Full free-breathing cardiovascular magnetic resonance imaging: enhancing efficiency and image quality in clinical practice

Abstract

Background

Conventional cardiovascular magnetic resonance (CMR) examinations require patients to repeatedly hold their breath, which can reduce examination efficiency and pose challenges for patients unable to do so. This study aimed to demonstrate the feasibility and effectiveness of a full free-breathing CMR protocol in clinical practice.

Methods

Patients prospectively enrolled in this study underwent a full free-breathing CMR exam on a 3T scanner between June 1 and June 30, 2024. Acquisition time and image quality were assessed. Cine and flow imaging were compared with those acquired with the conventional breath-holding CMR protocol. Other sequences, including T1/T2 mapping and late gadolinium enhancement (LGE), were evaluated quantitatively and qualitatively, respectively. Group comparisons were performed using the Wilcoxon signed-rank test or paired t-test. Consistency was assessed using Kappa statistics, Bland–Altman statistics, intraclass correlation coefficient (ICC), and linear regression.

Results

A total of 211 patients were evaluated (median age: 53 years [IQR: 38–63]; range: 10–82 years; 145 men). The mean acquisition time for full free-breathing CMR was 22.6 ± 3.7 min. The median image quality scores for cine and LGE images acquired with free-breathing CMR were 4 (IQR: 4–4) and 5 (IQR: 4–5), respectively. Compared with conventional breath-holding CMR, the end-diastolic volume (EDV), end-systolic volume (ESV), EDV index, and ESV index measured by free-breathing CMR were slightly higher (all P＜0.05), whereas the left ventricular ejection fraction and left ventricular mass were slightly lower (both P＜0.05). Nonetheless, the two methods demonstrated good agreement and correlation (r values: 0.85–0.99). Native T1 and T2 values in healthy subjects from free-breathing CMR were 1214.9 ± 16.7 ms and 38.4 ± 3.2 ms, respectively. Among the 211 patients, 147 were LGE positive. Except for five patients with image quality scores below 3, all others had scores of 3 or higher.

Conclusion

Full free-breathing CMR examinations are feasible and effective in clinical practice, significantly reduce scan time while maintaining high image quality.

Graphical abstract

1. Introduction

Cardiac magnetic resonance (CMR) imaging provides a comprehensive, "one-stop" evaluation of the cardiac structure, function, and myocardial tissue characteristics. Recognized as the gold standard for evaluating cardiac anatomy and function, CMR has become increasingly vital for the diagnosis and prognosis of cardiovascular diseases [1], [2]. Numerous studies have confirmed that CMR plays an important role in the diagnosis, differential diagnosis, and risk stratification of cardiovascular diseases [3], [4], [5], [6], [7]. However, CMR often involves multiple imaging protocols and planes, resulting in lengthy examination times. Although the actual image acquisition time is relatively short, repeated breath-holds and the recovery periods between them significantly extend the overall duration—accounting for nearly 50% of the total scan time. Furthermore, for patients unable to hold their breath, such as those with severe cardiac insufficiency, the elderly, individuals with hearing impairments, or infants under anesthesia, it is difficult to complete the examination and obtain high-quality images. Thus, optimizing the CMR scanning process to enable continuous, full free-breathing scans with reliable image quality is of great clinical value. In recent years, advancements in magnetic resonance imaging have enabled the use of compressed sensing (CS) and motion correction (MOCO) techniques to facilitate free-breathing CMR scans [8], [9], [10], [11], [12]. Furthermore, deep learning-based methods have been developed to facilitate free-breathing image acquisition [13], [14]. However, most existing studies have primarily focused on evaluating individual sequences, such as cardiac cine or late gadolinium enhancement (LGE). In contrast, this study introduces a new full free-breathing CMR protocol, including cardiac cine, flow imaging, T1 and T2 mapping, and LGE imaging, and investigates its feasibility and clinical utility.

