Clinical application of a full free-breathing cardiac MRI protocol: feasibility, efficiency, and diagnostic accuracy

Introduction

Cardiac magnetic resonance imaging (MRI) is an established non-invasive imaging modality for a wide range of clinical indications (1,2). It is considered a “one-stop” assessment of the heart, as it provides detailed information on cardiac anatomy and tissue characteristics, and serves as the gold standard for evaluating cardiac function (2). However, current standard cardiac MRI protocols rely on breath-holding for the acquisition of all key sequences, including cine imaging, T2-weighted imaging (T2WI), first-pass perfusion imaging, and late gadolinium enhancement (LGE) imaging (3). Conventional protocols require repeated breath-holds (BHs) for each sequence, which restricts their clinical application, particularly in patients unable to sustain prolonged BHs or follow breathing commands. Thus, a holistic full free-breathing (FB) cardiac MRI protocol needs to be established that matches BH standards in image quality (IQ), and spatial and temporal resolution, while accounting for patient-specific factors, facilitating its routine clinical adoption.

Recent technological advancements in real-time (RT) acquisition, compressed sensing (CS), motion correction (MoCo), and deep learning reconstruction have facilitated the development of FB cardiac MRI. FB cine imaging incorporating CS and/or MoCo has been shown to achieve IQ and quantitative assessment of ventricular function comparable to conventional BH cine imaging (4-7). FB T2WI has been implemented using radial k-space acquisitions combined with respiratory navigator gating, and has shown improved myocardial visibility, less motion sensitivity, and better IQ compared to BH T2WI, particularly in patients with limited breath-holding capacity (8). In addition, FB-LGE imaging enabled by MoCo reconstruction strategies has shown high IQ, although it tends to overestimate LGE mass in cases involving subepicardial or transmural enhancement (9).

To date, most previous studies on FB cardiac MRI have focused on validating the feasibility of individual sequences. The implementation of a fully integrated FB cardiac MRI protocol holds significant promise for streamlining clinical workflow and expanding patient eligibility. However, prolonged acquisition times can undermine its diagnostic effectiveness in real-world settings by exacerbating clinically relevant challenges, such as patient discomfort, inadvertent motion, respiratory arrhythmia, or transient arrhythmias. Suboptimal IQ resulting from such factors may in turn impact diagnostic confidence and interpretation reliability. It remains unclear whether an integrated, full FB cardiac MRI protocol that incorporates cine, T2WI, and LGE sequences is feasible and clinically applicable for both qualitative and quantitative cardiac assessment.

Thus, this study sought to develop a full FB cardiac MRI protocol for clinical practice and to assess its diagnostic feasibility. We hypothesized that FB cine, T2WI, and LGE acquisitions would achieve comparable IQ and allow accurate qualitative and quantitative analysis of cardiac function and LGE relative to standard BH acquisitions. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2638/rc).

Methods

Study population

This was a prospective study. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Beijing Anzhen Hospital (No. 2025005x), and all patients provided written informed consent. Patients referred to Beijing Anzhen Hospital for clinically indicated cardiac MRI were consecutively enrolled in the study. The exclusion criteria included refusal to participate and standard contraindications to MRI. Among the 97 participants enrolled from December 2024 to February 2025, three withdrew due to physical discomfort. Consequently, a final cohort of 94 patients (mean age 48±14.9 years, range 14–82; 74.47% male) was included in the analysis.

Imaging protocols

All cardiac MRI examinations were conducted on a 3.0-T scanner (uMR880, United Imaging Healthcare, Shanghai, China), equipped with respiratory navigator-based electrocardiographic (ECG) gating technology and a 32-channel cardiac surface phased-array coil. An artificial intelligence (AI)-based cardiac shim method was employed, which automatically establishes a homogeneous B0 field and confirms the F0 frequency across the heart, eliminating the need for a separate frequency scout or additional shim. FB cine imaging was performed using AI-assisted compressed sensing (ACS) to accelerate single-beat RT acquisition, with data collected exclusively during cardiac cycles within respiratory quiescence to suppress motion artifacts. Each slice was fully acquired within a single R-R interval. The acquired RT frames were subsequently interpolated to 25 cardiac phases.

FB black-blood T2WI employed ECG-triggered fast spin-echo acquisition with an ACS-accelerated single-shot acquisition strategy. Each slice was acquired over two consecutive cardiac cycles, with one cycle for data acquisition and one cycle for magnetization recovery. To improve IQ, multiple repetitions were acquired for each slice. Images with good quality were selected and corrected using non-rigid MoCo. The resulting motion-corrected images were then averaged to generate the final image for each slice (a schematic pulse sequence diagram of the FB T2WI sequence is presented in Figure S1). FB phase-sensitive inversion recovery (PSIR) imaging was conducted using dual gating with ECG and a cross-pair excitation diaphragmatic navigator, allowing data acquisition during periods of respiratory quiescence. A navigator acceptance window of ±3 mm was applied. The scanning protocol was identical for each person, as shown in Figure S2.

