Evaluation of three-dimensional navigator-gated whole heart MR coronary angiography: The importance of systolic imaging in subjects with high heart rates

2.1. Subjects 

Twenty-seven healthy adult volunteers were enrolled in this study (9 male/18 female; mean age 33±9 years). All subjects were in sinus rhythm. Patients of heart rate ≤65bpm were classified as low HR group, otherwise as high HR group according to the upper heart rate threshold of coronary motion-free images previously suggested by MDCT studies [9], [10], [11], [12]. The study protocol was approved by the institutional review board and all subjects gave informed consent before the enrollment.

2.2. MR imaging protocol 

All magnetic resonance (MR) imaging were performed with a commercial 1.5-T whole-body MR system (Avanto, Siemens Medical Solutions, Erlangen, Germany) [13]. A phased-array body matrix coil was used, with the subject in the supine position. A four-lead vector ECG was obtained for cardiac gating. The pulse sequence was an R-wave-triggered, single-slab 3D gradient-echo-sequence with no velocity compensation, employing a navigator spin-echo measurement to track the variation of z-position of the diaphragm during the scan. Initial survey images were obtained in three orthogonal directions to help determine the positions of the heart and diaphragm. Scout and reference images were taken during free breathing. To determine the motion of the coronary arteries, multiphase cine MR images were obtained in the transverse plane with a steady-state free precession (SSFP) pulse sequence (TE/TR/flip angle=2.8ms/5.6ms/25°, matrix size=192×156, number of cardiac phases=64 (shared phases), and acceleration factor=GRAPPA×3). The motion of right coronary artery (RCA) was visually assessed by scrolling through the transverse cine MR images, and the rest period of minimal displacement during diastole and systole were defined as optimal acquisition window and a subject-specific trigger delay time was determined accordingly [4], [8]. On the basis of these cine survey images, the whole-heart MRCA images without a gadolinium chelate agent administration were acquired in a trans-axial orientation using a magnetization-prepared 3D centric-ordered SSFP sequence using parameters as following: TE/TR/flip angle=1.6ms/3.2ms/90°, lines per heartbeat=15, readout bandwidth=678.5Hz/pixel, acceleration factor=GRAPPA×2, FOV=270mm×270mm×120mm, and acquisition matrix=256×256×80 (interpolated voxel size=0.55mm×0.55mm×0.75mm). Magnetization preparation included a T2 preparation pulse and a frequency-selective fat-saturation pulse. An automated shim procedure was applied to the volume of the entire heart. For real-time respiratory gating and positioning of the 3D volume in the craniocaudal direction, a navigator echo was acquired from a cylindrical region delineated by a two-dimensional excitation pulse perpendicular to the right hemidiaphragm, with a gating window of 4mm [4]. The 3D MRCA images were acquired with the identical section orientations, positions and pulse sequences in the diastolic and systolic phases. The R-R interval lengths were recorded beat by beat during cine MR imaging acquisition. The heart rate variation was calculated as [one standard deviation (S.D.)/mean of R–R interval]×100%.

2.3. Image display and data analysis 

All MR images data were transferred to a workstation with commercially available image reconstruction software (syngo MR version, Siemens). Two independent, blinded, experienced observers assessed the anonymized MRCA data sets separately, but non-blinded to the cardiac phases. In addition, in order to evaluate the intraobserver variation, all images were scored one month later by the same readers who were blinded to patients’ identity with the same approach. After analysis of interobserver agreement, discordant findings were resolved by consensus reading.

The MRCA images of diastole and systole were compared side-by-side with a slice-thickness of 1.5mm in axial sections. Imaging quality was visually assessed with a four-point scale which a score of (1) indicating poor visibility (coronary vessel barely seen, or was obscured by noise); (2) marginal visibility (coronary was visible, but confidence in the diagnosis was low, due to moderately blurred borders); (3) good visibility (the vessel was adequately visualized, with confidence in the diagnosis, only mildly blurred borders) and a score of 4, excellent visibility (the vessel was well depicted with sharply defined borders) [2], [3], [4]. For diagnostic performance assessment, scores of 1 or 2 was classified as non-diagnostic. The scores of left main coronary artery (LM), proximal, middle and distal portions of left descending artery (LAD), left circumflex artery (LCX) and RCA, first diagonal (D1) and first obtuse marginal (M1) branches were recorded in 16-segment model as American College of Cardiology/American Heart Association recommendation [14]. We additionally evaluated the diagnostic performance in the coronary segments diameter ≥2mm on per-vessel and per-patient basis.

In this study, axial images were used for comparison instead of 3D voxel-based post-processing algorithms in order to avoid reconstruction errors.