2. Methods

2.1. Study population

A total of 211 consecutive patients who underwent a full free-breathing CMR examination at our hospital between June 1 and June 30, 2024 were prospectively enrolled. All patients were randomized without pre-screening prior to the examination. The cohort included 145 males (68.7%) and 66 females (31.3%), with a median age of 53.0 years (interquartile range: 38.0–62.8 years; age range: 10–82 years). All patients underwent a full free-breathing CMR examination, including cine, T1 mapping, T2 mapping, flow imaging, rest perfusion, and LGE scans. In addition, 154 patients also underwent conventional breath-holding cine and flow imaging the day before, allowing for direct comparison with the corresponding free-breathing sequences. The remaining free-breathing protocols were not compared with their breath-hold counterparts. This study complied with the Declaration of Helsinki, was approved by the Ethics Committee of Fuwai Hospital. Written informed consent was obtained from all patients, and for minors, consent was obtained from their parents or legal guardians.

2.2. Image acquisition

CMR examinations were performed on a 3T scanner (MAGNETOM Vida, Version VA202A, Siemens Healthineers, Forchheim, Germany) using an 18-channel body coil. Image acquisition was performed during free-breathing using motion correction (MOCO) and optimized parameters for cardiac cine imaging, T1 and T2 mapping, perfusion and LGE (Fig. 1). The imaging protocol is designed to be able to adapt to the age and body type of the subject. For younger and smaller subjects, we adjusted imaging parameters, such as field of view, to optimize image quality while minimizing potential risks associated with CMR scans. The detailed scanning sequences are as follows: (1) cardiac cine imaging, prototype sequence with balanced steady-state free precession (bSSFP) readout, compressed sensing (CS) acceleration and respiratory MOCO was used to collect the left ventricular two-chamber, four-chamber, left ventricular outflow tract and multi-slices short-axis cine images covering the entire left ventricle [11]; (2) T1 mapping, using the modified Look-Locker inversion recovery (MOLLI) sequence, three levels of the base, middle and apex of the left ventricle short-axis view were collected [15], [16]; (3) T2 mapping [17], T2 preparation sequence was used, and the collection levels were the same as T1 mapping; (4) flow imaging, was acquired using a 2D phase contrast technique with a gradient echo (GRE) readout, which was acquired by simply averaged over several acquisitions, including four-chamber and left ventricular outflow tract in-plane flow acquisition, the velocity encoding range was set at 150 cm/s; (5) rest perfusion and LGE imaging, a bolus of 0.2 mmol/kg of Gd-DTPA (Gadovist, Bayer, Berlin, Germany) was administered intravenously at 4 mL/s. An initial dose of 0.05–0.1 mmol/kg was used for the rest perfusion scan, with the remainder administered 90 s thereafter. The rest perfusion was acquired by turbo fast low-angle shot (Turbo FLASH) sequence. After about 10 min, delayed enhanced imaging was performed using phase-sensitive inversion recovery (PSIR) sequences with MOCO [18], including left ventricular two-chamber, four-chamber, and multi-slice short-axis images. In these sequences, MOCO cine and mapping are research sequences, and the rest are commercial products. Detailed scanning parameters are shown in Table 1.

Fig. 1.

The workflow of a full free-breathing CMR examination. ECG electrocardiogram, PSIR phase-sensitive inversion recovery sequence, CMR cardiovascular magnetic resonance

Table 1.

Detailed scanning parameters for free-breathing CMR examination.

Parameters	BH Seg cine	FB MOCO cine	FB MOCO LGE	FB MOCO T1 mapping	FB MOCO T2 mapping	FB Flow	FB MOCO Perfusion
TR/TE (ms)	3.33/1.46	3.30/1.45	2.73/1.15	2.64/1.09	3/1.25	5.06/2.92	2.05/0.98
FOV (mm)	380*344	380*344	380*344	380*327	380*325	360*270	400*367
Image matrix	256*213	240*200	256*174	256*164	192*124	224*134	192*130
Reconstructed spatial resolution (mm)	1.48*1.48	1.58*1.58	1.48*1.48	1.48*1.48	1.98*1.98	1.61*1.61	2.08*2.08
Slice thickness (mm)	8	8	8	8	8	6	8
Flip angle (degrees)	min 40， as large as possible	min 40， as large as possible	55	35	12	20	10
Bandwidth (Hz/pixel)	930	947	1085	1085	1184	456	1002
Temporal resolution (ms)	40	46.2	237	187	138	50.6	162.6
Segments	12	14	87	82	62	5	65
Average	1	12	8	1	1	3	-
Total heartbeats	10	12	16	11	9	31	80
Acceleration factor	3	14.3	2	2	2	2	2
Magnetization preparation			TI range, 300–350 ms	TImin100, ΔTI 80 ms	T2prep 0, 35, 55 ms		TS 95 ms
Readout	bSSFP	bSSFP	bSSFP	bSSFP	GRE	GRE	GRE