Following localization imaging, short-axis T2WI images were acquired, followed by cine images of the short-axis and three long-axis views (4-chamber, 2-chamber, and 3-chamber). First-pass perfusion imaging was performed under FB conditions using a gradient echo sequence. Approximately 10 minutes after the administration of contrast agent (Magnevist, Bayer Healthcare, Berlin, Germany; 0.2 mmol/kg at 4 mL/s, followed by 20 mL saline chasing bolus), LGE images of the short axis and long axes (2-chamber and 4-chamber views) were acquired using a PSIR sequence. BH and FB sequences were scanned sequentially. The scan times for all sequences in each patient were recorded. Detailed scanning parameters of each sequence are listed in Table 1.

Image analysis

Two experienced radiologists (with 6 and 10 years of experience in cardiac MRI image interpretation, respectively) independently and randomly evaluated the images from all sequences in a blinded fashion, without knowledge of clinical information and the acquisition strategy (i.e., FB or BH). The IQ scores of the two observers were then averaged to yield a single value for each category. Qualitative analysis was performed on the T2WI and LGE images to assess the presence of edema and myocardial injury. In cases of disagreement, a consensus was reached through discussion with a third radiologist (with 15 years of experience). Perfusion imaging was excluded from the analysis because the corresponding BH sequences were not acquired. All patients received a single contrast injection, administered during the FB acquisition only.

IQ analysis

The overall IQ of all sequences was visually evaluated using a five-point Likert scale (5 = excellent IQ, 4 = normal IQ, 3 = presence of artifacts but sufficient IQ, 2 = severe artifacts in the area of the left ventricle, and 1 = completely non-diagnostic images). Only datasets with a score of 2 or higher were included in the analysis (10). Blood to myocardial contrast (BMC) was measured to quantitatively evaluate the IQ of the short-axis cine sequence. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated to evaluate the IQ of the T2WI and LGE sequences. For the T2WI images, the SNR was defined as the mean signal intensity of normal myocardium divided by background noise, while the CNR was calculated between normal myocardium and the left ventricular (LV) blood pool. For the LGE images, the SNR was defined as the mean signal intensity of the enhanced region divided by background noise. Additionally, the CNR was quantified for the following tissue contrasts: (I) LGE vs. normal myocardium; (II) LGE vs. LV blood pool; and (III) normal myocardium vs. LV blood pool. Mean signal intensities were measured in regions of interest (ROIs) within the LGE, normal myocardium, and LV cavity, each with a diameter of at least 2 mm. ROI placement followed specific guidelines: normal myocardium and LV blood pool were assessed at standardized sites (the septum and central LV cavity) on the mid-ventricular short-axis slice, while the LGE ROIs were placed on the slice with maximal enhancement, specifically in the area of highest signal intensity. Background noise was determined as the standard deviation (SD) of signal intensity measured in the air. The IQ results from two independent observers were averaged to yield a single value for each category.

Quantitative analysis of cardiac function and LGE

Biventricular function and LGE analyses were performed using Cvi42 software (version 5.2.0, Circle, Calgary, Canada). Prior to all the analyses, a random cohort of 15 patients was chosen to assess intra- and inter-observer agreement. For the intra-observer agreement assessment, one observer repeated all measurements after a 30-day interval. The endocardial trabeculations and papillary muscles were included in the ventricular cavity volumes. The biventricular function parameters included end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), ejection fraction (EF), and LV mass. LGE was defined as regions with a signal intensity exceeding 5 SDs above the mean signal of a remote myocardial ROI (11). The results were expressed as the LGE mass and LGE% [calculated as (LGE mass / LV mass) × 100%].

Statistical analysis

All statistical analyses were performed using SPSS software (version 25.0, IBM Corp., Chicago, IL, USA) and MedCalc statistical software (version 22.009, Ostend, Belgium). Sample size calculations were performed using PASS 2021 software. Based on a maximum allowable difference of 4, and assuming a mean difference of 1 with a SD of 1, the sample size calculation yielded a minimum requirement of 15 patients. Normally distributed continuous variables were expressed as the mean ± SD, and compared using the paired t-test. In contrast, non-normally distributed continuous variables were expressed as the median with the interquartile range (IQR), and compared using the Wilcoxon signed-rank test. Categorical variables were expressed as counts and percentages, and compared using the chi-square test or Fisher’s exact test, as appropriate. Bland-Altman analysis was used to evaluate the agreement of quantitative parameters between the BH and FB techniques. Inter- and intra-observer agreement was assessed with intra-class correlation coefficients (ICCs) for absolute agreement, using a two-way random-effects model on single measures (an ICC <0.40 indicated poor agreement; an ICC of 0.40–0.54, weak agreement; an ICC of 0.55–0.69, moderate agreement; an ICC of 0.70–0.84, good agreement; and an ICC of 0.85–1.00, excellent agreement) (12). A two-sided P value <0.05 was considered statistically significant.

