Visualization of Hemodynamics in a Silicon Aneurysm Model Using Time-Resolved, 3D, Phase-Contrast MRI

A Silicon Intracranial Aneurysm Model 3 Times Actual Size.

Rotational digital subtraction angiography was performed for a 67-year-old woman with an unruptured right middle cerebral artery bifurcation aneurysm. We obtained a 3D dataset of the right internal carotid angiogram with frame rate of 8.8, matrix of 512 × 512, and with 15 mL of iodinated contrast material (injection rate of 3 mL/s). The greatest dimension of this intracranial aneurysm was 10 mm. The aspect ratio (defined as depth/neck width)5,6 of the aneurysm was 4.8. This 3D dataset of the right internal carotid angiogram was reconstructed into multiple 2D axial sections with the thickness of 0.5 mm. These DICOM datasets were processed to reconstruct the lumen of the vessels in 3 dimensions and then transferred to a 3D printer by using powders and adhesive to produce a master cast of the vascular lumen. Based on this master cast, a realistic silicon model, 3 times actual size, containing the lumen of the original vessels was then constructed (S. Inagawa, unpublished data). Therefore, the size of the intracranial aneurysm in our model was 30 mm.

Flow Velocity Profile of the Right Middle Cerebral Artery of the Patient.

Flow velocity measurement was performed with a 1.5T MR unit (Signa Infinity TwinSpeed with Excite XI [version 11] GE Medical Systems, Milwaukee, Wis) using 2D cine phase contrast MR imaging. The imaging parameters were as follows: repetition time (TR)/echo time (TE)/number of excitation (NEX), 29/5.6/2; field of view (FOV), 100 × 100 mm; matrix, 192 × 128; thickness, 4 mm; velocity encoding (VENC), 80 cm/s; imaging time, 3 minute 5 seconds; imaging plane, sagittal plane; number of phases during one cardiac cycle, 30. Flow velocity profile of the M1 segment of the right middle cerebral artery of this patient was calculated by a workstation. Patient’s pulse rate was 84 beats/min, systolic maximum velocity was 44.8 cm/s, and average velocity was 26.2 cm/s. Reynolds number is the ratio of inertial to viscous forces in a fluid. The Reynolds number ranged from 102.1 to 350.5 and was 205.5 on the average velocity in the right middle cerebral artery of the patient. The Womersley parameter represents the ratio of the vascular diameter to the oscillating boundary layer thickness. The Womersley parameter in the middle cerebral artery of this patient was 2.24. The peak systolic and average velocity in this patient were lower than those of previously published data and they were thought to depend on the patient’s age.7,8

Phantom Circuit.

A phantom circuit was a closed circuit with a reservoir tank, a computer controller, a pulsatile pump (CardioFlow 1000MR; Shelly Medical Imaging Technologies Company), the silicon model, and simulation vessel. We ran a 53% weight glycerin solution (T1 value, 1005 ms; T2 value, 86 ms; kinematic viscosity, 14.6 × 10⁻⁶ m²/s, temperature; around 20° centigrade) in pulsatile flow with the cardiac cycle of 2 seconds by using the pump. We measured relaxation times with the MR scanner, viscosity with a viscosimeter, and temperature with a thermometer. Flow velocity measurement was done with the 1.5T MR unit using 2D cine phase contrast MR imaging. The imaging parameters were as follows: TR/TE/NEX, 29/5.2/1; FOV, 160 × 160 mm; matrix, 256 × 128; thickness, 4 mm; VENC, 150 cm/s; imaging time, 4 minutes 18 seconds; number of phases during 1 cardiac cycle, 20. Flow velocity profile of the M1 segment of the middle cerebral artery of this vascular model was calculated by a workstation. The Reynolds number ranged from 80.8 to 584.3 and was 262.5 on the average velocity in the right middle cerebral artery of the silicon model. The Womersley parameter was 2.41. Matching the Reynolds number and the Womersley parameter is the most important consideration in guaranteeing similarity of in vivo and in vitro hemodynamics. To match these nondimensional parameters the velocity and cardiac cycle were adjusted and were therefore different between our model and the patient. The Reynolds number and the Womersley parameter in our model were similar to that in humans so blood flow conditions in the patient were reproduced in our silicon model.

MR Scanner.

We used a 1.5T MR scanner (Signa Infinity Twinspeed with Excite XI [version 11]; General Electric) with a commercially available quadrature head coil.

Imaging Sequence.

Imaging sequence of 4D-Flow was derived from radio-frequency-spoiled gradient-echo sequence.3 Imaging was retrospectively gated by electrocardiogram wave produced by the controller of the pulsatile pump. The sequence used a segmented acquisition in the section direction, in which 4 sections were acquired per RR interval and thus, with the 3 encodings and one reference encoding, the true temporal resolution of the data was 4 × 4 × TR. The reconstruction retrospectively created the 20 temporal phases by using linear interpolation. The phase images encoded for x, y, and z directions were corrected by a correction for Maxwell phase effects9 and the Markl-Bammer correction,10 and 3D velocity data for each 3D voxel was generated. The data including velocity components in x, y, and z direction and time provided us with 4D-Flow information.

Imaging Parameter.

The imaging parameters were as follows: TR/TE/NEX, 5.8/2.1/1; flip angle, 15°; FOV, 140 × 140 × 108 mm; matrix, 160 × 160 × 36; voxel size, 0.88 × 0.88 × 3 mm; VENC, 20 cm/s; imaging time, 48 minutes; imaging plane, transverse plane; maximum gradient magnetic field, 40 mT/m; slew rate, 150 T/m/ms. VENC was set to 20 cm/s, because the intraaneurysmal flow velocity was estimated to be very low based on the large number of the aspect ratio of this aneurysm model.5,6

Postprocessing.

The data obtained by 4D-Flow were converted into a data format suitable for flow visualization software (EnSight; Computational Engineering International, Apex, NC) by personal computer (Intel Pentium4 CPU, 3.2 GHz, 2 GB of RAM, Linux OS) in approximately 90 minutes. We forwarded these converted data to another personal computer (Intel Pentium4 CPU, 3.2 GHz, 2 GB of RAM, Windows XP) and calculated time-resolved 2D velocity vector fields and time-resolved 3D streamlines with the use of visualization software “EnSight.” We processed each set of images within 5 minutes for a total processing time of approximately 1 hour.

We define some of our visualization tools: 2D velocity vector fields are arbitrary 2D planes extracted from the 3D imaging volume; 3- directional velocities are displayed over time as color-coded vector fields;4 3D streamlines are integrated traces along instantaneous velocity vector field, color coded according to the local velocity magnitude;4 the flow rate was shown by a color displayed in each legend.