Introduction

4D Flow is a rapidly evolving MR technique that shows clinical promise with a wide variety of cardiovascular applications, including congenital heart disease. As a volumetric technique, it can simplify acquisition of velocity data while simultaneously visualizing anatomic structures (1–4). Flow can be quantified at any level within the scanned field of view and 4D flow imaging enables calculation of cardiac volumes and biventricular function(3)(5). We hypothesized that flow measurements may be accurately performed across a variety of scanning parameters and field strengths typically used in clinical practice. We therefore established a multicenter clinical study of 4D Flow at sites using MRI for management of patients with suspected atrial septal defects (ASD). We investigated the consistency of blood flow measurements obtained from 4D Flow data at multiple locations along the systemic and pulmonary vasculature by multiple observers.

Methods

With local IRB approval and HIPAA compliance, we selected patients referred for evaluation of ASD at four academic centers between (March, 2015and month, year) for retrospective analysis. 26 patients (mean age 45 years, 4 male) were identified. 4D Flow acquisitions were performed on clinical MRI scanners (70% at 1.5T, 30% at 3T) (GE Healthcare, Wi) after injection of gadolinium–based contrast agent. The in-slice scanned spatial resolution was less than 2 mm and slice thickness was maximum 3 mm. Temporal resolution varied between 48 and 60ms and the maximum scanning time was 10 minutes. Image post-processing was done by 2 observers using cloud-based software (Arterys Inc, CA). After software background eddy-current correction, ASD were visualized within volumetric data sets with different rendering techniques (color-coded velocity overlay, vector overlay and streamlines) and classified. The standard planes used were long and short axis over atrial septum. Pulmonic and systemic flows were measured in 3 different ways, respectively and 3 shunt ratios (Qp/Qs) were calculated. More precisely, pulmonic flow was measured at the level of the pulmonary valve, of the main pulmonary artery and as a sum of the flow of right and left pulmonary arteries. Systemic flow was measured at the level of aortic valve, ascending aorta and as a sum of the flow of superior vena cava above azygos and descending aorta. In 8 patients, shunt volumes were directly quantified at ASD level with a cross-sectional cut of the flow jet. In 18 patients a second observer repeated the measurements. Statistical analysis was performed with SPSS software (IBM, NY). Quantitative data were compared with Bland-Altman tests and Pearson t-tests. Inter-observer variability was done with intra-class correlation coefficient (ICC).

Results

We identified 17 ASD secundum, 2 primum, 2 sinus venosus SVC and 2 sinus venosus IVC. Three of the patients had also PAPVR. Pulmonic flow measured at 3 levels had a mean of 9.25 (+/-4.49), 9.25 (+/-4.44) and 9.25 (+/-4.47) respectively. Systemic flow was 5.15 (+/1.73), 5.10 (+/-1.75) and 5.19 (+/-1.64) respectively. Mean Qp/Qs ratios were 1.83 (+/-0.80), 1.84 (+/-0.74) and 1.80 (+/-0.73). The Pearsons’ correlations between Qp/Qs ratios were very good, with r=0.94 (first and second measurement) and r=0.89 (first and third measurement). Mean shunt volumes at the 3 levels were 4.10 (+/-3.93) l/min, 4.15 (+/-3.74)l/min, and 4.05 (+/-3.84)l/min, and mean direct quantification of the ASD was 3.68 (+/-1.93)l/min. The Pearson’s correlations were r=0.97 (first and second measurement), r=0.93 (first and third measurement) and r=0.82 (first and direct shunt quantification). ICC for inter-observer variability for shunt ratios were 0.92, 0.94 and 0.79.

Discussion

4D Flow provides multiple possibilities for flow quantification (pulmonic and systemic). In the case of aliasing or turbulent flow at a specific level, the clinical question will still be answered by measuring at another location.

Conclusions

With this study, we show that 4D Flow is a feasible technique for visualization and quantification of flow volume and shunt volume and fractions. Quantitative data appear reproducible and consistent across multiple observers and measurement locations, from data obtained across multiple sites, which is essential for the implementation of 4D flow imaging in clinical practice.

Acknowledgements

No acknowledgement found.