Results

Study population

The clinical characteristics of the study cohort are summarized in Table 2. Approximately one-third of the study population was diagnosed with heart failure, including comorbidity of ischemic cardiomyopathy (n=4), inflammatory cardiomyopathy (n=2), and dilated cardiomyopathy (n=2). Among the 23 patients with coronary artery disease, 16 had chronic myocardial infarction (MI), and five had acute MI.

Imaging quality

The total scan time of the FB protocol was significantly shorter than that of the BH protocol (551±194 vs. 900±148 sec, P<0.001). The comparison of individual sequences also revealed significant time savings for the FB acquisitions [44±17 vs. 266±49 sec, P<0.001 for cine images; 127 sec (IQR 111, 148) vs. 298 sec (IQR 248, 322) for T2WI images; P<0.001]. Conversely, the scan time for the LGE sequences was comparable between the FB and BH techniques [327 sec (IQR 264, 427) vs. 352 sec (IQR 298, 383), P=0.385]. All images in the study cohort had overall IQ scores of 2 or higher. The overall IQ ratings showed no significant differences between the BH and FB images (4.19±0.87 vs. 4.10±0.57, P=0.121 for cine; 3.98±0.84 vs. 4.05±0.90, P=0.468 for T2WI; 4.29±0.73 vs. 4.27±0.69, P=0.385 for LGE). The BMC of the cine images did not differ significantly between the BH and FB images (4.66±1.29 vs. 4.57±0.90, P=0.528). The SNR and CNR values of the T2WI and LGE imaging were also comparable between the BH and FB protocols. Off-resonance artifacts were observed in nine balanced steady-state free precession (bSSFP) cine sequences acquired with BH and in eight sequences acquired with FB. The detailed IQ results are set out in Table 3.

Typical images are provided in Figures 1-3. Figure 1 includes images of a 64-year-old woman with chronic MI without arrhythmia. T2WI, cine, and LGE images acquired under both BH and FB conditions demonstrated comparably high IQ. Figure 2 includes images of a 54-year-old man with coronary artery disease, who experienced arrhythmia during the cardiac MRI examination. Due to poor BH cooperation, the BH-acquired T2WI, cine, and LGE images exhibited noticeable blurring and artifacts around the left ventricle. Conversely, the FB-acquired images for all three sequences achieved high IQ. Figure 3 includes images of a 53-year-old man with acute MI. The BH- and FB-acquired T2WI, cine, and LGE images were of similarly high IQ. The T2WI images, regardless of the acquisition method, showed hyperintensity in the anterior and septal walls, indicating myocardial edema, along with apical-septal hypointensity suggestive of intra-myocardial hemorrhage.

Figure 1 Images of a 64-year-old woman with chronic myocardial infarction without arrhythmia. BH-acquired T2WI (A-C), cine (G-I), and LGE (M-O) images, and FB-acquired T2WI (D-F), cine (J-L), and LGE (P-R) images all showed equally high image quality. Edema was not observed in either the BH or FB T2WI images. LGE images demonstrated an identical extent of sub-endocardial and transmural myocardial infarction. Quantitative analysis: BH: LVEDV = 223.72 mL, LVESV = 159.14 mL, LVSV = 64.58 mL, LVEF = 28.87%, LV mass = 128.37 g, RVEDV = 118.8 mL, RVESV = 37.2 mL, RVSV = 81.6 mL, RVEF = 68.69%, LGE% = 37.93%. FB: LVEDV = 224.57 mL, LVESV = 160.02 mL, LVSV = 64.55 mL, LVEF = 28.74%, LV mass = 126.86 g, RVEDV = 117.61 mL, RVESV = 36.37 mL, RVSV = 81.84 mL, RVEF = 69.58%, LGE% = 37.51%. BH, breath-hold; EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume; FB, free-breathing; LGE, late gadolinium enhancement; LV, left ventricular; RV, right ventricular; SV, stroke volume; T2WI, T2 weighted imaging.

