Journal of Cardiovascular Magnetic Resonance最新文献

Background: Cardiac functional metrics such as ejection fraction, strain, and valve excursion are important diagnostic and prognostic measures of cardiac disease. However, they ignore a large amount of systolic shape change information available from modern cardiovascular magnetic resonance (CMR) examinations. We aimed to automatically quantify multidimensional shape and motion scores from CMR, investigate covariates, and test their discrimination of disease in the UK Biobank compared against standard functional metrics.

Methods: An automated analysis pipeline was used to obtain quality-controlled three-dimensional left and right ventricular shape models in 38,858 UK Biobank participants, 5149 of whom had one or more diagnoses of cardiovascular or cardiometabolic disease. Principal component analysis was used to obtain a statistical shape atlas and quantify each participant's left and right ventricular shape at both end-diastole and end-systole simultaneously. Systolic strain was obtained from arc length changes computed from the shape model, and mitral/tricuspid annular plane systolic excursion (MAPSE/TAPSE) was computed from the displacement of the valves. Discrimination for prevalent disease was quantified using linear discriminant analysis area under the receiver operating characteristic curve.

Results: The first 25 principal component scores captured >90% of the total shape variance. Significantly stronger discrimination for atrial fibrillation, heart failure, diabetes, ischemic disease, and conduction disorders (p<0.001 for each) was obtained using shape scores compared with volumes, ejection fractions, strains, MAPSE, and TAPSE.

Conclusion: Automatically derived shape and motion z-scores capture more discriminative information on disease effects than standard metrics, including volumes, ejection fraction, strain and valve excursions.

Background: Cardiovascular magnetic resonance (CMR) is a complex imaging modality requiring a broad variety of image processing tasks for comprehensive assessment of the study. Recently, foundation models (FM) have shown promise for automated image analyses in natural images (NI). In this study, a CMR-specific vision FM was developed and then finetuned in a supervised manner for nine different imaging tasks typical to a CMR workflow, including classification, segmentation, landmark localization, and pathology detection.

Methods: A ViT-S/8 model was trained in a self-supervised manner using DINO on 36 million CMR images from 27,524 subjects from three sources (UK Biobank and two clinical centers). The model was then finetuned for nine tasks: classification (sequence, cine view), segmentation (cine SAX, cine LAX, LGE SAX, Mapping SAX), landmark localization, pathology detection (LGE, cardiac disease), on data from various sources (both public and three clinical datasets). The results were compared against metrics from state-of-the-art methods on the same tasks. A comparable baseline model was also trained on the same datasets for direct comparison. Additionally, the effect of pretraining strategy, as well as generalization and few-shot performance (training on few labeled samples) was explored for the pretrained model, compared to the baseline.

Results: The proposed model obtained similar performance or moderate improvements to results reported in the literature in most tasks (except disease detection), without any task-specific optimization of methodology. The proposed model outperformed the baseline in most cases, with an average increase of 6.8% points (pp) for cine view classification, and 0.1 to 1.8 pp for segmentation tasks. The proposed method also obtained generally lower standard deviations in the metrics. Improvements of 3.7 and 6.6 pp for hyperenhancement detection from LGE and 14 pp for disease detection were observed. Ablation studies highlighted the importance of pretraining strategy, architecture, and the impact of domain shifts from pretraining to finetuning. Moreover, CMR-pretrained model achieved better generalization and few-shot performance compared to the baseline.

Conclusions: Vision FM specialized for medical imaging can improve accuracy and robustness over NI-FM. Self-supervised pretraining offers a resource-efficient, unified framework for CMR assessment, with the potential to accelerate the development of deep learning-based solutions for image analysis tasks, even with few annotated data available.

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±16.7ms and 38.4±3.2ms, 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.

Background: Electrocardiogram (ECG)-triggered cardiovascular magnetic resonance (CMR) can be challenging in patients with ECG unreliability. Pilot tone (PT)-triggered CMR may offer a reliable alternative.

Purpose: To evaluate the feasibility of PT-triggered CMR and compare its performance with ECG-triggered imaging across various sequences in patients with common cardiovascular diseases.

Methods: This prospective study included 50 participants (26 males, 24 females; mean age 46.0±19.0y), including 15 with normal CMR findings and 35 with various cardiovascular diseases. All participants underwent both PT-triggered and ECG-triggered CMR on a 3T MRI system. Imaging included T2-weighted imaging (T2WI), T1-mapping, T2-mapping, cine, late gadolinium enhancement (LGE), and post-contrast T1-mapping sequences. Image quality and quantitative measurements were evaluated, including T2WI signal intensity, native T1-mapping, T2-mapping, and extracellular volume fraction (ECV) values, and comparative signal-to-noise ratio (compSNR) and comparative contrast-to-noise ratio (compCNR) of cine and LGE images, left/right ventricular function. Inter-reader agreement was evaluated using the intraclass correlation coefficient (ICC). Comparisons between the two methods were performed using paired t-test or the Wilcoxon signed-rank test.

