To conduct ultrasound VFI experiments, an anthropomorphic phantom of the human thoracic aorta and a flow system to simulate hemodynamics in the aorta under VA-ECMO treatment was developed (Fig. 1). The thoracic aorta phantom illustrated in Fig. 1 (hereafter referred to as the aorta phantom) was designed and fabricated based on a protocol of polyvinyl alcohol (PVA) wall-less phantoms [23] so that it resembles the internal geometry of the human thoracic aorta that was extracted from anonymized contrast-enhanced computed tomography images (Fig. 2 A). The use of human-origin data was approved by the local ethics committee at graduate school of engineering, Tohoku University. The fabricated phantom had four outlets: the ascending aorta (Ø26.9 mm), the brachiocephalic artery (BCA, Ø10.7 mm), the confluence of the left common carotid and left subclavian arteries (Ø7.9 mm), and the mid descending aorta (Ø23.7 mm). The fabrication process of the aorta phantom is described in Supplementary Content 1. In addition, as shown in Fig. 1, a flow system was built to generate complex flow dynamics, such as the mixing zone, in the aorta phantom. The ascending aorta inlet was connected to a cardiac pulsatile pump (A mechanical piston pump with two ball valves; valve and outlet diameter: 15.8 mm; 55-3305, Harvard Apparatus, MA, USA) via a 90 cm tube (inner diameter: 15.8 mm), while the descending aorta outlet and the aortic trifurcation outlets were connected to water tanks, which were placed at a height (80 cm) relative to the aorta phantom to mimic diastolic blood pressure. Blood pressure in the phantom was measured at the BCA outlet using a pressure monitoring system (TSD104 and MP160 system, BIOPAC Systems Inc., CA, USA). Furthermore, in order to simulate retrograde ECMO flow in the phantom, a centrifugal pump for ECMO treatment (hereafter referred to as the ECMO pump; CAPIOX SP-101, Termo Corp., Tokyo, Japan) with an outlet flow meter (DigiFlow 6710 M, KRONE Corp., Tokyo, Japan) was used. Note that no artificial lung module was connected to the system. A tube (inner diameter: 12 mm) from the ECMO pump outlet was inserted into the flow circuit without cannulas as shown in Fig. 1, so that the tip of the tube was located at the outlet of the descending aorta of the phantom. Finally, a blood-mimicking fluid consisting of 10% aqueous solution of glycerin (072-00621, FUJIFILM Wako Pure Chemical Corp., Osaka, Japan) (density: 1.02 kg/m3) and 0.5% of 8.0–13.0 μm silica particles (density: 2.1 kg/m3; Godd ball B-25C, Suzukiyushi Industrial Corp., Osaka, Japan) was used as the fluid to circulate in the aorta phantom to mimic the density and acoustic properties of human blood [24, 25]. Although the density of silica particles was higher than that of fluid, the negligible sedimentation velocity of the particles due to their small diameter ensures that this property does not affect the flow investigations conducted in this study.

Schematic diagram (left) and photograph (right) of the thoracic aorta phantom and flow system

A The thoracic aorta model printed with a 3D printer and ultrasound images of the aorta phantom acquired with a sector probe. B B-mode images of the aorta phantom taken with a linear array probe. On the upper-right side, the location of the probe for each measurement is shown. The diameters are shown in mm

To simulate the hemodynamics in the aorta phantom with a mixing zone at different locations, two ECMO flow rate settings were used in this study: low (0.35 L/min), and high (1.0 L/min). For both ECMO conditions, the same cardiac output (stroke volume: 30 mL; heart rate: 90 bpm; cardiac output: 2.7 L/min, systole/diastole ratio: 40/60%) was used for the pulsatile pump. In this pulsatile pump setting, the blood pressure at the BCA without ECMO flow was measured to be 81/57 mmHg, which was similar to low output syndrome (stroke volume < 40 mL) [26]. The flow rate at the outlet of the trifurcation aorta was measured to be 1.3 L/min in low ECMO flow and 1.7 L/min in high ECMO flow, respectively. Prior to VFI experiments, conventional pulse Doppler imaging in the aorta phantom was performed to show the basic flow properties in the phantom and the images are shown in Supplementary Content 2.

