A pulsatile 3D flow relevant to thoracic hemodynamics: CFD - 4D Flow MRI comparison

Application Challenge AC7-04 © copyright ERCOFTAC 2021

Test Data

Overview of Tests

The experimental method presented in this section is described in detail by Puiseux et al. (2019) [2]. However, the MRI machine and sequence parameters differ. The PC-MRI scans were performed on a 1.5T scanner (Siemens MAGNETOM Sola, Siemens Medical Systems, Erlangen, Germany). Even though the MRI measurements presented in this AC are new, the phantom setup used is the same and the sinusoidal pulsatile flow regime studied is very close to the one previously investigated.

Description of Experiment

A scheme presenting the experimental setup is shown on Fig. 1. The pulsatile flow is delivered by a computer-controlled pump (CardioFlow 5000MR, Shelley Medical Imaging Technologies, London, Ontario, Canada), which allows the user to define the desired waveforms. To control the flow rate it delivers, an ultrasonic flowmeter (UF25B100 Cynergy3 components Ltd, Wimborne, Dorset, UK) is used. Futhermore to reduce the swirling motion of the fluid entering the phantom, a flow straightener is positioned downstream of the flowmeter and upstream of the phantom.

Concerning the PC-MRI acquisitions, one acquisition was made for 4D Flow (duration of 42.6 min) and several for 2D cine PC-MRI (duration of 3.6 min each). For each acquisition, the full k-space was recorded (ie there was no parallel imaging acceleration), where the k-space refers to the complex-values Fourier transform of the MR images measured. Indeed the signal collected during the MR experiment corresponds to the Fourier transform of the transverse magnetization signal and is thus expressed in terms of spatial frequencies. The MR images are reconstructed through an inverse Fourier transform applied on this k-space. All scans were retrospectively cardiac gated, which allows to reconstruct a certain number of phases of the whole cardiac cycle. 20 phases were acquired for the 4D Flow scan and 30 for the 2D cine PC-MRI scan. Concerning the spatial resolution, isotropic voxels of 2 mm were acquired for the 4D Flow scan. For the 2D cine PC-MRI, the isotropic pixel size was 0.8 mm, with a slice thickness of 6 mm to enhance the signal-to-noise ratio.

These settings and additional MR information are given in the following table:

	4D Flow	2D cine PC-MRI
Acquisition Duration	42.6 min	3.6 min
Cardiac Phases Number	20	30
Acquisition Plane	Coronal (xz-plane)	Transverse (xy-plane)
Acquired voxel size	2 x 2 x 2 mm	0.78125 x 0.78125 x 6 mm
Flip Angle	7°	20°
VENC	0.7 m/s (in-plane velocity)	0.7 m/s (through-plane velocity)
VENC	0.2 m/s (through-plane velocity)	0.2 m/s (in-plane velocity)
TR	51.84 ms	111.04 ms
Effective TE	4.15 ms	4.49 ms

Table 1: MR sequences parameters

As briefly explains above, MRI theory relies on recording the signal emitted by spatially encoded protons. Each proton is sequentially encoded: first along the slice-selection direction (also known as first phase-encoding and in 2D corresponding to the slice thickness), then along the so-called (second) phase-encoding direction, and finally along the last direction named frequency-encoding direction. The acquisition plane parameter matches the plane normal to the slice-selection direction and is an important piece of information, as the phase-encoding directions are associated with two major artifacts in MRI, namely aliasing and displacement artifacts.

Boundary Data

Inlet and outlet of the phantom are both connected to the whole hydraulic circuit and care is taken to avoid any air running into it. Before the measurements, the fluid was freely running through the circuit. Besides the flowmeter measurements, the high-resolution 2D cine PC-MRI acquisitions at the inlet enable to check whether the prescribed pulsatile flow is achieved. The averaged flow rate obtained at the inlet from 2D cine PC-MRI acquisitions is displayed in Fig. 4, together with the flow rate obtained from 4D Flow MRI at the same location.

Figure 4: Flowrate at the inlet over one cardiac cycle (duration: 1s)

Measured Data

The data collected are magnitude and phase images proportional to the proton magnetization and are saved as DICOM (Digital Imaging and COmmunications in Medicine) files. The first post-processing step is to reorder the data and to compute the velocity field from the phase images. The resulting dataset is written as a structured 3D grid in the VTK format (Visualisation ToolKit, Kitware, Inc, CliftonPark, NY) and is available for download : File:070121 4Dflow RawData 1.tar.gz and File:070121 4Dflow RawData 2.tar.gz. All post-processing steps, which will be further expanded on in the coming sections, are performed using in-house VTK-based Python scripts.

Measurement Errors

PC-MRI images suffer from some limitations: spatio-temporal resolution, noise, or artefacts related to motion, hardware or sequences. It is indeed the purpose of this Application Challenge to be able to compare velocity fields obtained by MRI acquisitions with the same fields obtained via CFD simulations, in order to quantify errors from the first method. However, both techniques need to be post-processed to enable comparison. Concerning the experimental 4D Flow MRI data, the MR images have been corrected for geometric distortion, noise and phase aliasing and have been spatially registered onto the CFD model thanks to an Iterative Closest Point algorithm. Thereby both techniques are expressed in the same coordinate system, yet not on the same grid.

Contributed by: M. Garreau^a, T. Puiseux^a,b, R. Moreno^c, S. Mendez^a, F. Nicoud^a —
^aIMAG, University of Montpellier, CNRS UMR 5149, Montpellier, France
^bSpin Up, Toulouse, France
^cI2MC, INSERM/UPS UMR 1297, Toulouse, France
^dALARA Expertise, Strasbourg, France