CMR cardiovascular magnetic resonance, BH breath-holding, FB free-breathing, MOCO motion correction, LGE late gadolinium enhancement, TR repetition time, TE echo time, FOV field of view, TS saturation recovery time, bSSFP balanced Steady-State Free Precession, GRE gradient recalled echo

2.3. Image quality assessment

The image quality score was evaluated visually based strictly on a 5-point Likert score ranging from 1 to 5: 5=excellent image quality, minimal to no motion-related artifacts or image distortion; 4=good image quality, mild motion-related artifacts or image distortion; 3=adequate image quality, moderate motion-related artifacts or image distortion but sufficiently diagnostic for ventricular contour tracing; 2=fair image quality, marked motion-related artifacts or image distortion but partially diagnostic; and 1=completely nondiagnostic images. An image quality score≥3 was considered clinically diagnostic. Image quality was assessed by a radiologist with over 5 years of experience in cardiovascular imaging. For inter-reader agreement, images from 30 randomly selected patients were independently evaluated by a second senior radiologist with at least 10 years of diagnostic experience, who was blind to the first reader’s results.

2.4. Image analysis

Circle Cardiovascular Imaging CVI42 software (Circle Cardiovascular Imaging, Calgary, Alberta, Canada) was used by two imaging technicians to measure the parameters of left ventricular function under free-breathing and conventional breath-holding, and the two technicians were blind to each other’s results. Specific CMR parameters include left ventricular ejection fraction (LVEF), cardiac output (CO), left ventricular end-diastolic volume (LVEDV), left ventricular end-diastolic volume index (LVEDVi), left ventricular end-systolic volume (LVESV), left ventricular end-systolic volume index (LVESVi), stroke volume (SV), and left ventricular mass (LVM). T1, T2, and extracellular volume fraction (ECV) values were measured manually at the base, middle, and apex levels of the interventricular septum. 1–2 areas of interest were selected for each level, and then the average value was calculated. During the measurement, the myocardium close to the blood pool was avoided to prevent the influence of the blood pool on the measured values. The severity of regurgitation was assessed using a visual semi-quantitative method and categorized as mild, moderate, or severe based on the intensity of the regurgitant signal. To better demonstrate our findings, we focused on the flow imaging of the patients with primary valvular heart disease (VHD) and compared it with echocardiography. The evaluation of rest perfusion images was only qualitative. According to myocardial perfusion conditions, it was divided into reduced perfusion and no obvious perfusion abnormality, and quantitative measurement was not used. The qualitative evaluation of LGE included the presence of LGE, LGE site, and patterns such as subendocardial, subepicardial, intramural or transmural enhancement. All of these parameters were assessed by at least two physicians, one with >10 years of experience.

2.5. Statistical analyses

All statistical analyses were performed with IBM SPSS Statistics 20.0 (IBM Corporation, Armonk, New York) and MedCalc 20.022 software (MedCalc Software Ltd, Ostend and Belgium). Continuous variables are presented as mean ± standard deviation (SD) in the case of normal distribution and median (interquartile range, IQR) otherwise. Categorical variables are presented as the frequency and percentage. The scan time and image quality scores were compared using the Wilcoxon matched-pairs signed-rank test. The quantitative results of left ventricular function parameters were compared by the Wilcoxon matched-pairs signed-rank test or paired t test according to the tests of normality and homogeneity of variances. Linear regression and Bland–Altman analyses were used to evaluate the correlation and agreement of quantitative parameters between different techniques. Cohen’s kappa statistics were calculated to evaluate inter-observer variability in subjective image quality scores. Intra- and inter-observer reliability were assessed using the intraclass correlation coefficient (ICC) and Bland–Altman analyses. A P value < 0.05 was considered statistically significant.

