Validation of a novel three-dimensional ultrafast cardiac magnetic resonance imaging protocol in adolescents: a non-inferiority study compared with the two-dimensional gold standard

Abstract

Background

Cardiac magnetic resonance (CMR) is the gold standard for assessing cardiac anatomy and function. However, long acquisition times and multiple breath-holds pose significant challenges in pediatric imaging.

Objective

To perform a non-inferiority analysis comparing a novel 3-dimensional (D) ultrafast CMR protocol against the conventional 2-dimensional (D) gold standard for evaluating cardiac function, strain, and tissue characterization in adolescents.

Materials and methods

Thirty-nine adolescents (mean age 12.2±2.6 years) underwent both a standard 2-D protocol and a 3-D ultrafast protocol at 3-T. The 3-D protocol comprised enhanced sensitivity encoding by static outer volume subtraction (ESSOS) cine and 3-D late gadolinium enhancement. Image quality and diagnostic confidence were compared. A pre-specified non-inferiority margin (Δ) was used to assess functional and strain parameters. Agreement was evaluated using Bland-Altman analysis and intraclass correlation coefficients (ICCs).

Results

The total scan time was significantly shorter for the 3-D protocol compared to the 2-D protocol (75.8±10.0 s vs. 734.4±19.6 s, P<0.05). Image quality scores were comparable between protocols (median score 4.0, P=0.74). The 3-D protocol demonstrated statistical non-inferiority for all functional and strain metrics. Bland–Altman analysis showed minimal bias for key parameters, and the 95% confidence intervals for differences met the pre-specified non-inferiority margins. Intraclass correlation coefficients (ICCs) indicated good to excellent agreement for all parameters (n=39).

Conclusion

The novel 3-D ultrafast CMR protocol is non-inferior to the conventional 2-D gold standard for quantitative assessment of cardiac function and strain in adolescents. It offers comparable image quality with significantly reduced acquisition times, potentially improving clinical feasibility in pediatric populations.

Graphical abstract

Introduction

Cardiovascular diseases in pediatric populations encompass a range of conditions, including congenital heart disease, cardiomyopathy, myocarditis, Kawasaki disease, and arrhythmia. Recent advances in diagnostic and therapeutic techniques have led to a substantial increase in the survival rates of children diagnosed with cardiovascular disease [1]. Accurate characterization of cardiac anatomy, hemodynamics, function, and myocardial tissue before and after treatment is crucial for clinical decision-making and long-term follow-up [2]. Echocardiography is the primary imaging tool for evaluating cardiovascular disease in children. However, it exhibits certain limitations, especially in the detailed assessment of extracardiac vascular imaging, retrosternal structures, and myocardial tissue characteristics. Moreover, its effectiveness is influenced by operator dependence and consistency [3–5].

Cardiac magnetic resonance (CMR) is recognized as the gold standard for quantifying ventricular morphology and functional features. Accurate assessment of morphological and functional alterations is crucial for treatment planning, determining efficacy, and monitoring follow-up [5, 6]. Therefore, CMR has emerged as an essential diagnostic modality for evaluating cardiovascular disease in pediatric patients. Nevertheless, one limitation of CMR examination is its prolonged scanning time and the fact that specific sequences are acquired using the breath-holding method, which is not well-suited for children across various age groups [1, 5, 7]. Therefore, CMR examinations characterized by rapid acquisition, free breathing, reduced breath-holding times, and shortened scanning times are urgently needed, particularly for the diagnosis of cardiovascular disease in children [6].

Recently, isotropic 3-dimensional (D) cine using enhanced sensitivity encoding (SENSE) by static outer volume subtraction (ESSOS) and 3-D late gadolinium enhancement sequences have been introduced, enabling the accurate assessment of the heart anatomy, function, and myocardial characteristics in a shorter time compared with conventional 2-dimensional (D) sequences [7, 8]. The technical principle of ESSOS reconstruction is grounded in the acquisition of two interleaved datasets: a static dataset obtained with a relatively low SENSE factor to capture stable anatomical structures, and a dynamic dataset acquired with a higher SENSE factor for each cardiac phase. By automatically identifying and subtracting static regions in the image domain, ESSOS isolates the dynamic components for reconstruction before merging them back with the static background. This approach significantly reduces the effective SENSE factor during dynamic reconstruction, thereby optimizing the trade-off between spatial and temporal resolution [9]. This efficiency is particularly critical in pediatric and adolescent imaging, where reducing breath-hold duration directly translates to fewer motion artifacts and improved diagnostic reliability. However, whether this fast 3-D CMR protocol is suitable for pediatric and adolescent patients remains unclear. In addition, the strain analysis ability should be compared between 3-D and 2-D cine sequences. Therefore, our study aimed to validate a new 3-D ultrafast CMR protocol for determining cardiac anatomy, function, and late gadolinium enhancement.