2.4. Statistical analysis 

Continuous variables are expressed as mean±S.D., with categorical data presented as number and the percentages. For comparisons of demographic and scan parameters between low and high HR groups, Student's t-test and chi-square analyses (with Fisher's exact test in smaller sizes) were used for continuous and categorical variables, respectively. The kappa analyses were performed as measures of intraobserver and interobserver agreements in terms of image quality scores within each cardiac phase, valued as follows: 0–0.2 low, 0.21–0.40 moderate, 0.41–0.60 substantial, 0.61–0.80 good, and ≥0.81 perfect of agreement. The interpretations of consensus were used for subsequent analyses. The visual scores and diagnostic performance between two groups were compared by NcNemar's χ2-test. The comparison between diastolic and systolic images within each group was assessed with paired t-test.

All statistical tests were two-sided, and a p value <0.05 was considered as statistically significant. Statistic data analyses were done with the software Stata for Windows, Version 8 (Stata Press, College Station, TX).

3.1. Subject characteristics 

Of 27 subjects, the mean heart rate during the scan was 67.1±10.2bpm (range 53–110). Thirteen subjects were classified as the low HR group (HR 59.8±4.9bpm, range 53–65), and 14 were high HR group (HR 73.9±9.0bpm, range 68–110) according to predetermined HR threshold of 65bpm [9], [10], [11], [12]. There were no statistically significant differences in age, gender distribution, heart rate variation and total scan time between two groups (Table 1). All subjects successfully underwent the scans without complication, and nearly all MRCA of two cardiac phases were completed around 30min no matter baseline heart rates.

Table 1. Subjects and scan characteristics (n=27)

		Low HR group (n=13)	High HR group (n=14)
	Age	31.9±4.5	34.1±11.5
	Gender (male)	3 (23%)	6 (43%)
	Heart rate (bpm)	59.8±4.9 (53–65)	73.9±9.0 (68–110)*
	R–R interval (ms)	1008.8±85.1	821.4±78.9
	S.D. of R–R interval (ms)	32.8±30.0 (12–125)	33.3±45.0 (5–176)
	Heart rate variation (%)	3.2±2.8 (1.3–11.7)	4.0±5.4 (0.8–20.9)
	Scan time (s)
	Diastolic MR angiography	1011.5±157.3	918.9±212.3
	Systolic MR angiography	993.4±197.7	884.4±180.8
	Time of delay (ms)
	Diastolic MR angiography	616.2±61.0	488.1±85.2*
	Systolic MR angiography	191.2±26.6	161.0±44.2*

Data are mean values±S.D. (range) or number (%). HR: heart rate; MR: magnetic resonance.

* p<0.05 in low vs. high HR group.

3.2. Image quality evaluation 

3.2.3. Segment-by-segment analysis 

The results of quantitative angiographic scores in diastole and in systole for each coronary artery segment are summarized in Table 2. In low HR group, there were significantly poorer scores in proximal and distal segments of LAD and D1 at systole than at diastole. In high HR group, systolic MRCA had better image quality scores generally. There were no differences in MRCA image quality of diastolic images in low HR group and these of systole in high HR group in all major coronary arteries.

Table 2. Quantitative angiographic scores for each coronary artery segment (n=329 segments)

		Low HR group (n=13)		High HR group (n=14)
		Diastolic	Systolic	Diastolic	Systolic
	LCA
	LM	3.9±0.3	3.8±0.4	3.7±0.6	3.9±0.4
	LAD
	Proximal	3.7±0.5	3.4±0.5*	3.1±0.6†	3.4±0.6*
	Middle	3.2±0.6	3.0±0.4	2.8±0.7	3.2±0.5
	Distal	3.0±0.5	2.3±0.6*	2.3±0.7†	2.8±0.6*
	D1	3.3±0.5	2.8±0.4*	2.5±0.8†	2.9±0.7
	LCX
	Proximal	3.7±0.5	3.5±0.5	3.0±0.7†	3.5±0.6*
	Middle	2.9±0.6	2.8±0.6	2.5±0.8†	2.9±0.6
	Distal	2.7±0.5	2.4±0.7*	2.1±0.8	2.6±0.8*
	M1	3.2±0.8	2.8±0.8	2.4±0.7†	2.9±0.6*
	IM (n=5)	3.0±0.0 (n=2)	2.5±0.7	2.7±0.3 (n=3)	3.3±0.3*
	RCA
	Proximal	3.7±0.4	3.6±0.6	3.4±0.5	3.8±0.5*
	Middle	3.4±0.5	3.2±0.6	2.9±0.8†	3.3±0.7*
	Distal	3.3±0.7	3.2±0.8	2.8±0.8	3.1±0.7

Data are mean values±S.D. HR: heart rate; LCA: left coronary artery; LAD: left anterior descending artery; D1: first diagonal branch; LCX: left circumflex artery; M1: first obtuse marginal branch; IM: intermediate branch; RCA: right coronary artery.

* p<0.05 in diastolic vs. systolic images within each group. † p<0.05 in diastolic images between low and high HR groups.