Figure 2 Images of a 54-year-old man with coronary artery disease and arrhythmia during cardiac magnetic resonance examination. The BH-acquired T2WI (A-C), cine (G-I), and LGE (M-O) images were more blurred and displayed artifacts around the left ventricle due to poor BH cooperation, while the FB-acquired T2WI (D-F), cine (J-L), and LGE (P-R) images all achieved high image quality. Edema and LGE were absent in both the BH- and FB-acquired images. Quantitative analysis: BH: LVEDV = 101.68 mL, LVESV = 59.81 mL, LVSV = 41.87 mL, LVEF = 41.18%, LV mass = 111.31 g, RVEDV = 146.45 mL, RVESV = 100.74 mL, RVSV = 45.71 mL, RVEF = 31.21%. FB: LVEDV = 97.93 mL, LVESV = 55.15 mL, LVSV = 42.78 mL, LVEF = 43.68%, LV mass = 113.64 g, RVEDV = 146.2 mL, RVESV = 95.16 mL, RVSV = 51.04 mL, RVEF = 34.91%. BH, breath-hold; EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume; FB, free-breathing; LGE, late gadolinium enhancement; LV, left ventricular; RV, right ventricular; SV, stroke volume; T2WI, T2 weighted imaging.

Figure 3 Images of a 53-year-old man with acute myocardial infarction. The BH-acquired T2WI (A-C), cine (G-I), and LGE (M-O) images, and FB-acquired T2WI (D-F), cine (J-L), and LGE (P-R) images all showed equally high image quality. The T2WI images, regardless of the acquisition method, revealed hyperintensity in the anterior and septal walls, indicating myocardial edema, along with apical-septal hypointensity suggestive of intra-myocardial hemorrhage. The LGE images further demonstrated an identical extent of sub-endocardial and transmural myocardial infarction. Quantitative analysis: BH: LVEDV = 149.11 mL, LVESV = 96.89 mL, LVSV = 52.22 mL, LVEF = 35.02%, LV mass = 151.68 g, RVEDV = 113.17 mL, RVESV = 52.31 mL, RVSV = 60.86 mL, RVEF = 53.78%. FB: LVEDV = 147.27 mL, LVESV = 96.21 mL, LVSV = 51.06 mL, LVEF = 34.67%, LV mass = 150.75 g, RVEDV = 114.58 mL, RVESV = 53.59 mL, RVSV = 60.99 mL, RVEF = 53.28%. BH, breath-hold; EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume; FB, free-breathing; LGE, late gadolinium enhancement; LV, left ventricular; RV, right ventricular; SV, stroke volume; T2WI, T2 weighted imaging.

In a cohort of 33 patients with heart failure and/or arrhythmia, the overall IQ scores for FB imaging were slightly higher than those for BH imaging across the cine (4.18±0.64 vs. 3.76±1.00, P=0.073), T2WI (4.00±0.90 vs. 3.91±0.84, P=0.660), and LGE (4.12±0.78 vs. 4.09±0.84, P=0.962) images, although these differences did not reach statistical significance (see Table S1).

Qualitative and quantitative analysis of cardiac function and LGE

The results of the qualitative (myocardial edema and LGE presence) and quantitative (ventricular function and LGE quantification) analyses are set out in Table 4. For all the qualitative and quantitative results, there was no significant difference between the BH and FB images. Bland-Altman analysis (Figures 4,5) further demonstrated high agreement between the two protocols, with mean differences for all parameters approaching zero and the majority of data points falling within the 95% limits of agreement. These findings were consistent in a subgroup of 33 patients with heart failure and/or arrhythmia (Table S2). Additionally, all parameters showed excellent intra- and inter-observer agreement, ranging from 0.863 to 0.9998 (Table 5).

Figure 4 Bland-Altman plots applied to evaluate LVEDV (A), LVESV (B), LVSV (C), LVEF (D), LV mass (E), and LGE% (F) for agreement between FB and BH acquisition methods. The solid central line indicates the mean bias, and the upper and lower dashed lines represent the 95% limits of agreement (±1.96 × SD of the differences). BH, breath-hold; EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume; FB, free-breathing; LGE, late gadolinium enhancement; LV, left ventricular; SD, standard deviation; SV, stroke volume.

Figure 5 Bland-Altman plots applied to evaluate RVEDV (A), RVESV (B), RVSV (C), and RVEF (D) for agreement between the FB and BH acquisition methods. The solid central line indicates the mean bias, and the upper and lower dashed lines represent the 95% limits of agreement (±1.96 × SD of the differences). BH, breath-hold; EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume; FB, free-breathing; RV, right ventricular; SD, standard deviation; SV, stroke volume.