Materials and methods

Subjects

The study protocol complied with the Declaration of Helsinki and was approved by the local Institutional Review Board (No. KS20220585). The parents or guardians of all participants provided written informed consent.

A total of 50 adolescents were screened, and 43 were finally enrolled in the program. Inclusion criteria included the following: (1) ability to undergo magnetic resonance examination alone or with a guardian; (2) ability to cooperate with breath-holding; and (3) requirement for cardiovascular magnetic resonance (CMR) for clinical reasons. The exclusion criteria were as follows: (1) claustrophobia; (2) inability to remain still; (3) inability to tolerate scanning noise; and (4) severe arrhythmia.

Cardiac magnetic resonance protocols

Cardiac magnetic resonance (CMR) examinations were conducted using a 3-T scanner (Ingenia CX, Philips Healthcare, Best, the Netherlands) equipped with a 16-element phased-array cardiac coil. Routine breath-hold 2-D cine and late gadolinium enhancement sequences were acquired. The 3-D CMR protocol included enhanced sensitivity encoding (ESSOS) and 3-D late gadolinium enhancement.

Routine standard 2-D cine scans were performed before injecting the gadolinium contrast agent. Balanced turbo field echo (TFE) steady-state free precession sequences with the following parameters were used: field of view (FOV), 270×270×96 mm; voxel size, 1.8×1.8 mm; repetition time (TR), 2.8 ms; echo time (TE), 1.41 ms; flip angle, 45°; SENSE 2.8, and a slice thickness of 5 mm, with no gap between the slices.

Enhanced sensitivity encoding by static Outer volume subtraction (ESSOS) was acquired after the intravenous administration of 0.10 mmol/kg gadoteric acid contrast agent (Gadovist; Bayer AG, Leverkusen, Germany). The parameters of the ESSOS sequence included a nonregulated 3-D coronal volume with an FOV of 300×321×340 mm (FH-AP-RL). The 3-D k-space was acquired via centric spiral ky-kz profile order with a voxel size of 2.60×2.66×2.74 mm (reconstructed to 1.34×1.34×1.70 mm) and 20 cardiac phases (triggered retrospectively). Nonselective radiofrequency excitation pulses were utilized to obtain a short TR (2.8 ms) with a relatively high flip angle (45°) while using full echo acquisition (TE=1.35 ms).

After 10 min of gadoteric acid contrast agent injection, a Look-Locker scan was performed to determine the appropriate inversion time for acquiring 3-D and 2-D late gadolinium enhancement images. 3-D late gadolinium enhancement imaging was conducted in the sagittal orientation using inversion recovery spoiled TFE acquisition (TR=2.2 ms; TE=1.08 ms; flip angle=7°). A 3-D volume of 350×350×141 mm (FH-AP-LR) was acquired, with a spatial resolution of 2.2×2.2×2.2 mm. The acquisition was triggered at end-diastole with a mean shot interval of 185 ms, and the entire acquisition was accelerated using a parallel acquisition factor of 4 (2×2 in the AP and LR directions), resulting in a breath-hold time of 13 s. For comparison, multiple 2-D slices were acquired using an equivalent acquisition technique. The in-plane resolution of the 2-D images was 1.55×0.55 mm, the slice thickness was 8 mm, and the images were acquired with an FOV of 380×380 mm and a parallel acquisition factor of 2, resulting in a 15-s breath-hold per slice.