Overall, there were 91.8% (145/158) of segments at diastolic images of low HR group were classified as assessable with adequate image quality, and 82.3% (130/158) were assessable in systolic images. On the contrary, 69.0% (118/171) of segments of diastolic images in high HR group were classified as assessable, while 88.3% (151/171) were assessable in systole. There was a significant reduction of number of unevaluable segments in the systolic MRCA in patients of high HR group.

In terms of image quality and vascular distribution, LM and proximal portions of major coronary arteries were generally satisfactory (mean scores ≥3), and the middle and distal portions of LCX had the trends toward lower image quality when comparing to the corresponding portion of LAD and RCA, although not reaching statistical significance (p=NS).

3.2.4. Vessel-by-vessel analysis 

The analysis of diagnostic performance in MRCA by vessel-based is shown in Table 3 and Fig. 1. In segments ≥2mm, image quality was classified as non-diagnostic only in 8.8% (7/80) of vessels from diastolic MRCA in low HR group, and 9.2% (8/87) of systolic images in high HR group (p=NS).

Table 3. Vascular distribution of non-diagnostic imaging quality on MR coronary angiography (low HR group=13; high HR group=14 subjects)

		Diastolic images		Systolic images
		Low HR	High HR	Low HR	High HR
	LM	0 (0%)	1 (7%)	0 (0%)	0 (0%)
	LAD	1 (7%)	5 (36%)*	0 (0%)	1 (7%)
	D1	0 (0%)	8 (57%)*	2 (15%)	3 (21%)
	LCX	1 (7%)	7 (50%)*	3 (23%)	1 (7%)
	M1	3 (23%)	8 (57%)*	3 (23%)	2 (14%)
	IM	0 (0%) (n=2)	0 (0%) (n=3)	1 (50%) (n=2)	0 (0%) (n=3)
	RCA	2 (15%)	4 (29%)	2 (15%)	1 (7%)

Values are n (%). HR: heart rate; LCA: left coronary artery; LAD: left anterior descending artery; D1: first diagonal branch; LCX: left circumflex artery; M1: first obtuse marginal branch; IM: intermediate branch; RCA: right coronary artery.

* p<0.05 between low and high HR groups.

View Large Image Download to PowerPoint

Fig. 1. Diagnostic performance of diastolic (a) and systolic (b) images of MR coronary angiography by vessel-based analysis (167 vessels; 80 of low HR group and 87 of high HR group). The image quality of diastolic phase was significantly better in low HR group, and there was no significant difference in diagnostic rate of systolic images between two groups. MR: magnetic resonance; HR: heart rate.

3.2.5. Patient-by-patient analysis 

In a subject basis analysis, there were 77% (10/13) of low HR group with adequate image quality at diastole, while 79% (11/14) of patients with higher heart rates have adequate image quality at systole (p=NS). The imaging quality during systole was significantly better than those during diastole in high HR group by subject-based analysis, consistent with segmental and vascular-based analyses.

Single slices of images examples are shown in Fig. 2, Fig. 3. The MRCA image quality at systole was slightly decreased in a patient with HR of 56bpm (Fig. 2), while the image quality was significantly improved in all coronary arteries of systolic images in a patient with a HR of 110bpm (Fig. 3).

View Large Image Download to PowerPoint

Fig. 2. Cross-section MR coronary angiographic images in 34-year-old woman (heart rate of 56bpm) obtained in the diastolic (a) and systolic (b) phases. The image quality of two phases is similar. MR: magnetic resonance; LAD: left descending artery; LCX: left circumflex artery; RCA: right coronary artery.

View Large Image Download to PowerPoint

Fig. 3. Cross-section MR coronary angiographic images in 32-year-old woman (heart rate of 110bpm) obtained in the diastolic (a) and systolic (b) phases. Image quality during systole is significantly improved compared with that in diastole, permits visualization of the coronary arteries. MR: magnetic resonance; LAD: left descending artery; LCX: left circumflex artery; RCA: right coronary artery.

3.3. Influence of heart rate variation on motion artifacts 

The mean HR variation during canning was 3.6±4.3%. There were two cases of HR variation more than 10%; including one had a HR of 56bpm with variation of 11.7% and other had 71bpm with variation of 20.1%. The image quality scores were not significantly compromised when comparing with their matching group. Yet, the influence of HR variation on the image quality in different cardiac phases of MRCA was not clearly identified due to limited cases with higher heart variation in the present study.

4.1. Study limitation 

This study enrolled young healthy volunteers; therefore the degree of stenosis was not able to be evaluated. Due to the cardiac size were unequal in different phases, the vessel lengths were unable to compare. In addition, the interpretation of coronary stenosis could be influenced by the different diameters of coronary arteries during systole and diastole, or in patients with myocardial bridge. The sample size was small and range of heart rates was limited (53–110bpm) and uneven distributed, it was insufficient to set up the threshold of heart rates in systole and diastole. Finally, although the experienced readers were blinded to the demographic data of subjects, they were aware of the cardiac phases and bias could exist.