Assessment of Left Ventricular Function and Mass on Free-Breathing Compressed Sensing Real-Time Cine Imaging

Abstract

Background: Compressed sensing (CS) cine magnetic resonance imaging (MRI) has the advantage of being inherently insensitive to respiratory motion. This study compared the accuracy of free-breathing (FB) CS and breath-hold (BH) standard cine MRI for left ventricular (LV) volume assessment.

Methods and Results: Sixty-three patients underwent cine MRI with both techniques. Both types of images were acquired in stacks of 8 short-axis slices (temporal/spatial resolution, 41 ms/1.7×1.7×6 mm³) and compared for ejection fraction, end-diastolic and systolic volumes, stroke volume, and LV mass. Both BH standard and FB CS cine MRI provided acceptable image quality for LV volumetric analysis (score ≥3) in all patients (4.7±0.5 and 3.7±0.5, respectively; P<0.0001) and had good agreement on LV functional assessment. LV mass, however, was slightly underestimated on FB CS cine MRI (median, IQR: BH standard, 83.8 mL, 64.7–102.7 mL; FB CS, 79.0 mL, 66.0–101.0 mL; P=0.0006). The total acquisition times for BH standard and FB CS cine MRI were 113±7 s and 24±4 s, respectively (P<0.0001).

Conclusions: Despite underestimation of LV mass, FB CS cine MRI is a clinically useful alternative to BH standard cine MRI in patients with impaired BH capacity.

Retrospective electrocardiogram (ECG)-gated cine magnetic resonance imaging (MRI) with breath hold (BH) is generally considered the standard method for assessment of left ventricular (LV) function.^1,2 This imaging sequence, however, requires multiple BH scans to cover the entire LV, thus prolonging the duration of cardiovascular MRI (CMR). In addition, many patients who undergo CMR have dyspnea, and difficulty tolerating BH. Poor BH technique can lead to poor image quality and low accuracy of LV functional assessment. Several acceleration techniques have been developed to reduce the BH duration in cine MRI.^3–5 Recently, compressed sensing (CS) with sparse sampling and iterative reconstruction was proven to reduce MRI acquisition time drastically.⁶ Clinical studies have reported the utility of this technique in cine MRI for assessment of LV function with 1.5-T and 3.0-T MRI scanners.^7–10 Prospective ECG-triggered CS cine MRI, however, might be unable to capture the first and last phases of the cardiac cycle, because the acquisition window is set to a fixed length at the beginning of the scan.^8–10 This often leads to underestimation of end-diastolic volume (EDV) and, consequently, variations in stroke volume (SV) and ejection fraction (EF) relative to ECG-gated retrospective standard cine MRI.¹⁰ To overcome this limitation, imaging data have been acquired over 2 heart beats, thus capturing the complete end-diastole between the first and second heart beats.¹⁰ Given that CS cine MRI has the advantage of being inherently insensitive to respiratory motion because of single-shot acquisition,^11,12 it is suitable for free-breathing (FB) imaging. The purpose of this study was therefore to compare the accuracy of FB full cardiac cycle CS and BH standard cine MRI at 3.0 T for LV volume assessment in patients with cardiac conditions.

Methods

Subjects

This prospective study was approved by the ethics review board of Saiseikai Matsuyama Hospital. All participants gave written consent for participation in the study. Seventy-one consecutive patients who were scheduled to undergo CMR for various cardiac conditions were enrolled. All participants underwent cine MRI with BH standard and FB CS sequences. Patients with arrhythmia (n=5) or severely impaired BH capacity (n=3) were excluded. Finally, 63 patients were included for analysis of image quality and LV volume. Table 1 lists the subject characteristics.

Table 1. Patient Characteristics

Characteristics	Patients (n=63)
Sex (F/M)	17/46
Age (years)	71.1±9.0
Height (cm)	160.6±9.4
Weight (kg)	61.9±12.8
Body mass index (kg/m2)	23.8±3.4
Heart rate (beats/min)	62.0±10.2
Clinical diagnosis
Coronary artery disease	42 (67)
Cardiomyopathy	13 (21)
Valve disease	3 (5)
Other	5 (8)

Data given as mean±SD or n (%).