Image analysis

All CMR data were transferred to an offline workstation with the commercial post-processing software CVI42 (Circle Cardiovascular Imaging, Calgary, Canada). All images were independently analyzed by two cardiac radiologists (W.C. and S.L., with 10 years and 15 years of experience in cardiovascular imaging, respectively) who were blinded to the clinical data. Subjective image quality was rated on a 5-point Likert scale, as follows: 5=excellent image quality, interpretable with no artifacts; 4=good image quality, interpretable with minimal artifacts; 3=average image quality, interpretation mildly compromised by image artifacts; 2=below-average image quality, interpretable but moderately compromised; and 1=poor image quality, uninterpretable images [6, 10]. Detailed qualitative scoring (scores of 0–3) was performed according to the standardized criteria of the European Cardiovascular Magnetic Resonance Registry, considering image artifacts or impaired image quality, with lower scores indicating better image quality [4, 11, 12]. To provide a robust assessment of observer variability, reliability was assessed in the full cohort of 39 patients, with a minimum of 1 week between evaluations.

In all cardiac phases, 3-D cine images were reformatted by radiologists in the short-axis view using a slice thickness of 5 mm with no gap, mimicking the coverage of 2-D cine images. Images were obtained from 2-D and 3-D cine for functional and strain analysis, whereas the presence of late gadolinium enhancement was obtained from 2-D and 3-D late gadolinium enhancement images. The left ventricular end-diastolic volume (LVEDV), right ventricular end-diastolic volume (RVEDV), left ventricular end-systolic volume (LVESV), right ventricular end-systolic volume (RVESV), left ventricular mass (LVM), left ventricular stroke volume (LVSV), and right ventricular stroke volume (RVSV) were measured.

To measure strain, the endocardial and epicardial borders were traced on the entire stack of 2-D and ESSOS cine short-axis and two- and four-chamber long-axis images using artificial intelligence (AI). Subsequently, these contours were tracked throughout the cardiac cycle. Manual correction was performed when the output provided by the AI was inadequate. The following myocardial strain parameters were derived: peak systolic global longitudinal strain (GLS), peak global radial strain (GRS), and peak global circumferential strain (GCS) [12]. The presence of late gadolinium enhancement was evaluated, and diagnostic confidence was assigned a score. Where present, late gadolinium enhancement was semiautomatically quantified using the full-width half-maximum method [13].

Statistical analysis

Statistical analyses were performed using SPSS software, version 26 (IBM, Armonk, NY). Normality was assessed using the Shapiro–Wilk test. Continuous variables are presented as mean±standard deviation (SD) for normally distributed data and as median with interquartile range (IQR) for non-normally distributed data. Categorical variables are expressed as frequencies and percentages. Continuous variables were compared between groups using paired t-tests or Wilcoxon signed-rank tests, as appropriate. The primary analysis was a non-inferiority test comparing the 3-D ESSOS protocol to the 2-D gold standard for cardiac function parameters, adhering to regulatory guidelines for new medical diagnostics [14, 15]. Pre-specified non-inferiority margins (Δ) were defined based on clinical rationale and established CMR thresholds: LVEDV 15 mL, LVESV 8 mL, LVSV 10 mL, LVCO 0.8 L/min, LVM 10 g, LVEF 5%, RVEDV 15 mL, RVESV 8 mL, RVSV 10 mL, RVCO 0.8 L/min, RVEF 5%, and 3% for all strain parameters (LVGLS, LVGCS, LVGRS) [16–18]. Non-inferiority was established if the relevant limit of the 95% confidence interval (lower or upper bound, as appropriate) met the pre-specified margins. Specifically, considering the potential for underestimation, non-inferiority was confirmed if the lower bound of the 95% CI was greater than the negative margin (-Δ) [14].

Agreement was evaluated using Bland–Altman analysis to determine mean bias and 95% limits of agreement (LoA). Inter- and intra-observer reliability were assessed using intraclass correlation coefficients (ICCs), with values >0.75 considered good and >0.90 excellent. A two-sided P-value <0.05 was considered statistically significant.

Results

Participant characteristics

The screening and enrollment process is illustrated in Fig. 1. Of 50 screened individuals, 43 were enrolled. Two participants withdrew, and two were excluded due to suboptimal image quality, resulting in a final cohort of 39 patients (22 males, 17 females; mean age 12.2±2.6 years). Clinical indications included chest symptoms (n=10), suspected/confirmed myocarditis (n=8), hypertrophic cardiomyopathy (n=5), hypertension (n=2), abnormal ECG (n=5), congenital heart disease (n=4), and history of cardiac surgery (n=5). Demographic data are summarized in Table 1.