Imaging of the flow dynamics in the aorta phantom under two VA-ECMO conditions was performed using a VFI framework, so-called vector projectile imaging (VPI) [16]. To perform VPI measurements in the experiment, a research-purpose ultrasound platform (Vantage 256, Verasonics Inc., WA, USA) equipped with a 5 MHz linear array transducer (L11-5v, Verasonics Inc.) was utilized. The linear array probe was placed at three positions corresponding to the ascending aorta, distal arch, and mid descending aorta in the phantom (Fig. 2B) to image the flow dynamics in the entire thoracic aorta. The path length from the root of ascending aorta to those three imaging positions were approximately 50, 110 and 180 mm, respectively.

The detailed principle and algorithm of the flow vector estimation and VPI framework can be found in [15, 16]. In brief, the ultrasound system was configured to emit plane-wave ultrasound pulses from two different transmission angles (−10° and 10°) alternatively at a pulse repetition frequency of 10 kHz in order to simultaneously acquire two color-Doppler cineloops, which correspond to the Doppler shifts of each transmission angle. Those multiple color Doppler data was used to construct a linear equation based on Doppler equation to derive flow vectors (i.e., flow direction and speed at each pixel in each frame). Finally, the flow vectors were estimated by solving the linear equation with the least-squares method [15]. The final flow vector data was visualized by a dynamic vector rendering technique proposed in the VPI framework. In this study, for each ECMO flow condition, three VPI cineloops (respectively, acquired from three probe positions) were obtained. Each cineloop visualized a region of 38 mm (width) × 40 mm (depth) from the probe for a duration of 1.5 s that involved at least one cardiac cycle of flow data. These three cineloops were synchronized and synthesized into one video file (Fig. 3A-C) using the blood pressure waveforms (Fig. 3a) recorded at the BCA simultaneously in each VPI measurement.

Screenshots of the VPI cineloops of low ECMO flow condition (upper row) and high ECMO flow condition (lower row). The timing of each image corresponds to the A, B, and C labels in a showing a blood pressure waveform. b Definition of anterograde and retrograde flow directions for each image. c Range of anterograde and retrograde flow vectors. Speed of blue and gray region is above 0.25 m/s

Using the time in the blood pressure curve (Fig. 3a), a cardiac cycle (approx. 0.66 s) of VPI data was divided into three phases: late diastole (< 0.1 s), systole (0.1–0.4 s), and early diastole (> 0.4 s). In our experiment setting, it was not possible to determine the start time of the systolic action of the cardiac pump. In addition, the pressure curve showed a bi-modal shape due to incomplete (e.g., bouncing) ball valve actions in the cardiac pump that made identifying the systole phase difficult. Therefore, in this study, we defined the start time of the systole as that of the first observation of the anterograde flow vectors in the proximal end of the ascending aorta images (at 0.1 s) and the systole duration as 0.3 s from the pressure curve observation.

The visualized flow vectors in the thoracic aorta were further analyzed to assess temporal changes in the flow vector distribution. For this purpose, the flow vectors were categorized into five groups: fast anterograde, slow anterograde, fast retrograde, slow retrograde, and other flow vectors. For each imaging position, the anterograde and retrograde flow directions were defined as the flow direction from the pulsatile pump and the ECMO pump, respectively (Fig. 3b). For each VPI cineloop, flow vectors directed to ± 30° with respect to the anterograde or retrograde direction were categorized into the anterograde flow group or the retrograde flow group, and the other flow vectors were placed in the other flow vectors group. In addition, the flow vectors in the anterograde and retrograde flow groups were further divided into the fast or slow group using a threshold of 0.25 m/s (Fig. 3c). After categorization, the area of each flow group was counted for each frame, and the ratio of each flow group with respect to the flow area in the frame was calculated. To compare the ratio of respective flow groups between two ECMO flow conditions, mean and standard deviation of those ratio values in each cardiac phase (i.e., late diastole, systole, or early diastole) were derived for respective imaging positions and ECMO flow conditions. Welch’s t-test was used for the statistical analyses.