Data Acquisition

All CMR was acquired with a clinical 3.0-T MRI scanner (MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany) with 48 receivers and 2 parallel transmit channels. Scout images were acquired to plan the cardiac axial views. Short-axis BH standard cine images were acquired in stacks of 8 contiguous slices with adequate slice gaps to cover the entire LV, using a segmented balanced steady-state free-precession sequence. Standard cine MRI was acquired during BH after inspiration, and this served as the reference standard. Immediately after BH standard cine MRI, FB CS real-time cine MRI was acquired using a prototype sequence with sparse incoherent sampling of k-space and non-linear iterative SENSE-type image reconstruction. Details regarding data acquisition and reconstruction of CS real-time cine MRI have been given in a previous report.¹⁰ In the present study, temporal and spatial resolution as well as slice orientation were kept identical between the 2 cine MRI protocols. Imaging parameters are summarized in Table 2.

Table 2. Imaging Parameters

	BH standard MRI	FB CS MRI
Sequence type	2-D cine true FISP	2-D cine true FISP
ECG mode	Retrospective gating	Prospective triggering
Repetition time (ms)	3.2	3.2
Echo time (ms)	1.4	1.4
FOV (mm)	350×350	350×350
Image matrix	208×166	208×166
Reconstructed spatial resolution (mm)	1.7×1.7	1.7×1.7
Temporal resolution (ms)	41	41
Slice thickness (mm)	6	6
No. slices	8	8
Slice gap (mm)	3.6–4.8	3.6–4.8
Flip angle (°)	50	50
Bandwidth (Hz/pixel)	1,145	960
Cardiac phases	25	19–31
No. BH	4	0
Acceleration factor	3	12.8
No. iterations	–	80

BH, breath-hold; CS, compressed sensing; ECG, electrocardiogram; FB, free-breathing; FISP, fast imaging with steady-state free precession; FOV, field of view; MRI, magnetic resonance imaging.

Image Analysis

Both types of cine images were visually assessed for image quality, with focus on the adequacy of contour detection for LV volumetry, and scored on a 5-point scale: 1, non-diagnostic; 2, poor; 3, adequate; 4, good; and 5, excellent (Figure 1). Images with score ≥3 were considered acceptable for LV volume analysis. For measurement of LV volume, stacks of 8 contiguous short-axis cine MR images were assessed independently by 2 radiologists with 7 and 5 years of experience in cardiac MRI. A dedicated software package (SYNAPSE VINCENT; Fujifilm, Tokyo, Japan) was used for quantitative assessment of EDV, end-systolic volume (ESV), SV, EF, and LV end-diastolic (LVED) mass in both cine MRI datasets. Epicardial and endocardial contours rendered by automated analysis were reviewed and manually corrected, if necessary.¹⁰ Papillary muscles and trabeculations of the LV were included in the LV chamber volume.^13,14

Figure 1.

Image quality score: representative cine images. (A) Score 1; (B) score 2; (C) score 3; (D) score 4; and (E) score 5.

Statistical Analysis

On the basis of distribution, continuous data are expressed as mean±SD or median (IQR), as appropriate. Paired t-test was used to compare scan times between the 2 methods. Comparison of image quality, EDV, ESV, SV, EF, and LVED mass between BH standard and FB CS cine MRI was done using Wilcoxon matched-pair signed-rank test. Linear regression and Bland-Altman analyses were performed to assess the correlation and agreement between these LV measurements and to evaluate inter- and intraobserver variabilities in CS cine MRI. P<0.05 was considered statistically significant. All statistical analysis was performed using commercially available software (JMP version 11; SAS Institute, Cary, NC, USA).

Sample size was calculated on the basis of the primary outcome of difference in EF between the 2 cine MRI methods. On sample size calculation analysis, 51 participants were required in order to obtain an absolute difference >6% in EF, with 80% power and a 2-sided significance level of 0.05, assuming a common SD of 15% for mean EF. This EF margin was considered as being within the clinically acceptable range on the basis of previous findings.^15–19