3. Results

3.1. Basic clinical characteristics of the study population

A total of 211 patients were included in the study, including 145 males (68.7%) and 66 females (31.3%). The median age of patients was 53.0 years (IQR: 38.0–62.8 years; range: 10–82 years). All patients underwent a full free-breathing CMR examination, and 154 patients also underwent cine and flow imaging in a conventional breath-holding state the day before. Forty-eight patients had arrhythmia during the examination, and ten patients could not effectively cooperate with breath-holding CMR examination. Around 201 patients had definite cardiovascular diseases, including hypertrophic cardiomyopathy (HCM), dilated cardiomyopathy (DCM), coronary artery disease (CAD), and so on, all of these diseases were diagnosed based on CMR features and clinical information. And ten cases had no definite signs of cardiac abnormality. Other basic clinical characteristics are shown in Table 2.

Table 2.

Basic clinical characteristics of the study population (n = 211).

Characteristic	Results
Age (y), median (IQR)	53.0 (38.0, 62.8)
Male, n (%)	145 (68.7)
Height (m), median (IQR)	1.70 (1.62, 1.75)
Weight (kg), mean ± SD	72.35 ± 14.07
Body mass index (kg/m²), mean±SD	25.38 ± 3.98
Heart rate (bpm), median (IQR)	66.0 (58.0, 76.0)
Diagnosis, n (%)
Hypertrophic cardiomyopathy	59 (28.0)
Dilated cardiomyopathy	34 (16.1)
Hypertensive heart disease	25 (11.8)
Coronary artery disease	20 (9.5)
Valvular heart disease	16 (7.6)
Arrhythmia-related heart disease	11 (5.2)
Arrhythmogenic cardiomyopathy	10 (4.7)
Other cardiac diseases†	26 (12.3)
Healthy subjects	10 (4.7)
Arrhythmia during the examination, n (%)	48 (22.7)*
Poor breath-holding during the examination, n (%)	10 (6.5)#

^†Other cardiac diseases include cardiac amyloidosis, congenital heart disease, left ventricular noncompaction, myocarditis, cardiac mass, and so on. IQR interquartile range

^*In 211 cases with free-breathing CMR examination, while in 154 cases with conventional breath-holding CMR examination, there were 29 cases (18.8%) had arrhythmia during the examination. CMR cardiovascular magnetic resonance

^#In 154 cases who also underwent breath-holding CMR examination

3.2. Scan time and image quality of cine

The average total scan time for 211 patients with full free-breathing CMR was 22.6 ± 3.7 min, with short-axis cine images being 2.8 ± 0.4 min. In 154 patients who also underwent breath-holding CMR examination, the average acquisition time of short-axis cine was 4.8 ± 1.4 min, which was significantly longer than that of free-breathing CMR examination (P＜0.001).

The median image quality score for free-breathing short-axis cine imaging was 4 (IQR: 4–4) points. Among the 154 patients who also underwent breath-holding CMR, the image quality score for short-axis cine was 5 (IQR: 4–5) points, which was statistically higher than that of the free-breathing short-axis cine imaging (P＜0.001). Notably, no free-breathing short-axis cine scan received an image quality score below 3 points. Typical cases of image quality points are shown in Fig. 2. Among the 30 randomly selected patients, the inter-reader agreement for the image quality score of free-breathing short-axis cine imaging demonstrated a weighted kappa of 0.88.

Fig. 2.

Typical images display with different scores ranging from 2 to 5. A-D, cardiac cine images, E-H, LGE images. Our results lack images with an image quality score of 1 point, and the scores of cardiac cine images obtained by free-breathing CMR were all greater than or equal to 3 points. Panel A was obtained by conventional breath-holding CMR examination, and Panels B-H were obtained by free-breathing CMR examination. LGE late gadolinium enhancement, CMR cardiovascular magnetic resonance

It is worth mentioning that among the 154 patients who underwent breath-holding short-axis cine imaging, 10 patients had poor breath-holding capacity, and 29 patients had arrhythmia during the examination. Among these 10 patients, the image quality scores were 3.7 and 4.3 points for breath-holding and free-breathing short-axis cine imaging, respectively (P = 0.107). Among the 29 patients with arrhythmia, the image quality scores were 3.6 and 4.1 points for breath-holding and free-breathing short-axis cine imaging, respectively (P = 0.016).

