Measure the accuracy of a novel, rapid Cartesian four-dimensional flow-sensitive magnetic resonance imaging (4D Flow MRI) acquisition using ultrasound flow probes and two-dimensional (2D) phase contrast (PC) as reference standards in patient specific models of total cavo-pulmonary connection (TCPC). Congenital heart disease (CHD) is a prevalent condition affecting almost 1% of all births[1]. In these children, characterization of flow patterns, anatomy, and function guides treatment and prognosis. Cardiac MRI enables evaluation of anatomic detail with same scan acquisition of 4D flow allowing large volume hemodynamic analysis including flow patterns and function. Increasing the implementation of this powerful technology will require short scan times and improved post processing efficiency. A novel technique for time resolved 3D velocity encoding using k-t acceleration and variable density signal averaging Cartesian acquisition offers a method of 4D flow analysis while addressing the inherent difficulties of Cartesian encoding including scan time, spatial resolution, and motion. The Cartesian acquisition is advantageous given improved post processing time but is yet to be validated in phantoms and in the clinical setting. A TCPC connection model was utilized to create a more complex flow pattern and more closely simulate in-vivo conditions.

Patient specific models: Three physical models of TCPC (Figure 1) were created using 3D printing according to an IRB-approved and HIPAA-compliant protocol. The TCPC was segmented from images acquired as part of a clinical cardiac magnetic resonance angiography in MIMICs (Materialise, Leuven, Belgium). The segmented TCPC volume was saved in STL format and printed using a 3D printer. Polyethylene tubing was attached to the inlets and outlets of the TCPC and then connected to a perfusion pump.

MRI: MRI was performed on a clinical 3T scanner (GE Discovery MR 750, Waukesha, WI) with a 32-channel phase array body coil. Each TCPC model was imaged separately secured in a saline bath. Water was pumped through the TCPC models at four different flow rates (1, 2, 3, and 4 L/min). 2D PC-MRI sequence and 4D Flow MRI were performed at each flow rate.

k-t Accelerated Cartesian 4D Flow MRI: kat ARC [2], a spatiotemporal-correlation-based autocalibrating parallel imaging method with cardiac motion adaptive temporal window selection, was used for fast imaging. As shown in Figure 1, data was collected with a variable density random (VDR) k-t sampling scheme to improve overall reconstruction accuracy and reduce coherent residual artifacts [3]. In reconstruction, a static tissue removal scheme [3] was used to identify voxels with no flow or motion and remove signal from such static voxels before kat ARC processing to reduce residual aliasing artifacts at high acceleration.

Variable density signal averaging: Signal averaging scheme was used to suppress respiratory motion. The number of excitations (NEX) varies based on k-space location for scan efficiency, with linearly decreasing NEX from the highest at center k-space toward outer k-space (Figure 2). Such NEX scheme at near-center k-space is termed CNEX.

Analysis: 2D PC flow was quantified using ReportCard 2.0 (Advanced Workstation, GE Healthcare, Waukesha, WI). 4D flow MRI was analyzed in Ensight (CEI Inc. Apex, NC). 2D planes through the Glenn, Fontan, LPA, and RPA (Figure 3). Bland-Altman analysis was performed to compare 4D Cartesian acquisition to flow sensor, 2D PC MRI, and PC VIPR.

Bland Altman analysis (Figure 4) reveals minimal differences between 2D flow and Cartesian 4D flow measurements with bias of -0.087±0.19. Correlation of 2D PC and 4D flow to US probe flow rate (Figure 5) confirms excellent correlation between MRI flow measurements an US measured flow with slopes of 1.1941 L/min and 1.183 L/min respectively (R2 of 0.9871 and 0.9365, respectively). These findings confirm the in vitro accuracy of both 2D PC and 4D flow in patient specific models of CHD. The improved post processing efficiency and limited resolution and temporal trade-off of 4D flow using k-t acceleration and variable density signal averaging promises to be a viable, widely used method for the evaluation of complex CHD. 4D flow with k-t acceleration and variable density signal averaging demonstrates excellent correlation with US probe flow measurements and negligible difference with 2D PC measurements in a 3D printed TCPC model utilizing realistic geometries for CHD. The benefits of increased post-processing efficiency in Cartesian acquisition promises more widespread use of 4D flow in treatment planning of CHD.