Discussion

The FB technique is clinically important for patients who cannot cooperate or endure long breath-holding. In this study, we proposed a full FB cardiac MRI scanning protocol (including cine, T2WI, and LGE sequences) and assessed its clinical feasibility in 94 consecutive patients with different cardiac disorders. Compared with the standard BH scanning protocol, this protocol achieved significantly shorter scan times, similar IQ, identical qualitative identification of myocardial edema and enhancement, and comparable quantitative measurements of biventricular function and LGE size. In patients with heart failure and/or arrhythmia, the FB imaging achieved slightly higher overall IQ scores (4.18±0.64 vs. 3.76±1.00, P=0.073 for cine; 4.00±0.90 vs. 3.91±0.84, P=0.660 for T2WI; 4.12±0.78 vs. 4.09±0.84, P=0.962 for LGE), while maintaining accurate qualitative and quantitative results. Specifically, the qualitative identification of myocardial edema and enhancement on FB sequences was identical to that on BH sequences, and the mean differences for biventricular function parameters approached zero, with the majority of data points falling within the 95% limits of agreement. Further, all quantitative analysis results showed excellent intra- and inter-observer agreement, indicating that full FB cardiac MRI examination is feasible with high reproducibility, and could thus be widely applied in routine clinical practice.

The core technological innovations of our full FB cardiac MRI protocol lie in AI-ACS and MoCo. CS exploits spatiotemporal redundancy across multiple cardiac cycles to enhance spatial or temporal resolution (13), and has been widely adopted in cine imaging to accelerate acquisition. Previous studies have shown its feasibility under both BH (14,15) and FB conditions (10,16). More recently, MoCo techniques have been incorporated across multiple cardiac MRI sequences to mitigate artifacts and blurring caused by cardiac and respiratory motion (17). Our proposed full FB cardiac MRI protocol integrates CS and MoCo techniques, and employs AI to assist acceleration and IQ improvement during single-beat RT cine acquisition and single-shot T2WI acquisition, allowing for significant rapid scanning and accurate diagnosis.

Previous studies have explored FB implementations for individual sequences, highlighting benefits such as reduced scan times and generally comparable performance; however, their findings remain inconsistent. In relation to IQ, several studies reported slightly inferior subjective scores for FB cine acquisitions compared with conventional BH acquisitions (7,18-20). However, the majority of published works indicate that FB imaging with CS and MoCo yields IQ comparable to that of BH imaging (4,6,21,22). Some studies have even demonstrated superior subjective IQ in FB acquisitions across cine (5), T2WI (8), and LGE (9,23) sequences. Our study demonstrated comparable IQ between FB and BH acquisitions across all sequences, which aligns with the majority of previous reports. In contrast to some earlier studies (8,9,23) that reported superior IQ for FB-acquired T2WI and LGE images, we observed no significant difference overall, although a modest improvement with FB was observed in subgroups with heart failure and/or arrhythmia. This discrepancy in findings may be attributed to differences in the study populations. For instance, Liu et al. (8) included cohorts with high proportions of arrhythmia (35–40%) and patients with known signal loss on conventional BH T2WI. Similarly, Yu et al. (9) enrolled a substantial proportion of patients (23.5%) with arrhythmia or breath-holding difficulties, and Olivieri et al. (23) focused specifically on pediatric patients, who are more prone to arrhythmias and unable to perform BHs. Conversely, in our cohort, only nine of the 94 patients (9.57%) had arrhythmia, and the majority of heart failure patients (24 out of 28) demonstrated satisfactory BH capacity. Given that arrhythmias and an inability to follow BH commands are primary sources of artifacts in BH images, the relatively low prevalence of these factors in our cohort may explain the limited additional benefit of FB acquisitions in relation to IQ.

Similarly, most evaluations report consistent biventricular functional measurements between FB and BH methods (4-7,19-22,24), though minor discrepancies exist in certain parameters. For instance, Reiter et al. (25) reported higher LV volumes and lower EF during FB in participants without heart failure, which may be attributable to the absence of CS and MoCo in their FB acquisition. Sudarski et al. (10) found that RT true Fast Imaging with Steady-state Precession cine without breath-holding yielded accurate LV volumes and EF, but overestimated LV mass, whereas Kido et al. (16) reported a slight underestimation. A recent study by Yan et al. (18) showed comparable biventricular metrics between FB and BH cine imaging except for significantly smaller LV EDV in FB acquisitions. The inconsistencies among the aforementioned studies may be due to sampling or measurement bias. Research on FB T2WI and LGE imaging remains less extensive than that on cine imaging. The finding of Liu et al. (8) that FB T2WI significantly improved myocardial visibility and reduced motion sensitivity is inconsistent with our results, which showed that FB T2WI had a comparable but not superior ability to identify myocardial edema. This discrepancy is likely attributable to differences in the study cohort as noted above. Our findings align with previous studies (9,26-28) that reported strong agreement in LGE quantification between FB MoCo-LGE and conventional BH-LGE, confirming the considerable evolution of this technique. This consistency was observed even in our subgroup of patients with heart failure and/or arrhythmia.