3.3. Quantitative measurements of left ventricular function parameters

The values of EDV, ESV, EDVI, and ESVI measured by free-breathing CMR were slightly higher than those measured by breath-holding, while the values of LVEF and LVM were slightly lower than those measured by breath-holding (Table 3). Although there was a statistical difference between the two groups, the difference between the two groups was small, and the Bland–Altman plots showed good agreement between left ventricular function parameters measured by the free-breathing CMR and breath-holding CMR. Linear regression showed a strong positive correlation between breath-holding and free-breathing CMR for all the left ventricular function parameters. The data distribution of the two groups of measurements is shown in Fig. 3. In 30 randomly selected patients, the intra- and inter-reader variabilities of left ventricular function parameters obtained by free-breathing CMR are shown in Table 4, and their ICCs were excellent, ranging from 0.9982 to 0.9998 and 0.9956 to 0.9998, respectively. The Bland–Altman plots of intra- and inter-group consistency are shown in Supplementary materials (Fig. S1, S2).

Table 3.

Comparison of left ventricular function parameters under breath-holding and free-breathing state (n = 154).

	Breath-holding mode	Free-breathing mode	P value
EDV (mL)	170.0 (136.8, 226.8)	171.1 (137.6, 233.1)	＜0.001
ESV (mL)	77.1 (49.9, 152.0)	81.6 (52.8, 158.3)	＜0.001
EDVI (mL/m²)	90.6 (72.8, 126.0)	92.0 (77.1, 128.4)	＜0.001
ESVI (mL/m²)	42.4 (26.8, 84.1)	46.4 (27.9, 88.3)	＜0.001
SV (mL)	84.2 (65.0, 99.4)	79.1 (64.3, 98.6)	0.309
CO (L/min)	5.3 (4.3, 6.6)	5.1 (4.2, 6.5)	0.526
EF (%)	51.1 (32.5, 64.7)	49.7 (31.1, 63.4)	＜0.001
LV Mass(g)	117.1 (93.0, 145.8)	112.1 (90.0, 144.5)	0.011

EDV end-diastolic volume, ESV end-systolic volume, EDVI end-diastolic volume index, ESVI end-systolic volume index, SV stroke volume, CO cardiac output, EF ejection fraction, LV left ventricular

Data are means ± standard deviation or medians (interquartile range).

Fig. 3.

Scatter plots, Bland–Altman plots, and linear regression plots for left ventricular function parameters obtained by breath-holding and free-breathing CMR. The first column was scatter plots, the second was Bland–Altman plots, and the third was linear regression plots. 1 represents the value measured by the breath-holding CMR, and 2 represents the value measured by the free-breathing CMR. EDVI end-diastolic volume index, LVEF left ventricular ejection fraction, LVM left ventricular mass, CMR cardiovascular magnetic resonance

Table 4.

Intra- and inter-observer variability for left ventricular function parameters of free-breathing CMR examination (n = 30).

	Intra-observer consistency		Inter-observer consistency
	ICC	95% CI	ICC	95% CI
EDV (mL)	0.9998	0.9995–0.9999	0.9996	0.9984–0.9998
ESV (mL)	0.9998	0.9995–0.9999	0.9998	0.9993–0.9999
EDVI (mL/m²)	0.9997	0.9991–0.9999	0.9997	0.9991–0.9999
ESVI (mL/m²)	0.9998	0.9995–0.9999	0.9996	0.9991–0.9998
SV (mL)	0.9982	0.9963–0.9992	0.9956	0.9906–0.9979
CO (L/min)	0.9984	0.9966–0.9992	0.9961	0.9918–0.9981
LVEF (%)	0.9993	0.9986–0.9997	0.9988	0.9975–0.9994
LVM (g)	0.9993	0.9984–0.9997	0.9987	0.9972–0.9994

CMR cardiac magnetic resonance, ICC intraclass correlation coefficient, CI confidence interval, EDV end-diastolic volume, ESV end-systolic volume, EDVI end-diastolic volume index, ESVI end-systolic volume index, SV stroke volume, CO cardiac output, LVEF left ventricular ejection fraction, LVM left ventricular mass

3.4. T1 mapping, T2 mapping and ECV

All 211 patients underwent T1 mapping and T2 mapping imaging under a full free-breathing state, the image quality scores of native T1, T2, and ECV maps were 5 (4, 5), 5 (4, 5), and 4 (4, 5), respectively. It should be noted that ECV maps could not be generated for 6 patients due to a slice mismatch between the pre- and post-contrast T1 maps, so the analysis of ECV results was based on 205 patients. There were ten healthy individuals, whose native T1, T2, and ECV values were 1214.9 ± 16.7 ms, 38.4 ± 3.2 ms, and 24.8 ± 2.3%, respectively. For other cardiovascular diseases, the distribution of T1, T2, and ECV values is shown in Table 5 and Supplementary materials (Figs. S3-5).