Results: No significant differences were observed in scanning times (p=.253-.864) or image quality (ICC: .589-1.000, p=.057-1.000) between PT- and ECG-triggered scans and images. Quantitative assessments showed good to excellent consistency (ICC=.843-.987). While PT-triggered LGE images showed higher compCNR (14.14±7.68 vs. 13.24±7.52, p=.016), other quantitative parameters showed no significant differences between PT- and ECG-triggered images. Six participants with hypertrophic cardiomyopathy or heart valve disease experienced false R-wave triggering during ECG gating, leading to motion artifacts, which were not visible in PT-triggered images.

Conclusion: PT-triggered cardiac MRI provides comparable image quality and quantitative assessments to ECG-triggered sequences and may offer advantages in minimizing motion artifacts, particularly in patients with conditions affecting ECG reliability, making it a promising alternative for cardiac MRI synchronization.

Background: Timing of pulmonary valve intervention (PVI) for pulmonary regurgitation in patients with repaired tetralogy of Fallot (TOF) is guided by right ventricle (RV) volumetric/function assessment on cardiac magnetic resonance (CMR) indexed to actual body surface area (BSA). However, different RV contouring techniques and BSA formulae exist with significant variability in reported measurements. We aimed to review the protocols reported in CMR studies of PVI in TOF.

Methods: A search of electronic databases (Embase and MEDLINE) was performed to identify studies published between February 28, 2010 and February 28, 2025 which assessed adults with repaired TOF before and after PVI using CMR. RV contouring methods and BSA formulae were reviewed.

Results: After screening 610 references, a total of 27 studies met the criteria and were included. All studies were of only level III or level IV (lowest) levels of evidence. Most studies (81%, 22/27) did not specify the RV contouring technique used and none defined the RV basal slice. Of the five studies describing the RV contouring technique, four excluded trabeculations/papillary muscles from the RV volume and one included these structures. No studies reported the formula used to calculate actual BSA.

Conclusion: RV contouring technique and BSA methodology utilized in CMR studies of PVI in TOF is poorly reported. Given the importance of severity thresholds for RV volumes in triggering intervention in clinical practice, clear reporting and standardization of RV contouring and BSA methodology should be mandatory. Further research into the optimal RV thresholds for PVI based on clear contouring methods is required.

Introduction: Mitral annular disjunction (MAD) is a pathologic fibrous separation of the mitral valve hinge point from the ventricular myocardium. The aims of this study were to describe the range of MAD distance by cardiovascular magnetic resonance (CMR) in children and young adults with connective tissue disorders (CTDs) versus a healthy control sample, and to assess the MAD distance as a predictor of adverse cardiovascular outcomes.

Methods: This was a retrospective, single-center study of healthy subjects and patients with Marfan syndrome, Loeys-Dietz syndrome, Ehlers-Danlos syndrome, or nonspecific CTD who underwent CMR between January 01, 2000 and January 01, 2020. The MAD distance was measured from the 2-chamber, 4-chamber, and left ventricular outflow tract views in systole and diastole and analyzed as absolute values as well as indexed to BSA and height. The primary outcome was a composite defined as the presence of significant ventricular arrhythmias, cardiac arrest, and/or death. Age-adjusted odds ratios with 95% confidence intervals and c-statistic are reported. Classification and Regression Tree analysis was performed to identify the most discriminating binary threshold to predict the occurrence of the composite outcome.

Results: Around 30 healthy control subjects and 254 patients with CTD met inclusion criteria. The mean ± SD age at initial CMR was 17±6years for patients with CTD and 14±3years for controls. The mean MAD distance was larger in patients with CTD compared to the control sample, and the maximum MAD distance in the control sample was 3.6 mm. Median follow-up in the CTD group was 5 years (IQR 3-11years). Thirty-four (15%) patients met the composite outcome. Systolic MAD distance was positively associated with the composite outcome. The optimal binary threshold for height-indexed maximum systolic MAD distance was 0.033 mm/cm with an event rate of 18.6% at/above threshold versus 2.6% below threshold (AUC 0.74). The association was independent of other important clinical predictors.

Conclusion: A small MAD distance can be measured in healthy children and young adults. Children and young adults with CTD have a longer MAD distance than healthy control subjects, and a longer MAD distance is associated with adverse outcomes.