The present study addressed the limitations of previous studies that primarily focused on individual sequences and showed that our proposed full FB cardiac MRI protocol is feasible in a diverse cohort of patients with various cardiac conditions. This confirms the practical viability of FB technology throughout complete cardiac MRI examinations under real-world conditions. By eliminating the need for repeated BHs, FB cardiac MRI significantly reduces the coordination burden on both patients and technologists, streamlines the scanning process, and enhances overall efficiency. It avoids interruptions between BH maneuvers and reduces the number of repeated acquisitions, which is a particular advantage for frail or vulnerable patients. Overall, the proposed integrated FB cardiac MRI protocol has the potential to standardize clinical workflow, improve consistency, and broaden accessibility, thereby reliably extending high-quality cardiac MRI capabilities to a wider referral population, including those with diverse cardiovascular diseases and limited ability to BH.

This study had several limitations. First, as a single-center study with a modest sample size, there is a potential risk of selection bias, despite the use of consecutive enrollment to minimize this effect. Second, FB-LGE and conventional BH-LGE images may differ visually to experienced readers; however, we employed objective quantitative metrics (the CNR and LGE mass) to mitigate interpretive subjectivity. Third, we acknowledge the limitations of using a background ROI-based method for SNR evaluation, especially in the context of parallel imaging. However, this approach is still widely reported in clinical application studies (29-32) evaluating parallel imaging, CS, and AI-accelerated MRI, particularly when the aim is to provide a pragmatic assessment of IQ under routine reconstruction conditions rather than a strict physical quantification of the SNR. The primary goal of the present study was to provide a practical and consistent comparative assessment of the SNR across different strategies of cardiac MRI (FB and BH acquisitions) using clinically acquired datasets. In this context, the SNR values reported in our study were intended as auxiliary, descriptive indicators of relative IQ, rather than absolute SNR measurements. Notably, the primary conclusions of the study are supported by qualitative image assessment and a clinically relevant evaluation of image interpretability, and do not rely solely on SNR comparisons. Further, the acquisition order of the BH- and FB-LGE images was not randomized, and the subsequently acquired FB-LGE may have benefited from greater contrast washout, potentially improving blood pool-myocardial differentiation. Nevertheless, we believe that this effect was minimal due to the established high reproducibility of LGE quantification. Finally, a quantitative comparison of myocardial edema between FB and BH T2WI was not feasible, as only five patients exhibited edema. Future studies should adopt a multi-center design with a larger cohort, randomize the acquisition order of the FB and BH-LGE sequences, and include quantitative assessments of myocardial edema to further validate our findings.

Conclusions

A full FB cardiac MRI protocol is feasible in clinical practice, providing IQ and diagnostic accuracy comparable to that of conventional BH imaging, while significantly reducing acquisition time. This FB cardiac MRI protocol can be implemented in routine clinical practice to streamline workflow and improve efficiency without compromising diagnostic accuracy, particularly for vulnerable patients who cannot perform BHs.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2638/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2638/dss