Fig. 1.

Participant flow diagram illustrating the enrollment, exclusion, and final analysis cohort

Table 1.

Participant characteristics

Characteristic	Value
Total no. of participants	39
Male (n)	22
Female (n)	17
Mean age (y)	12.2±2.6
Mean height (cm)	158.4±15.1
Mean weight (kg)	48.9±14.3

Comparison of acquisition time

The 3-D ultrafast CMR protocol (ESSOS + 3-D late gadolinium enhancement) required a total scanning time of 75.8±10.0 s. The 3-D cine sequence required repetition in only two patients. These repetitions were necessary due to significant bulk body motion during the scan, which resulted in image blurring and artifacts that hindered accurate quantification. In contrast, the conventional 2-D sequences required 734.4±19.6 s. The 3-D acquisition time was significantly shorter than that of the 2-D protocol (P<0.05).

Image quality and late gadolinium enhancement diagnostic confidence

Representative cases are shown in Figs. 2 and 3. Subjective general image quality scores were identical between sequences (median 4.0; P=0.74). No significant differences were found for specific artifacts (respiratory/cardiac ghosting, blurring) (all P>0.05) (Table 2). Diagnostic confidence for late gadolinium enhancement was high and comparable between techniques (P>0.99).

Fig. 2.

Two- (a-c) and 3- (d-f) dimensional cardiac magnetic resonance enhanced sensitivity encoding by static outer volume subtraction (ESSOS) images in a 12-year-old boy diagnosed with abnormal ECG but no structural heart disease. There is no significant difference in image quality or cardiac measurements. For the 2-D protocol: (a) four-chamber cine, (b) short-axis two-chamber cine, and (c) axial late gadolinium enhancement. For the 3-D protocol: (d) four-chamber cine, (e) short-axis two-chamber cine, and (f) axial late gadolinium enhancement. No myocardial hyperintensity is observed in either sequence

Fig. 3.

Two- (a-c) and 3- (d-f) dimensional cardiac magnetic resonance enhanced sensitivity encoding by static outer volume subtraction (ESSOS) images in a 14-year-old boy diagnosed with confirmed myocarditis. There is comparable image quality and diagnostic performance. For the 2-D protocol: (a) four-chamber cine, (b) short-axis two-chamber cine, and (c) axial late gadolinium enhancement. For the 3-D protocol: (d) four-chamber cine, (e) short-axis two-chamber cine, and (f) axial late gadolinium enhancement. Both sequences (c and f) clearly show subtle high signal intensity in the right ventricular insertion point (arrows), consistent with myocardial fibrosis or inflammation. There is no significant difference in the quantification of these late gadolinium enhancement areas between the two protocols

Table 2.

Comparison of image quality scores, qualitative analysis and semiautomatic quantification of late gadolinium enhancement between 2-D and 3-D protocols

Measure variable	2-D	3-D	P
Image quality scores
General image quality	4.0 (4.0, 5.0)	4.0 (4.0, 5.0)	0.74
Detailed qualitative scoring
Left ventricular coverage	0 (0, 0)	0 (0, 0)	>0.99
Wraparound	0 (0, 0)	0 (0, 0)	>0.99
Respiratory ghost	0 (0, 2)	0 (0, 2)	0.94
Cardiac ghost	0 (0, 1)	1 (0, 2)	0.15
Image blurring or mistriggering	1 (0, 2)	1 (0, 2)	0.16
Metallic artifacts	0 (0, 0)	0 (0, 0)	0.99
Signal loss (coil inactive)	0 (0, 0)	0 (0, 0)	0.99
Section thickness	0 (0, 0)	0 (0, 0)	0.99
Gap	0 (0, 0)	0 (0, 0)	0.99
Diagnostic confidence for late gadolinium enhancement
Diagnostic confidence	2.0 (2.0, 2.0)	2.0 (2.0, 2.0)	>0.99