Table 5.

The native T1, T2, and ECV values in different cardiovascular diseases and healthy subjects.

	Native T1 value (ms)	T2 value (ms)	ECV value (%)
Arrhythmogenic cardiomyopathy	1284.5 ± 28.1	39.2 ± 1.8	26.6 ± 2.0
Arrhythmia-related heart disease	1248.2 ± 43.6	39.4 ± 2.3	25.4 ± 2.0
Cardiac amyloidosis	1503.3 ± 36.1	48.9 ± 2.9	49.3 ± 6.4
Coronary artery disease	1291.0 ± 35.5	39.8 ± 2.0	33.1 ± 6.1
Dilated cardiomyopathy	1276.7 ± 37.0	39.8 ± 2.0	30.0 ± 4.9
Hypertrophic cardiomyopathy	1280.1 ± 48.4	40.0 ± 2.3	29.0 ± 5.2
Hypertensive heart disease	1270.0 ± 46.6	39.2 ± 2.2	27.7 ± 4.4
Myocarditis	1243.3 ± 48.4	40.6 ± 3.8	31.3 ± 4.4
Healthy subjects	1214.9 ± 16.7	38.4 ± 3.2	24.8 ± 2.3
Valvular heart disease	1294.7 ± 57.8	40.4 ± 2.1	30.4 ± 4.5
Other*	1261.1 ± 50.0	40.3 ± 3.5	27.9 ± 3.0

^*Other include congenital heart disease, left ventricular noncompaction, and cardiac mass, and so on. ECV extracellular volume fraction

3.5. Flow imaging

In this study, all 211 patients underwent free-breathing flow imaging, which routinely included four-chamber and left ventricular outflow tract flow imaging. Besides, 154 patients also underwent conventional breath-holding flow imaging the day before, which was compared to free-breathing flow imaging. The image quality scores of breath-holding and free-breathing flow imaging were 4 (4, 5) and 4 (4, 5) points, respectively, with no significant difference between groups”. In these 154 patients with both breath-holding and free-breathing flow imaging, the results were consistent between the two groups in 135 cases, the weighted kappa of the agreement between breath-holding and free-breathing flow imaging was 0.84. For specific details of valve lesions, please refer to the supplementary materials (Table S1). Furthermore, among the 16 patients with primary VHD, 9 had aortic insufficiency (Fig. 4), 3 had aortic stenosis, 2 had mitral valve insufficiency, and 2 had tricuspid insufficiency. The results were consistent with echocardiography, and the weighted kappa was 0.90. Typical cases are shown in Supplementary materials (Fig. S6).

Fig. 4.

A 62-year-old male patient with valvular heart disease underwent a full free-breathing CMR. A-C, cardiac cine imaging, the left ventricle is significantly enlarged (left ventricle diameter about 70 mm), interventricular septal thickness about 9–10 mm. D-F indicates the pre-T1 map, T2 map, and ECV map, respectively. G-I, flow imaging, moderate regurgitation signal is seen in the aortic valve during diastole (regurgitation fraction about 43%, panels G and I, red arrows), and the aortic valve normally opens during systole (Panel H). J-L, LGE imaging, a little intramural halo enhancement can be seen in the proximal interventricular septum. CMR cardiovascular magnetic resonance, ECV extracellular volume fraction, LGE late gadolinium enhancement

3.6. Rest perfusion imaging

All 211 patients had MOCO perfusion imaging, and the image quality was excellent with 5 (4, 5) points. 39 patients (18.5%) had hypoperfusion, of which 18 had CAD, 11 had HCM, and 10 had other cardiomyopathies, such as DCM and cardiac amyloidosis (CA). The segments with reduced perfusion corresponded to the areas with delayed enhancement. No significant perfusion abnormalities were observed in other patients.