Funding: This work was supported by a grant from the National Natural Science Foundation of China (No. 82471937, to H.W.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2638/coif). K.X. is an employee of Shanghai United Imaging Healthcare Co., Ltd. J.D. is an employee of Beijing United Imaging Research Institute of Intelligent Imaging. H.W. received a grant from the National Natural Science Foundation of China (No. 82471937). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethical Committee of Beijing Anzhen Hospital (No. 2025005x), and all patients provided written informed consent.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Nguyen ET, Ordovas K, Herbst P, Kozor R, Ng MY, Natale L, Nijveldt R, Salgado R, Sanchez F, Shah D, Stojanovska J, Valente AM, Westwood M, Plein S. Competency based curriculum for cardiovascular magnetic resonance: A position statement of the Society for Cardiovascular Magnetic Resonance. J Cardiovasc Magn Reson 2024;26:100006. [Crossref] [PubMed]
2. Leiner T, Bogaert J, Friedrich MG, Mohiaddin R, Muthurangu V, Myerson S, Powell AJ, Raman SV, Pennell DJ. SCMR Position Paper (2020) on clinical indications for cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2020;22:76. [Crossref] [PubMed]
3. Kramer CM, Barkhausen J, Bucciarelli-Ducci C, Flamm SD, Kim RJ, Nagel E. Standardized cardiovascular magnetic resonance imaging (CMR) protocols: 2020 update. J Cardiovasc Magn Reson 2020;22:17. [Crossref] [PubMed]
4. Takakado M, Kido T, Ogawa R, Takimoto Y, Tokuda T, Tanabe Y, Kawaguchi N, Pang J, Komori Y, Kido T. Free-breathing cardiovascular cine magnetic resonance imaging using compressed-sensing and retrospective motion correction: accurate assessment of biventricular volume at 3T. Jpn J Radiol 2023;41:142-52. [Crossref] [PubMed]
5. Longère B, Abassebay N, Gkizas C, Hennicaux J, Simeone A, Rodriguez Musso A, Carpentier P, Coisne A, Pang J, Schmidt M, Toupin S, Montaigne D, Pontana F. A new compressed sensing cine cardiac MRI sequence with free-breathing real-time acquisition and fully automated motion-correction: A comprehensive evaluation. Diagn Interv Imaging 2023;104:538-46. [Crossref] [PubMed]
6. Lin L, Li Y, Wang J, Cao L, Liu Y, Pang J, An J, Jin Z, Wang Y. Free-breathing cardiac cine MRI with compressed sensing real-time imaging and retrospective motion correction: clinical feasibility and validation. Eur Radiol 2023;33:2289-300. [Crossref] [PubMed]
7. Wang Z, Wang Z, An J, Yuan Y, Pang J, He Y. Cardiac compressed sensing real-time cine for the assessment of left ventricular function: a large sample size analysis. Quant Imaging Med Surg 2024;14:8785-97. [Crossref] [PubMed]
8. Liu J, Jin H, Chen Y, Fu C, Chen C, Zeng M. Free-breathing BLADE acquisition method improves T2-weighted cardiac MR image quality compared with conventional breath-hold turbo spin-echo cartesian acquisition. Acta Radiol 2021;62:341-7. [Crossref] [PubMed]
9. Yu Y, Chen Y, Zhao S, Ge M, Yang S, Yun H, Bi X, Fu C, Zeng M, Jin H. Role of free-breathing motion-corrected late gadolinium enhancement technique for image quality assessment and LGE quantification. Eur J Radiol 2021;135:109510. [Crossref] [PubMed]
10. Sudarski S, Henzler T, Haubenreisser H, Dösch C, Zenge MO, Schmidt M, Nadar MS, Borggrefe M, Schoenberg SO, Papavassiliu T. 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. [Crossref] [PubMed]
11. Bondarenko O, Beek AM, Hofman MB, Kühl HP, Twisk JW, van Dockum WG, Visser CA, van Rossum AC. Standardizing the definition of hyperenhancement in the quantitative assessment of infarct size and myocardial viability using delayed contrast-enhanced CMR. J Cardiovasc Magn Reson 2005;7:481-5. [Crossref] [PubMed]
12. Schober P, Mascha EJ, Vetter TR. Statistics From A (Agreement) to Z (z Score): A Guide to Interpreting Common Measures of Association, Agreement, Diagnostic Accuracy, Effect Size, Heterogeneity, and Reliability in Medical Research. Anesth Analg 2021;133:1633-41. [Crossref] [PubMed]
13. Tsao J, Boesiger P, Pruessmann KP. k-t BLAST and k-t SENSE: dynamic MRI with high frame rate exploiting spatiotemporal correlations. Magn Reson Med 2003;50:1031-42. [Crossref] [PubMed]
14. Vincenti G, Monney P, Chaptinel J, et al. Compressed sensing single-breath-hold CMR for fast quantification of LV function, volumes, and mass. JACC Cardiovasc Imaging 2014;7:882-92. [Crossref] [PubMed]
15. Kido T, Kido T, Nakamura M, Watanabe K, Schmidt M, Forman C, Mochizuki T. 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. [Crossref] [PubMed]