Cardiac function and strain parameters

The 3-D ESSOS sequence demonstrated non-inferiority to the 2-D sequence for all key parameters (Table 3). For LVEF, the mean difference was −1.2% (95% CI: −2.69% to 0.29%); the lower bound (−2.69%) was greater than the negative margin (−5%), and the upper bound was well below the 5% margin. Similarly, volumetric parameters met non-inferiority criteria: LVEDV (mean difference 1.5 mL; upper bound 3.87 mL <Δ15 mL), LVESV (mean difference 2.0 mL; upper bound 4.17 mL <Δ8 mL), and LVM (mean difference −1.8 g; upper bound 1.38 g <Δ10 g). Notably, for LVCO, the mean difference was −0.15 L/min with a 95% CI of −0.45 L/min to 0.15 L/min, falling entirely within the ±0.8 L/min margin.

Table 3.

Non-inferiority comparative analysis between the 2-D group and the 3-D group

Variable	Mean diff (3-D–2-D)	Δ	Lower limit of 95% CI	Upper limit of 95% CI	Non-inferiority
LVEDV (mL)	1.5	15	−0.87	3.87	✓
LVESV (mL)	2.0	8	−0.17	4.17	✓
LVSV (mL)	2.3	10	−0.23	4.83	✓
LVCO (L/min)	−0.15	0.8	−0.45	0.15	✓
LVM (g)	−1.8	10	−4.98	1.38	✓
LVEF (%)	−1.2	5	−2.69	0.29	✓
RVEDV (mL)	−0.5	15	−1.33	0.33	✓
RVESV (mL)	1.4	8	−1.15	3.95	✓
RVSV (mL)	−0.6	10	−2.75	1.55	✓
RVCO (L/min)	−0.05	0.8	−0.25	0.15	✓
RVEF (%)	−0.54	5	−1.18	0.10	✓
LVGLS (%)	0.7	3	−2.89	4.29	✓
LVGCS (%)	0.0	3	−2.63	2.63	✓
LVGRS (%)	0.3	3	−2.36	2.96	✓

Right ventricular parameters also demonstrated non-inferiority: RVEDV (mean difference −0.5 mL) and RVSV (mean difference −0.6 mL) had confidence intervals within margins. RVCO showed a minimal mean difference of −0.05 L/min. For myocardial strain, LVGLS showed a mean difference of 0.7%. The lower bound of the 95% CI (−2.89%) was greater than the pre-specified negative margin (−3%), thereby meeting the non-inferiority criterion. LVGCS (mean difference 0.0%) and LVGRS (mean difference 0.3%) also demonstrated good consistency.

Bland–Altman plots (Figs. 4, 5, and 6) demonstrated minimal bias and excellent agreement for all left ventricular, right ventricular, and myocardial strain parameters. Intraclass correlation coefficient (ICC) analysis for the full cohort (n=39) corroborated these findings, with absolute agreement and consistency ICC values exceeding 0.80 for most parameters (Table 4).

Fig. 4.

Bland–Altman plots comparing left ventricular parameters between 2-D and 3-D enhanced sensitivity encoding (SENSE) by static outer volume subtraction (ESSOS) protocols. The parameters include (a) left ventricular end-diastolic volume, (b) left ventricular end-systolic volume, (c) left ventricular stroke volume, (d) left ventricular cardiac output, (e) left ventricular mass, and (f) left ventricular ejection fraction

Fig. 5.

Bland–Altman plots comparing right ventricular parameters between 2-D and 3-D enhanced sensitivity encoding (SENSE) by static outer volume subtraction (ESSOS) protocols. The parameters include (a) right ventricular end-diastolic volume, (b) right ventricular end-systolic volume, (c) right ventricular stroke volume, (d) right ventricular cardiac output, and (e) right ventricular ejection fraction

Fig. 6.

Bland–Altman plots comparing left ventricular strain parameters between 2-D and 3-D enhanced sensitivity encoding (SENSE) by static outer volume subtraction (ESSOS) protocols. The parameters include (a) global longitudinal strain, (b) global circumferential strain, and (c) global radial strain

Table 4.

