proper digital arterial segments based on the various aforementioned artifacts. No comparable non-CE MRA techniques have been published in the literature. Fazan et al. and Leslie et al. measured diameter using DSA and a postmortem study [29, 48]. These non-MRA studies confirm the accuracy of our CSA results.

SNR and CNR of the third common proper artery

The SNR and CNR were determined in the third part of our study. We decided to measure the signal intensity in the third common proper artery because it gave the highest median CSA on both hands according to our results. The area of ROIs drawn in the radial artery and in the background for calculating SNR and CNR depended on individual CSA of the third common proper artery on each hand of a HV. Here, too, no comparable study has been published in which SNR or CNR was determined by using non-CE MRA techniques. Winterer et al. [2] presented median SNR values of 21–36 for radial and ulnar artery in 14 HVs using CE MRA at 1.5 Tesla.

Diagnostic quality of the optimized Cartesian QISS-MRA

Except for some arterial segments, the optimized 2D ECG triggered Cartesian QISS-MRA provided a “good to excellent” diagnostic quality in about 90% of segments. Cartesian QISS-MRA does not require careful adjustment of the velocity-encoding setting in contrast to phase contrast MRA technique [8]. The visualization of wrist and hand arteries using Cartesian QISS-MRA is not based on image subtraction with or without flow dephasing gradients (which must also be adjusted in magnitude, duration and timing), in comparison with FSD-based bSSFP. The measurement time for Cartesian QISS-MRA is shorter (~1 s/slice) than for other non-CE MRA techniques in the published literature. Therefore, the image quality is less affected by patient movement and user-dependent factors, and the measured spatial resolution is much higher than that of previous non-CE MRA methods.

Based on detailed grading of image quality of the arterial segments, we found a substantial interobserver agreement for both hands. This value could be increased if the grading steps were reduced from 5 to 3 (grades 0–2 as nondiagnostic, grade 3 as good, and grade 4 as excellent). There is no comparable interobserver agreement data in the literature concerning the grading of arterial hand and wrist segments.

Potentially clinical aspects

The development and technical optimization of pulse sequences for CE or non-CE hand MRA techniques focused on acquisition time, spatial resolution, SNR, and CNR for obtaining selective arterial images without venous contamination. The evaluation of our data showed that the 2D ECG triggered Cartesian QISS-MRA is able to depict arterial segments of the hand without venous contamination. This fact suggests that the technique can be applied as part of a clinical routine without any venous contamination problems, for instance, for diagnosing various arterial abnormalities in the upper extremities, planning cardiosurgical bypass surgery using the radial artery graft [49], for accessing the transradial artery in interventional radiology and neuroradiology [50], and also for serial follow-up imaging. There are two non-CE MRA techniques comparable to Cartesian QISS-MRA. Bode et al. [12] presented the results of a modified non-CE balanced gradient echo pulse sequence, which simultaneously portrays both arteries and veins of upper extremities vessels, including the wrist. However, it would be difficult to use this technique for diagnosing arterial disease with regard to the small-vessel diameter and the adjacent vascular structures in the hand and in the fingers. In a study by Fan et al. [11], angiograms of hands using a non-CE FSD-based MRA technique were presented. Apart from banding artifacts at the wrist, venous contamination was more evident at the wrist and palm on these angiograms, while QISS-MRA did not produce any venous contamination in our study.

Study limitations

Our study was focused on a detailed technical analysis for visualizing 36 wrist and hand arterial segments using an optimized 2D ECG triggered Cartesian QISS-MRA technique. Therefore, only HVs without any pathologic condition were included in our study and, thus, invasive intra-arterial CE angiography as the clinical reference standard was not applicable. Due to deposition concerns and possible side effects associated with gadolinium-based contrast agents, we did not perform additional CE-MRA in this pilot study.

Conclusions

In conclusion, by using the optimized 2D ECG triggered Cartesian QISS-MRA protocol presented here non-CE angiography of the wrist and hand arteries provided a high in-plane spatial resolution and an excellent visualization of small digital arteries. In future studies, this pulse sequence should be evaluated in patients with various hand arterial diseases.

Highlights:

• High-resolution, non-contrast-enhanced magnetic resonance angiography of the wrist, hand and digital arteries using optimized implementation of 2D electrocardiogram (ECG) triggered Cartesian quiescent interval slice selective (QISS) at 1.5 T

• Optimized 2D ECG triggered Cartesian QISS-MRA allows complete suppression of venous signal intensity in the wrist, hand and digital arteries

• Optimized 2D ECG triggered Cartesian QISS-MRA allows for the first time the classification of superficial palmar arch (SPA) and deep palmar arch (DPA) variants using a MRA technique

• 5 new SPA and 6 new DPA variants could be classified using optimized 2D ECG triggered Cartesian QISS-MRA in comparison with previous studies using contrast enhanced computed tomography angiography and using fixed cadaver hands