16. Kido T, Kido T, Nakamura M, Watanabe K, Schmidt M, Forman C, Mochizuki T. Assessment of Left Ventricular Function and Mass on Free-Breathing Compressed Sensing Real-Time Cine Imaging. Circ J 2017;81:1463-8. [Crossref] [PubMed]
17. Zaitsev M, Maclaren J, Herbst M. Motion artifacts in MRI: A complex problem with many partial solutions. J Magn Reson Imaging 2015;42:887-901. [Crossref] [PubMed]
18. Yan X, Luo Y, Chen X, Chen EZ, Liu Q, Zou L, Bao Y, Huang L, Xia L. From Compressed-Sensing to Deep Learning MR: Comparative Biventricular Cardiac Function Analysis in a Patient Cohort. J Magn Reson Imaging 2024;59:1231-41. [Crossref] [PubMed]
19. Wang J, Li X, Lin L, Dai JW, Schmidt M, Forman C, An J, Jin ZY, Wang YN. Diagnostic efficacy of 2-shot compressed sensing cine sequence cardiovascular magnetic resonance imaging for left ventricular function. Cardiovasc Diagn Ther 2020;10:431-41. [Crossref] [PubMed]
20. Zucker EJ, Sandino CM, Kino A, Lai P, Vasanawala SS. Free-breathing Accelerated Cardiac MRI Using Deep Learning: Validation in Children and Young Adults. Radiology 2021;300:539-48. [Crossref] [PubMed]
21. Kocaoglu M, Pednekar AS, Wang H, Alsaied T, Taylor MD, Rattan MS. Breath-hold and free-breathing quantitative assessment of biventricular volume and function using compressed SENSE: a clinical validation in children and young adults. J Cardiovasc Magn Reson 2020;22:54. [Crossref] [PubMed]
22. Ghanbari F, Morales MA, Street JA, Rodriguez J, Johnson S, Pierce P, Carty A, Ngo LH, Hoeger CW, Tsao CW, Manning WJ, Nezafat R. Free-breathing, Highly Accelerated, Single-beat, Multisection Cardiac Cine MRI with Generative Artificial Intelligence. Radiol Cardiothorac Imaging 2025;7:e240272. [Crossref] [PubMed]
23. Olivieri L, Cross R, O’Brien KJ, Xue H, Kellman P, Hansen MS. Free-breathing motion-corrected late-gadolinium-enhancement imaging improves image quality in children. Pediatr Radiol 2016;46:983-90. [Crossref] [PubMed]
24. Zou Q, Xu HY, Fu C, Zhou XY, Xu R, Yang MX, Yang ZG, Guo YK. Utility of single-shot compressed sensing cardiac magnetic resonance cine imaging for assessment of biventricular function in free-breathing and arrhythmic pediatric patients. Int J Cardiol 2021;338:258-64. [Crossref] [PubMed]
25. Reiter C, Reiter U, Kräuter C, Nizhnikava V, Greiser A, Scherr D, Schmidt A, Fuchsjäger M, Reiter G. Differences in left ventricular and left atrial function assessed during breath-holding and breathing. Eur J Radiol 2021;141:109756. [Crossref] [PubMed]
26. Piehler KM, Wong TC, Puntil KS, Zareba KM, Lin K, Harris DM, Deible CR, Lacomis JM, Czeyda-Pommersheim F, Cook SC, Kellman P, Schelbert EB. Free-breathing, motion-corrected late gadolinium enhancement is robust and extends risk stratification to vulnerable patients. Circ Cardiovasc Imaging 2013;6:423-32. [Crossref] [PubMed]
27. Cha MJ, Cho I, Hong J, Kim SW, Shin SY, Paek MY, Bi X, Kim SM. Free-Breathing Motion-Corrected Single-Shot Phase-Sensitive Inversion Recovery Late-Gadolinium-Enhancement Imaging: A Prospective Study of Image Quality in Patients with Hypertrophic Cardiomyopathy. Korean J Radiol 2021;22:1044-53. [Crossref] [PubMed]
28. Li J, Kong H, Wang Z, Yuan Y, An J, He Y. Comparison of fast and standard segmented techniques for detection of late gadolinium enhancement in acute myocardial infarction: a prospective clinical cardiovascular magnetic resonance trial. Quant Imaging Med Surg 2025;15:5769-80. [Crossref] [PubMed]
29. Taron J, Weiss J, Notohamiprodjo M, Kuestner T, Bamberg F, Weiland E, Kuehn B, Martirosian P. Acceleration of Magnetic Resonance Cholangiopancreatography Using Compressed Sensing at 1.5 and 3 T: A Clinical Feasibility Study. Invest Radiol 2018;53:681-8. [Crossref] [PubMed]
30. Guo Y, Lin L, Zhao S, Sun G, Chen Y, Xue K, Yang Y, Chen S, Zhang Y, Li G, Zhu Y, Vliegenthart R, Wang Y. Myocardial Fibrosis Assessment at 3-T versus 5-T Myocardial Late Gadolinium Enhancement MRI: Early Results. Radiology 2024;313:e233424. [Crossref] [PubMed]
31. Liu J, Li W, Li Z, Yang J, Wang K, Cao X, Qin N, Xue K, Dai Y, Wu P, Qiu J. Magnetic resonance shoulder imaging using deep learning-based algorithm. Eur Radiol 2023;33:4864-74. [Crossref] [PubMed]
32. Bauer JS, Monetti R, Krug R, Matsuura M, Mueller D, Eckstein F, Rummeny EJ, Lochmueller EM, Raeth CW, Link TM. Advances of 3T MR imaging in visualizing trabecular bone structure of the calcaneus are partially SNR-independent: analysis using simulated noise in relation to micro-CT, 1.5T MRI, and biomechanical strength. J Magn Reson Imaging 2009;29:132-40. [Crossref] [PubMed]