Absolute-agreement and consistency-of-agreement intraclass correlation coefficients for comparisons between 2-D and 3-D Extended Segmented Static Output Sampling protocols

Measured variable	ICC
	Absolute agreement (95% CI)	Consistency agreement (95% CI)
Left ventricle
LVEDV (mL)	0.96 (0.92, 0.98)	0.96 (0.93, 0.98)
LVESV (mL)	0.88 (0.84, 0.93)	0.92 (0.87, 0.95)
LVSV (mL)	0.95 (0.91, 0.98)	0.97 (0.95, 0.99)
LVCO (L/min)	0.92 (0.85, 0.96)	0.93 (0.88, 0.96)
LVM (g)	0.85 (0.81, 0.92)	0.92 (0.87, 0.95)
LVEF (%)	0.95 (0.91, 0.97)	0.95 (0.92, 0.97)
Right ventricle
RVEDV (mL)	0.94 (0.89, 0.97)	0.94 (0.90, 0.97)
RVESV (mL)	0.93 (0.88, 0.96)	0.95 (0.92, 0.97)
RVSV (mL)	0.98 (0.96, 0.99)	0.98 (0.97, 0.99)
RVCO (L/min)	0.88 (0.80, 0.93)	0.89 (0.82, 0.94)
RVEF (%)	0.93 (0.88, 0.96)	0.93 (0.89, 0.97)
Myocardial strain
GLS (%)	0.90 (0.84, 0.95)	0.90 (0.85, 0.95)
GCS (%)	0.85 (0.79, 0.91)	0.89 (0.83, 0.94)
GRS (%)	0.83 (0.75, 0.89)	0.85 (0.80, 0.91)

Discussion

This study highlights the significant clinical advantages of the 3-D ultrafast CMR protocol over conventional 2-D cine sequences in adolescent patients. First, the 3-D protocol achieves diagnostic non-inferiority across all core cardiac functional and strain parameters, matching the reliability of the 2-D gold standard that underpins clinical decision-making. Second, it preserves equivalent image quality and diagnostic confidence, ensuring no compromise in identifying clinically relevant findings such as myocardial fibrosis. Third, it delivers a dramatic reduction in scan time and simplifies the imaging workflow. These features collectively position the 3-D ultrafast CMR protocol as a robust, practical alternative for routine clinical practice, addressing the challenges of patient cooperation and expanding access to high-quality cardiac imaging in pediatric populations.

To our knowledge, this is the first study to rigorously validate the non-inferiority of the ESSOS-based 3-D ultrafast CMR protocol against the 2-D gold standard in an adolescent cohort. Unlike previous studies that reported only a lack of statistical difference, we employed a rigorous framework with predefined margins (Δ). Our results demonstrate that 3-D ESSOS meets clinically acceptable equivalence standards while achieving an approximate 90% reduction in scan time, marking a paradigm shift toward a “fast, accurate, and patient-friendly” model in pediatric CMR [19].

The core strength of this study lies in establishing statistical non-inferiority. For LVEF, the upper bound of the 95% CI for the mean difference (0.29%) was far below the 5% margin—a result superior to 3-D-2-D differences reported in adult literature. Volumetric parameters (LVEDV, LVESV) also showed excellent performance, with variability significantly lower than that reported by Maredia et al. [20] using k-t SENSE technology. Notably, our LVM measurement bias was only −1.8 g, demonstrating higher precision than the 3 g deviation reported in deep learning-accelerated 3-D CMR studies [21, 22]. This is likely attributable to the optimized signal-to-noise ratio provided by ESSOS static outer-volume subtraction. Furthermore, hemodynamic assessment proved highly reliable; LVCO and RVCO demonstrated minimal mean biases of −0.15 L/min and −0.05 L/min, respectively, with confidence intervals falling entirely within the strict ±0.8 L/min margin. Right ventricular assessment—historically challenging in pediatric CMR—also achieved non-inferiority, confirming the feasibility of rapid imaging for accurate biventricular quantification.