3.7. LGE imaging

All 211 patients underwent a full free-breathing LGE imaging, with an average acquisition time of 4.1 ± 0.8 min. The median image quality score for the LGE images was 5 (IQR: 4–5), with only five patients receiving a score below 3. Among 30 randomly selected patients, the inter-observer agreement for the image quality score of free-breathing LGE images showed a weighted kappa of 0.87. There were 147 patients with LGE positive. Among them, there were 109 cases of intramural enhancement, most of which were found in non-ischemic cardiomyopathies, such as HCM, DCM and hypertensive heart disease (HHD, Fig. 5). There were 30 cases of subendocardial or transmural enhancement, which were common in coronary artery disease (Fig. 6). And there were 8 cases of subepicardial enhancement, which were common in myocarditis and arrhythmogenic cardiomyopathy (Supplementary materials, Fig. S7). In particular, the five patients with free-breathing LGE images scored below 3 points showed subendocardial or transmural LGE. In these 5 patients, their LVEF averaged 18.7% (range: 16.0%–20.2%). We added additional breath-holding delayed enhancement scans or black blood sequences to verify the presence of subendocardial or transmural LGE in these 5 patients. The LGE image scores of other patients were ≥3 points, which could meet the clinical diagnosis.

Fig. 5.

A 52-year-old male patient with HCM underwent a full free-breathing CMR. A-D, cardiac cine imaging, the proximal and middle interventricular septum thickened, with a maximum thickness of 23 mm, and the SAM sign can be seen in the anterior mitral valve leaflet during the systolic period (Panel C, red arrow). E-H indicates the pre-T1 map, T2 map, post-T1 map, and ECV map, respectively. I, flow imaging, high-velocity blood flow can be seen during systole of the left ventricular outflow tract (red arrow). J-L, LGE imaging, focal enhancement (red arrows) can be seen in the right ventricular insertion of the interventricular septum. HCM hypertrophic cardiomyopathy, CMR cardiac magnetic resonance, ECV extracellular volume fraction, LGE late gadolinium enhancement, SAM systolic anterior motion

Fig. 6.

A 60-year-old male patient with an old myocardial infarction underwent a full free-breathing CMR. A-D, cardiac cine imaging, the middle and distal anteroseptal and apical ventricular wall were significantly thinner, and the left ventricle was enlarged (about 66 mm) with reduced ejection fraction (30%). E-G is the pre-T1 map, T2 map, and ECV map, respectively. H, flow imaging, a small regurgitation signal can be seen in the aortic valve during diastole (red arrow). I-L, LGE imaging showed transmural enhancement in the ventricular septum and the apex of the left ventricle, and low signal area was seen in the ventricular septum enhanced signal, MVO was considered (red arrows), and subendocardial myocardial enhancement was seen in the lateral and inferior walls of the left ventricle. CMR cardiac magnetic resonance, ECV extracellular volume fraction, LGE late gadolinium enhancement, MVO microvascular obstruction

4. Discussion

In this prospective study, we demonstrated the feasibility and effectiveness of a full free-breathing CMR protocol in clinical practice. Additionally, the cine and flow measurements were validated by comparing free-breathing with breath-hold CMR techniques. The free-breathing CMR technique enables comprehensive assessment of cardiac structure, function, and myocardial tissue characteristics, offering significant clinical value. The main findings are as follows: (1) The full free-breathing CMR protocol has an average acquisition time of 22.6 ± 3.7 min, effectively reducing scan time and enhancing CMR utilization; (2) This protocol meets diagnostic requirements, with most image quality scores ≥3, and demonstrates superior image quality in patients who have difficulty breath-holding or experience arrhythmias during the examination.

CMR is a valuable tool for diagnosing and assessing cardiovascular diseases, yet it is underutilized compared to other imaging methods [1], [3]. This is partly due to lengthy CMR exam times and long patient wait times affecting its use. Accelerated CMR techniques could address these issues [8], [9], [19], [20], [21]. Additionally, conventional CMR exams require breath-holding, which can be difficult for elderly patients or those with heart issues, creating a need for free-breathing CMR methods. Previous research has explored technologies like compressed sensing and deep learning for free-breathing CMR but primarily focused on single sequences such as cardiac cine or LGE [8], [14]. This study was not limited to the comparison of individual sequences; more importantly, it introduced a novel, fully free-breathing protocol for acquiring all CMR images, enabling comprehensive assessment of cardiac structure and function—an approach not previously reported.