Results

All 63 patients successfully underwent BH standard and FB CS cine MRI. The total acquisition time for standard BH and for FB CS cine MRI was 113±7 s (range, 100–130 s; including recovery time between BH of 12 s), and 24±4 s (range, 16–34 s), respectively (P<0.0001). The image quality score of FB CS cine MRI (3.7±0.5) was worse than that of BH standard cine MRI (4.7±0.5; P<0.0001). FB CS cine MRI for all patients, however, had acceptable image quality (score ≥3) for LV volume analysis. Interobserver agreement on image quality for BH standard (kappa score, 0.73) and FB CS (kappa score, 0.72) cine MRI was good for both. Representative sets of standard BH and FB CS cine MRI of a patient with suspected myocardial ischemia are shown in Figure 2. BH standard cine MRI was used as the standard reference for LV volumetry. Comparison of EDV, ESV, SV, LVED mass, and EF between the 2 imaging methods is given in Table 3. There were no significant differences in EDV, ESV, SV, or EF between BH standard and FB CS cine MRI. Relative to BH standard cine MRI, however, a small but significant underestimation of LVED mass was noted on FB CS cine MRI. Nevertheless, there was good agreement between the 2 methods for all measurements on linear regression analysis (Figure 3). On Bland-Altman analysis, the mean differences in LV measurements between BH standard and FB CS cine MRI were as follows: EDV, 2.1 mL (95% CI: −32.6 to 36.7 mL); ESV, 2.2 mL (95% CI: −18.8 to 23.2 mL); SV, −0.2 mL (95% CI: −23.6 to 23.3 mL); LVED mass, −4.3 g (95% CI: −22.5 to 13.9 g); and EF, −1.1% (95% CI: −11.9 to 9.6%; Figure 4). The interobserver (−4.4 to 3.8%) and intraobserver (−7.9 to 4.6%) variabilities for FB CS cine MRI measurements were excellent, with slopes of regression ranging from 0.88 to 1.02 and from 0.91 to 0.97, respectively (Table 4).

Figure 2.

Representative (A,C) breath-hold (BH) standard and (B,D) free-breathing (FB) compressed sensing (CS) cine magnetic resonance imaging (MRI) of a 66-year-old female patient with suspected myocardial ischemia. (A,B) End-diastolic and (C,D) end-systolic short-axis MRI of the left ventricle. Both observers rated (A,C) BH standard CS cine MRI quality as excellent (i.e., score, 5) and (B,D) FB CS cine MRI quality as good (i.e., score, 4).

Table 3. LV Volumetry

	BH standard (n=63)	FB CS (n=63)	P value
EDV (mL)	120.5 (103.2–155.3)	124.7 (102.9–157.1)	0.37
ESV (mL)	45.3 (31.0–85.2)	48.2 (31.6–77.6)	0.11
SV (mL)	73.1 (59.4–82.9)	70.1 (62.0–83.3)	0.92
LVED mass (g)	83.8 (64.7–102.7)	79.0 (66.0–101.0)	0.0006*
EF (%)	62.2 (45.9–69)	61.2 (44.7–67.9)	0.11

Data given as median (IQR). *P<0.05. EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume; LV, left ventricular; LVED, left ventricular end-diastole; SV, stroke volume. Other abbreviations as in Table 2.

Figure 3.

Correlation between BH standard and FB CS cine magnetic resonance imaging (MRI) for parameters in the measurement of left ventricular (LV) volume: (A) end-diastolic volume (EDV); (B) end-systolic volume (ESV); (C) stroke volume (SV); (D) LV end-diastole (LVED) mass; and (E) ejection fraction (EF). Abbreviations as in Figure 2.

Figure 4.

Mean differences in left ventricle (LV) measurement between breath-hold standard and free-breathing compressed sensing cine magnetic resonance imaging: (A) EDV; (B) ESV; (C) SV; (D) LVED mass; and (E) EF. () Mean difference; () 95% CI of the mean difference; () 95% limits of agreement (i.e., mean ±1.96 SD). Abbreviations as in Figure 3.