Myocardial perfusion imaging plays a central role in the management of patients with known or suspected coronary artery disease (CAD) and increasingly in patients with suspected ischemia with normal coronary arteries (INOCA) as well as anomalous origins of the coronary arteries and Kawasaki disease. Stress perfusion cardiovascular magnetic resonance (CMR) is recognized by international guidelines, with several Class 1 indications for the detection of abnormal myocardial blood flow in these clinical scenarios and offers excellent diagnostic accuracy and independent prognostic value. While visual interpretation of the perfusion data is the prevailing analysis method in clinical practice, quantitative perfusion CMR is at least as accurate for the detection of significant obstructive CAD and provides a more accurate estimation of the total ischemic burden in patients with CAD. Moreover, quantitative myocardial perfusion analysis provides unique insights into the pathophysiology of myocardial ischemia, including microvascular disease in INOCA. Quantitative perfusion CMR can be fully automated, is user-independent, and may facilitate more widespread use of the modality. The aim of this Society for Cardiovascular Magnetic Resonance (SCMR) expert consensus document is to provide recommendations for the acquisition and analysis of quantitative myocardial perfusion CMR to facilitate standardization of methodology. This paper also discusses research and development goals to address current limitations, to ensure data reliability and validity, to create the basis for future multi-vendor and multicenter research, and to broaden the clinical use of quantitative perfusion CMR.

Background: Exercise cardiovascular magnetic resonance (ExCMR) imaging using supine in-scanner ergometer has shown promise in differentiating pathological dilated cardiomyopathy (DCM) from physiological exercise-induced cardiac remodeling. Since 2020, the National Heart Centre Singapore (NHCS) has incorporated ExCMR into its clinical workflow for patients with suspected DCM. This study aims to compare the costs associated with ExCMR versus conventional CMR in the evaluation of DCM.

Methods: A retrospective analysis was conducted on patients referred for conventional CMR between 2016 and 2019, and those referred for ExCMR from 2020 to 2023. Both imaging modalities followed standardized protocols, with ExCMR incorporating additional assessments during peak exercise. Costs were recorded in Singapore dollars (SGD) prior to the application of healthcare subsidies.

Results: The total cost for conventional CMR was SGD 1831.36, while ExCMR was associated with a higher initial cost of SGD 2336.48. However, ExCMR resulted in significantly fewer abnormal imaging findings and a reduced need for follow-up investigations (6.5% (9/139) vs 56.8% (71/125), p<0.001). A decision tree analysis and probabilistic sensitivity analysis (PSA) revealed that diagnosing 1000 suspected DCM patients with ExCMR could result in a cost savings of approximately SGD 182,323 compared to conventional CMR, with a 64% probability of being cost-effective.

Conclusion: These findings indicate that ExCMR offers a physiologically informative approach for diagnosing DCM, with the potential to reduce overdiagnosis of cardiac dilatation in active, healthy adults. Although further research is necessary to assess long-term outcomes, ExCMR appears to be a cost-effective imaging modality for DCM diagnosis, warranting reconsideration of its perceived higher cost.

Background: Cardiac magnetic resonance imaging (CMR) studies contain a wealth of information on a patient's cardiovascular status. The ability to extract this data from free-text reports could serve to automate clinical decision support tools and generate data for retrospective clinical knowledge discovery, and clinical operational purposes. Few studies have examined the automatic extraction of data from free-text CMR reports, and the existing studies that do have key limitations, including small sample size and disease-specific data extraction. Existing studies also fail to extract features associated with the cardiovascular conditions that reflect nuances in natural language, such as uncertainty, severity, subtype, and anatomical locations of the condition. The goal of this study was to build a broad named entity recognition model to automatically extract a broad variety of common CMR findings and their associated attributes from CMR reports.

Methods: We fine-tuned a Large Language Model Meta AI (LLaMA) model trained to identify 34 cardiovascular conditions and their associated attributes, including certainty, severity, location, and subtype of the condition. This model was trained on 1778 MRI reports and tested on 397 reports in an held-out test set and another 428 reports from another site in our hospital system with independent radiology practice and scanners.

Results: Our model shows robust performance in predicting the mention of the 31 cardiovascular conditions (average F1=0.85). It also showed strong performance predicting attributes, including certainty (average F1=0.97) and severity (average F1=0.97). Model performance on the external validation set was generally slightly lower than the internal validation set, but performance was still strong (average F1=0.78 for mention, 0.97 for certainty, and 0.96 for severity).

Conclusion: CMR-LLaMA has strong performance identifying a variety of concept mentions and moderate accuracies in extracting a selection of other associated attributes. NLP models can be used to automate the extraction of data from CMR reports to potentially assist with clinical and research workflow.