Assessment of Atrial and Ventricular Function Using Static 3D Whole-Heart SSFP and 4D Flow Cardiovascular Magnetic Resonance Imaging

Background: The atria play a crucial role in ventricular filling and overall cardiac performance. Although ventricular volume and function have traditionally been regarded as the primary determinants of cardiovascular prognosis, changes in atrial volume and function are now recognized as significant contributors to various cardiovascular conditions, including repaired Tetralogy of Fallot and atrioventricular valve regurgitation. However, conventional cardiovascular magnetic resonance (CMR) methods for assessing atrial volume and function are limited in accuracy, as they typically rely on a single 2-dimensional (2D) long-axis or four-chamber view. In this study, we aimed to develop a novel technique for 3D quantification of atrial and ventricular volumes and function using commercially available static 3D whole-heart SSFP and 4D-flow CMR images.

Methods: Ten patients (6 male, median age 16 [7–32]-year-old) with congenital heart disease who were referred for gadolinium-based contrast-enhanced CMR exams at our institution were prospectively recruited for this study after obtaining informed consent. For each patient, post-contrast static 3D whole-heart SSFP and 4D flow CMR images were acquired. The static 3D whole-heart images were automatically segmented using an nn-UNet deep-learning network to delineate the atria and ventricles and to generate a 3D mesh model of the heart and great vessels. The 4D flow images were used to capture non-rigid 3D deformation of the heart across 30 cardiac phases. The computed 3D deformation fields were then applied to the static 3D mesh model to generate a dynamic, beating heart model capturing a full cardiac cycle. This model was used to calculate atrial and ventricular volumes and function. The computed atrial and ventricular volume and function results were then compared with those measured using conventional clinical methods from 2D long-axis, four-chamber, and short-axis views.

Results: Figure 1 presents an example of the 3D beating heart model at end-diastole and end-systole, along with atrial and ventricular volumes across 30 cardiac phases computed from the beating heart model. Tables 1 and 2 provide statistical comparisons of atrial and ventricular volumes and function derived from conventional clinical 2D images versus those obtained from the 3D beating heart model.

Conclusion: This study presents insights into the assessment of 3D atrial and ventricular volumes and function using static 3D whole-heart SSFP and 4D flow CMR images. The findings demonstrate that this technique may provide estimations of atrial and ventricular volume and function. Future work will focus on increasing accuracy, precision, and incorporating blood flow dynamics into a beating heart model to enhance physiological relevance and clinical applicability.

Acknowledgments: We would like to thank NHLBI for supporting this study (R01HL149807).

Table1: Comparison of Atrial Volume and Function Derived from Conventional 2D Images, and 3D Whole-Heart SSFP and 4D Flow Imaging. * Statistically significant. Values are presented as mean ± standard deviation. EDV: end-diastolic volume; EF: ejection fraction; and ESV: end-systolic volume.

Cardiac Function	Left Atrium (LA)			Right Atrium (RA)
	EDV	ESV	EF	EDV	ESV	EF
2D 4 chamber or 2 chamber	50 ± 29	25 ± 23	67 ± 11	57 ± 32	34 ± 22	56 ± 18
3D static SSFP & 4D Flow	82 ± 46	32 ± 28	65 ± 13	120 ± 66	52 ± 35	58 ± 13
p-value	0.001*	0.037*	0.621	0.003*	0.086	0.907
Correlation coefficient	0.943	0.958	0.490	0.692	0.528	0.360

Table 2: Comparison of Ventricular Volume and Function Derived from Conventional 2D Images, and 3D Whole-Heart SSFP and 4D Flow Imaging. * Statistically significant. Values are presented as mean ± standard deviation. EDV: end-diastolic volume; EF: ejection fraction; and ESV: end-systolic volume.

Cardiac Function	Left Ventricle (LV)			Right Ventricle (RV)
	EDV (mL)	ESV (mL)	EF (%)	EDV (mL)	ESV (mL)	EF (%)
2D short-axis view	150 ± 78	55 ± 34	65 ± 8	148 ± 68	73 ± 39	54 ± 10
3D static SSFP & 4D Flow	200 ± 96	71 ± 37	65 ± 5	227 ± 125	99 ± 67	59 ± 9
p-value	< 0.001*	0.026*	0.330	0.007*	0.041*	0.121
Correlation coefficient	0.956	0.849	0.789	0.895	0.900	0.578

Figure 1: Generated 3D Beating Heart Model at Two Cardiac Phases—End-Diastole and End-Systole—With Atrial and Ventricular Function. (A) 3D beating heart model at end-diastolic phase; (B) 3D beating heart model at end-systolic phase; (C) Left and right ventricular volumes and left and right atrial volumes throughout the cardiac cycle. The 3D beating heart model is generated from commercially available post-contrast static 3D whole-heart SSFP images and 4D flow images acquired from a 33-year-old female with coarctation of aorta.