Table 4. Variability of FB CS Cine MRI Measurement

	Difference	Variability (%)	R2	Slope	P value
Intraobserver variability
EDV (mL)	−0.8±4.6	−0.6±3.4	0.99	1.02	<0.001
ESV (mL)	−2.9±5.9	−4.4±8.8	0.98	0.97	<0.001
SV (mL)	2.0±6.1	3.8±10.5	0.89	0.88	<0.001
LVED mass (g)	−2.4±6.1	−3.0±7.2	0.96	0.98	<0.001
EF (%)	1.7±3.5	−4.4±9.8	0.96	0.94	<0.001
Interobserver variability
EDV (mL)	−0.4±5.7	−0.2±4.2	0.98	0.97	<0.001
ESV (mL)	−3.4±5.5	−5.3±10.9	0.99	0.97	<0.001
SV (mL)	3.0±6.1	4.4±9.3	0.87	0.97	<0.001
LVED mass (g)	−6.9±7.1	−7.9±8.3	0.95	0.91	<0.001
EF (%)	2.3±3.9	4.6±7.7	0.94	0.95	<0.001

Data given as mean±SD. Abbreviations as in Tables 2,3.

Discussion

Our previous study demonstrated that single BH CS real-time cine MRI could evaluate LV volume with excellent accuracy using acquisition of CS cine MRI data over 2 heart beats.¹⁰ Given, however, that many patients who undergo CMR have dyspnea or shortness of breath, image acquisition over 2 heart beats increases the required BH time as well as patient burden. FB MRI could be an alternative approach in patients who cannot hold their breath adequately. In this study, we evaluated the accuracy of FB full cardiac cycle CS cine MRI in LV volumetry. FB CS cine MRI LV volumetry was in good agreement with that of BH standard cine MRI. There were no significant differences in EDV, ESV, SV, or EF between the 2 methods. This suggests that it is possible to capture the complete end-diastole between the first and second heart beats by acquiring data over 2 heart beats and to successfully trace endocardial contours on CS cine MRI even during FB. In comparison with standard cine MRI, standard real-time cine MRI usually has reduced spatial and temporal resolution, which might lead to inaccuracy in the evaluation of cardiac function with highly accelerated cine MRI.²⁰ In this study, we achieved identical temporal and spatial resolution in both the reference standard technique and real-time cine MRI using the CS technique, which was reflected in the good agreement for LV volumetry between the 2 techniques.

With regard to differences in LV volumetry between the 2 techniques, the 95% CI for all LV volume measurements obtained with FB CS cine MRI were wider compared with BH CS cine MRI in our previous study.¹⁰ These differences might be explained by the difference in breathing during imaging. Given that BH standard cine MRI was acquired at inspiration, the effect of negative intrathoracic pressure on LV volume²¹ and the slight differences in short-axis slice orientation between BH and FB MRI²² might have had some effect on this result. The blurry image due to FB might also slightly impair the accuracy of the determination of endocardial contour. The differences, however, between BH standard and FB CS cine MRI were within the acceptable range for LV volumetry.

Relative to standard BH cine MRI, we observed a significant underestimation of LVED mass in FB CS cine MRI. For measurement of LVED mass, both endocardial and epicardial contours are defined at end-diastole. In the present study, there were no significant differences in EDV between the 2 methods. This therefore suggests that the underestimation of LVED mass on FB CS cine MRI might have been mainly caused by inaccuracy in the definition of epicardial contours. One of the reasons for this inaccuracy might be the lower contrast between the myocardium and outer surrounding structures, such as the lungs or liver, in comparison with the contrast between the myocardium and LV cavity in balanced steady-state free precession cine MRI. Furthermore, the high sub-sampling feature of CS might lead to reduced definition of borders, which would make accurate delineation of the epicardial border even more challenging. The combination of these effects might have led to the significant underestimation of LVED mass on FB CS cine MRI in the present study.

In comparison with BH standard cine MRI, FB CS cine MRI drastically reduced the total acquisition time. Although BH standard cine MRI is a well-established technique with proven clinical reliability, rapid FB CS cine MRI is more time efficient and beneficial for patients with difficulty in holding breath.

In this study, we excluded patients with arrhythmia or severely impaired BH capacity (8/71 enrolled patients; 11%) because the reference sequence used for LV volume analysis relies on regular heart beat and adequate BH. Standard cine MRI of such patients has been reported to often exhibit severe artefacts, which makes LV volume analysis challenging.^23,24 CS real-time cine MRI is a useful approach in such patients because single-shot acquisition is inherently insensitive to motion artefacts caused by arrhythmia or breathing.^11,12 We believe that the CS technique can improve the adaptability of CMR and help promote its widespread clinical use.