This study found that the mean scan time for free-breathing CMR was 22.6 min, offering a comprehensive assessment of cardiac structure, function, and tissue characteristics. Traditionally, breath-holding CMR takes about one hour in our center. Thus, free-breathing CMR significantly reduces examination time and enhances CMR use. Additionally, free-breathing CMR can enhance image quality in patients who struggle to hold their breath. Results indicated that patients unable to hold their breath or with arrhythmias had higher image quality scores with free-breathing CMR than with conventional methods.

This study demonstrated that the EDV, ESV, EDVI, and ESVI measured by full free-breathing CMR were slightly higher than those assessed by conventional breath-holding CMR, while the LVEF and LVM were slightly lower, which is similar to some previous studies [9], [20], [22], [23]. These differences may be attributed to the slightly lower spatial and temporal resolution of free-breathing CMR and the marginally inferior image quality of free-breathing cine compared to conventional breath-holding cine, which can affect the accuracy of endocardial and epicardial delineation. Additionally, different breathing patterns may also contribute to these discrepancies [24], [25]. Nevertheless, the differences in left ventricular function parameters between the two techniques were small and not clinically significant.

The results of this study showed that it was difficult to identify some subendocardial myocardial enhancement in free-breathing LGE imaging. The possible reasons were considered as poor cardiac function in these patients and delayed clearance of contrast agents, resulting in poor contrast between the blood pool and the enhanced myocardium, which affected the observation of LGE. In addition, technical factors, such as the high signal-to-noise ratio caused by multiple averages and increased blood pool signal, can also affect the observation of subendocardial myocardial enhancement [12], [26]. For these patients, we may address this issue by extending the scan time, using black blood sequences, or by taking additional breath-holding delayed enhanced images.

Importantly, the primary innovation of this study lies in the introduction of a novel workflow for full free-breathing cardiac MRI, rather than in the direct comparison of individual imaging sequences. Exception for the comparison of cine and blood flow sequences, other sequences such as T1 mapping, T2 mapping, and LGE were not directly compared in this study. Nevertheless, the observed distribution characteristics of T1 values and LGE patterns across various cardiovascular diseases are consistent with findings from previous studies and our earlier reports [27], [28], [29], [30]. Future research will focus on further evaluating and comparing these individual sequences.

5. Limitations

First, not all patients underwent conventional breath-holding CMR; nevertheless, the sample size of this study was large enough to have little impact on the results. To the best of our knowledge, this is the largest sample size study ever conducted. Second, comparisons for sequences like T1/T2 mapping and LGE were not included, further study is needed. Third, there are only ten healthy individuals in this study, and the normal T1, T2, and ECV values given in this study may have a certain selection bias. However, according to our clinical experience, the T1, T2, and ECV values obtained from these participants are consistent with the results of the breath-holding scan. Fourth, respiratory through-plane motion, not corrected by MOCO, may cause slice misalignment and affect 2D measurements. While minimal in large cohorts, it can impact individual accuracy. Future studies should consider advanced motion-compensated reconstruction algorithms to address this.

In conclusion, the full free-breathing CMR examination is both feasible and effective in clinical practice. It significantly reduces scan time while maintaining high image quality, offering a comprehensive, one-stop evaluation of the myocardial structure, function, and myocardial tissue characteristics. This approach holds promising clinical applications with broad potential for widespread use.

Funding

This study was supported by National High Level Hospital Clinical Research Funding (Nos. 2022-GSP-QZ-5 and 2025-GSP-GG-5), CAMS Innovation Fund for Medical Sciences (CIFMS)(2025-I2M-C&T-A-008), High-end Medical Equipment Promotion and Application Project (No. 2024TGYY04), National Key R&D Program of China (Nos. 2021YFF0501400, 2021YFF0501401 and 2021YFF0501404), Fundamental Research Funds for the Central Universities (No. 3332024040).

Author contributions

M.J.L. was a major contributor in the conception and design of the work; K.Y. and C.C. were major contributors in writing the manuscript and the interpretation of data; S.H.Z. substantially revised the manuscript; F.T., G.Y., X.L.Y., and J.H.L. were major contributors in the acquisition and analysis of data. J.A., X.M.B., J.N.P., and K.C. were major contributors in the technical support. All authors read and approved the final manuscript.

Declaration of competing interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material

.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.