Regarding myocardial strain, we are the first to demonstrate in adolescents that 3-D strain analysis maintains high consistency with the gold standard (LVGLS bias of 0.7%), offering a reliable tool for detecting subclinical dysfunction. Although the confidence interval for LVGLS was slightly wider, the lower bound (−2.89%) remained within the clinically acceptable non-inferiority range (>−3%), indicating that the 3-D protocol does not clinically underestimate strain values. Comprehensive agreement analysis further supported these findings. Bland–Altman plots demonstrated narrow 95% LoA for all parameters, supporting interchangeability between the two techniques. For LVEF, the mean bias was −1.2% (95% LoA: −10.2% to 7.8%), notably narrower than values reported in previous pediatric CMR imaging validation studies [23–25]. The minimal bias and uniform distribution of differences across the measurement range confirmed the absence of systematic error proportional to disease severity, a critical consideration for monitoring progressive conditions such as cardiomyopathy. ICC analysis corroborated these results, with absolute agreement values >0.80 for all key parameters, demonstrating excellent reliability and exceeding ICCs reported for free-breathing 3-D sequences [26].

The 3-D ESSOS sequence represents a transformative advancement in pediatric CMR imaging, combining ultrafast acquisition, high-resolution coverage, and comprehensive functional assessment while overcoming key limitations of conventional 2-D multislice strategies. In pediatric practice, the 3-D ESSOS sequence reduces breath-hold requirements, streamlining the imaging process, enhancing patient comfort, and improving compliance—key advantages for pediatric populations, who often struggle with prolonged stillness and have limited tolerance for lengthy examinations [27, 28]. Additionally, the sequence’s nondirectional acquisition protocol eliminates the need for prescan cardiac axis alignment, a step that accounts for 28% of examination time in pediatric 2-D imaging. By removing this workflow barrier, the 3-D ESSOS sequence not only shortens the total scan time but also minimizes motion artifacts and errors caused by poor patient cooperation, directly addressing 2 key challenges in pediatric CMR imaging: anxiety-driven noncompliance and motion-related image degradation. These advances collectively improve imaging efficiency, enhance the patient experience, and preserve high image quality.

Notably, the 3-D ESSOS sequence preserves image quality comparable to that of the 2-D cine sequence, with no considerable differences in left ventricular coverage, respiratory ghosts, cardiac ghosts, or image blurring. This consistency is remarkable considering its single-breath-hold acquisition and shortened scan time, highlighting that increased efficiency does not compromise diagnostic clarity [28, 29]. Collectively, the 3-D ESSOS sequence with its ultrafast single-breath-hold acquisition, isotropic resolution, comprehensive anatomical coverage, and robust functional, strain, and late gadolinium enhancement assessment represents a uniquely valuable tool in pediatric CMR imaging. Its benefits are particularly pronounced in the diagnosis and postoperative follow-up of complex congenital heart diseases, where detailed anatomical information on the heart and great vessels is essential [30]. By overcoming the limitations of 2-D imaging while maintaining or improving diagnostic performance, the 3-D sequence addresses unmet needs in pediatric cardiac imaging, providing a rapid, reliable, and patient-centered solution tailored to the unique challenges of evaluating pediatric and adolescent populations.

This study had several limitations. The sample size was relatively small and consisted of adolescents from a single center, and the cohort included heterogeneous cardiac pathologies. Second, while this study utilized two experienced radiologists for image analysis, incorporating a broader reader pool, such as cardiologists, could further enhance the generalizability of these findings. Additionally, the number of patients with late gadolinium enhancement was small, preventing extensive quantitative analysis of fibrosis burden, though the comparable diagnostic confidence remains a notable preliminary finding. Future multi-center studies with larger, more diverse pediatric cohorts are needed to confirm these results and specifically investigate the potential benefits of ESSOS imaging in complex anatomical assessments.

Conclusion

This study validates the novel 3-D ultrafast CMR protocol as statistically non-inferior to the conventional 2-D gold standard for assessing cardiac function and strain in adolescent patients. The protocol offers comparable diagnostic quality while substantially reducing scanning time, thereby enhancing patient comfort and broadening the clinical applicability of CMR in pediatric cardiology.

Author contribution

L.X. and R.W. contributed to the study conception and design. Y.Z. and X.T. performed the MRI examinations and data acquisition. Material preparation and data analysis were performed by W.C., S.Liu, H.W., and S.Li. W.Li assisted with patient recruitment and clinical consultation. J.L. provided technical support for the ESSOS sequence optimization. The first draft of the manuscript was written by W.C. and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Data availability

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

Declarations

Competing interests

Jianxiu Lian is an employee of Philips Healthcare. The other authors have no competing interests to declare.