There are some limitations to this study. First, FB CS cine MRI was reconstructed using the MR scanner central processing unit at the end of data acquisition. The reconstruction time for all short-axis slices was approximately 3 min. Shorter reconstruction times, however, are desirable for achieving real-time visualization. We expect that more efficient CS reconstruction techniques with graphic processing unit-based reconstruction algorithms²⁵ can reduce the reconstruction time and overcome this limitation. Second, although the differences between the measurements of BH standard and FB CS cine MRI were within the acceptable range, there were relatively wide 95% limits of agreement in EDV and EF measurements. In this study, we used identical 8-slice short-axis cine MRI, performed as routine examination. The slice number is smaller than that in the previous report by Sudarski et al.⁷ When a smaller number of cine slice images is used to evaluate LV volume, the difference in determination of the lowest slice has a larger effect on the variability of measurement. Hence, in the present study this small slice number affected the accuracy of the determination of the lowest slice, and increased the difference in EDV and EF. We expect that the variability in the measurement can be reduced by increasing the slice number. Finally, given that regional myocardial wall motion was not assessed in the present study, further evaluation is required to determine the accuracy of FB CS cine MRI for assessment of regional myocardial wall motion abnormalities.

Conclusions

FB CS cine MRI might be a suitable alternative to standard BH cine MRI for assessment of LV function for patients who cannot tolerate BH. The CS technique is time efficient and can potentially improve the clinical utility of CMR.

Acknowledgments

The authors are grateful to Yoshiaki Komori and Yuta Urushibata (Siemens Healthcare, Tokyo, Japan) for their help with optimization of sequence parameters and image quality.

Disclosures

C.F. and M.S. are employees of Siemens Healthcare. The other authors declare no conflict of interest.

References

• 1.   Moon JC, Lorenz CH, Francis JM, Smith GC, Pennell DJ. Breath-hold FLASH and FISP cardiovascular MR imaging: Left ventricular volume differences and reproducibility. Radiology 2002; 223: 789–797.
• 2.   Grothues F, Smith GC, Moon JC, Bellenger NG, Collins P, Klein HU, et al. Comparison of interstudy reproducibility of cardiovascular magnetic resonance with two-dimensional echocardiography in normal subjects and in patients with heart failure or left ventricular hypertrophy. Am J Cardiol 2002; 90: 29–34.
• 3.   Kozerke S, Plein S. Accelerated CMR using zonal, parallel and prior knowledge driven imaging methods. J Cardiovasc Magn Reson 2008; 10: 29.
• 4.   Jahnke C, Nagel E, Gebker R, Bornstedt A, Schnackenburg B, Kozerke S, et al. Four-dimensional single-breath-hold magnetic resonance imaging using kt-BLAST enables reliable assessment of left- and right-ventricular volumes and mass. J Magn Reson Imaging 2007; 25: 737–742.
• 5.   Eberle HC, Nassenstein K, Jensen C, Schlosser T, Sabin GV, Naber CK, et al. Rapid MR assessment of left ventricular systolic function after acute myocardial infarction using single-breath-hold cine imaging with the temporal parallel acquisition technique (TPAT) and 4D guide-point modelling analysis of left ventricular function. Eur Radiol 2010; 20: 73–80.
• 6.   Lustig M, Donoho D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn Reson Med 2007; 58: 1182–1195.
• 7.   Sudarski S, Henzler T, Haubenreisser H, Dösch C, Zenge MO, Schmidt M, et al. Free-breathing sparse sampling cine MR imaging with iterative reconstruction for the assessment of left ventricular function and mass at 3.0 T. Radiology 2017; 282: 74–83.
• 8.   Feng L, Srichai MB, Lim RP, Harrison A, King W, Adluru G, et al. Highly accelerated real-time cardiac cine MRI using k-t SPARSE-SENSE. Magn Reson Med 2013; 70: 64–74.
• 9.   Vincenti G, Monney P, Chaptinel J, Rutz T, Coppo S, Zenge MO, et al. Compressed sensing single-breath-hold CMR for fast quantification of LV function, volumes, and mass. JACC Cardiovasc Imaging 2014; 7: 882–892.
• 10.   Kido T, Kido T, Nakamura M, Watanabe K, Schmidt M, Forman C, et al. Compressed sensing real-time cine cardiovascular magnetic resonance: Accurate assessment of left ventricular function in a single-breath-hold. J Cardiovasc Magn Reson 2016; 18: 50.
• 11.   Voit D, Zhang S, Unterberg-Buchwald C, Sohns JM, Lotz J, Frahm J. Real-time cardiovascular magnetic resonance at 1.5 T using balanced SSFP and 40 ms resolution. J Cardiovasc Magn Reson 2013; 15: 79.
• 12.   Aandal G, Nadig V, Yeh V, Rajiah P, Jenkins T, Sattar A, et al. Evaluation of left ventricular ejection fraction using through-time radial GRAPPA. J Cardiovasc Magn Reson 2014; 16: 79.
• 13.   van Geuns RJ, Baks T, Gronenschild EH, Aben JP, Wielopolski PA, Cademartiri F, et al. Automatic quantitative left ventricular analysis of cine MR images by using three-dimensional information for contour detection. Radiology 2006; 240: 215–221.
• 14.   Sievers B, Kirchberg S, Bakan A, Franken U, Trappe HJ. Impact of papillary muscles in ventricular volume and ejection fraction assessment by cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2004; 6: 9–16.
• 15.   Pattynama PM, Lamb HJ, Van der Velde EA, van der Wall EE, de Roos A. Left ventricular measurements with cine and spin echo MR imaging: A study of reproducibility with variance component analysis. Radiology 1993; 187: 261–268.
• 16.   Cohn PF, Levine JA, Bergeron GA, Gorlin R. Reproducibility of the angiographic left ventricular ejection fraction in patients with coronary artery disease. Am Heart J 1974; 88: 713–720.
• 17.   Gordon EP, Schnittger I, Fitzgerald PJ, Williams P, Popp RL. Reproducibility of left ventricular volumes by two-dimensional echocardiography. J Am Coll Cardiol 1983; 2: 506–513.
• 18.   Dorosz JL, Lezotte DC, Weitzenkamp DA, Allen LA, Salcedo EE. Performance of 3-dimensional echocardiography in measuring left ventricular volumes and ejection fraction: A systematic review and meta-analysis. J Am Coll Cardiol 2012; 59: 1799–1808.
• 19.   Malm S, Frigstad S, Sagberg E, Larsson H, Skjaerpe T. Accurate and reproducible measurement of left ventricular volume and ejection fraction by contrast echocardiography: A comparison with magnetic resonance imaging. J Am Coll Cardiol 2004; 44: 1030–1035.
• 20.   Muthurangu V, Lurz P, Critchely JD, Deanfield JE, Taylor AM, Hansen MS. Real-time assessment of right and left ventricular volumes and function in patients with congenital heart disease by using high spatiotemporal resolution radial k-t SENSE. Radiology 2008; 248: 782–791.
• 21.   Buda AJ, Pinsky MR, Ingels NB Jr, Daughters GT 2nd, Stinson EB, Alderman EL. Effect of intrathoracic pressure on left ventricular performance. N Engl J Med 1979; 301: 453–459.
• 22.   Hori Y, Yamada N, Higashi M, Hirai N, Nakatani S. Rapid evaluation of right and left ventricular function and mass using real-time true-FISP cine MR imaging without breath-hold: Comparison with segmented true-FISP cine MR imaging with breath-hold. J Cardiovasc Magn Reson 2003; 5: 439–450.
• 23.   Klinke V, Muzzarelli S, Lauriers N, Locca D, Vincenti G, Monney P, et al. Quality assessment of cardiovascular magnetic resonance in the setting of the European CMR Registry: Description and validation of standardized criteria. J Cardiovasc Magn Reson 2013; 15: 55.
• 24.   Marks B, Mitchell DG, Simelaro JP. Breath-holding in healthy and pulmonary-compromised populations: Effects of hyperventilation and oxygen inspiration. J Magn Reson Imaging 1997; 7: 595–597.
• 25.   Saybasili H, Herzka DA, Seiberlich N, Griswold MA. Real-time imaging with radial GRAPPA: Implementation on a heterogeneous architecture for low-latency reconstructions. Magn Reson Imaging 2014; 32: 747–758.