In the present study, comparing volume measurement by CEUS with thermodilution in a controlled in-vitro set-up, we have demonstrated that there is a good correlation between volumes measured by CEUS and by thermodilution (rs = 0.94 with the LDRW double fit method). Using the Bland-Altman analysis, there was a good level of agreement (bias −108 ± 67 mL with the LDRW double fit method) between both methods with only a modest underestimation of the true volume by CEUS (bias −40 ± 28 mL with the LDRW double fit method). Interestingly, compared to the gold standard of thermodilution, CEUS demonstrated to provide a more accurate measure of a known volume in this in-vitro set-up with a bias of only −40 ± 28 mL compared to 84 ± 62 mL for the thermodilution method.
In general, the thermodilution technique overestimated all volumes (Figure 4). The overestimation of the thermodilution volumes can be explained, in the first place, by loss of heat to the surroundings due to conduction across the tube wall. Heat loss is more marked at low flows and large volumes (Figure 4) due to longer contact-time and larger surface area, respectively. In order to minimize this effect, we isolated the whole set-up and the temperature was kept stable in a narrow range around 37°C. Despite this, the volume overestimation was consistently present even at small volumes (356 mL) and high flows (4 liters per minute), when heat loss is expected to be minimal.
In our study, we used a TEE matrix probe, as this is a more realistic set-up to perform volume measurements by CEUS during open heart surgery and the perioperative phase. In general the proposed methods are also feasible by TTE, as shown in studies by Mischi [7, 8]. In comparison to TTE, TEE is closer to the heart with minimal ultrasound attenuation in between. Whether the results with a TEE or TTE probe are interchangeable needs to be investigated in future research.
As no calibration is available for the adopted TEE probe in literature, we calibrated the probe by determining the relationship between the SonoVue® concentration and measured acoustic intensity. The calibration was performed at two different temperatures, namely ambient room temperature (20°C) and the typical temperature used for cold thermodilution (4°C). To this purpose, we diluted SonoVue® in saline at 4°C to obtain sufficient signal-to-noise ratio for the thermodilution IDC and because this temperature is routinely used for clinical thermodilution measurements. We found UCA bubbles to be stable longer at lower temperatures , however at the cost of a reduced echogenicity (Figure 2). This is in line with the reported decrease in bubble stability at higher temperatures [17, 18]. At higher temperature the UCA bubbles expand 5% by gas expansion and show increased acoustic backscatter . Our calibration showed no attenuation below 1.0 mg/L at 4°C. Higher doses will produce a non-stationary (concentration dependent) shadow effect that will influence the IDC quantification, possibly affecting the MTT. We calculated our diluting volume for SonoVue® as the total volume in the circuit, which was over 1 liter. Nevertheless at the lowest volume of the tube network (356 mL) and highest flow (4 L/min) we registered some attenuation in the IDC, probably due to a low effective diluting volume of less than 1 L. In spite of this, the MTT reproducibility at these settings was high with an intraclass-correlation of ICC = 0.99 and an average difference with the true volumes of +2 mL for CEUS and for thermodilution +74 mL (Figure 4). Therefore, the influence of attenuation on the MTT assessment seems negligible.
The IDCs are fitted using dedicated models that can possibly influence volume measurement thus leading to over- and/or underestimation. With this respect, the LDRW model seems superior as the fitting is based on both the ascending and descending slopes of the IDC which makes it less sensitive to noise. Still, this may lead to underestimation of the IDC tail compared to more common models like mono-exponential and power-law model . In a study of Ugander et al. magnetic resonance imaging was evaluated as a method for estimating pulmonary blood volume . An in-vitro validation of pulmonary blood volume measurements was carried out, showing a mean difference between the measured volumes and true volumes of 10 ± 2% for the peak-to-peak method and 4 ± 3% for the center of gravity method . The center of gravity and peak-to-peak methods do not use model fitting for the MTT estimation, and they are more sensitive to low signal-to-noise ratios and contrast recirculation. Moreover, they do not provide a physical interpretation of the investigated convective diffusion process, as provided by the LDRW model . The LDRW model provides better fits of skewed IDCs , which are present at high flows and small volumes. Our mean difference for CEUS and the true volumes was −6.5 ± 2.8% using the LDRW double fit method and −8.5 ± 2.9% using LDRW impulse response method. A volume underestimation of −3.3 ± 2.3% was also found by different models and settings in another in-vitro study for volume quantification with magnetic resonance imaging . In particular a slightly lower accuracy and volume underestimation by the impulse response method was reported; however, this was not confirmed by intra-thoracic blood volume measurements in the volunteers. The advantage of the impulse response method consists of making the measurement robust to recirculation and variations in the injection function. Moreover, the identification of the full transpulmonary dilution impulse response brings additional information, possibly adding diagnostic value to the analysis .
Another finding to be discussed relates to the volume underestimation by CEUS. This is likely to be explained with the bubble transport kinetics. It has been reported that bubbles, especially for laminar flow (Reynolds < 2000), show a velocity profile that differs from that of the carrier fluid, leading to a shorter MTT. In particular, while the carrier fluid shows a typical parabolic flow profile, bubbles are reported to travel with a “flatter” profile, whose average velocity over the tube cross section is higher than that of the carrier fluid [23, 24]. Additional explanations, to be verified in future studies, might relate to the concentration profile of bubbles across the tube.
In-vivo studies have another carrier fluid, blood, which could influence the indicator dilution curve. As blood is a more viscous carrier fluid than water, the Reynolds number will decrease. As a result, the diffusion coefficient is expected to be lower. The diffusion coefficient influences the parameter λ of the LDRW model, which equals the Peclet number divided by 2. The Peclet number represents the ratio between diffusion and convection time. An increase in viscosity produces therefore an increase in λ, whereas μ, which is representative of the MTT, is not affected .
Our study may provide a good alternative for volume measurement in the perioperative setting and in critical care. The safety of SonoVue® has been investigated in different settings. In a large retrospective study on assessment of adverse events in 28 Italian Centers, serious adverse events were found in 0.0086% [26, 27]. SonoVue® microbubbles are composed of SF6 gas with a phospholipid monolayer shell. The elimination of the SF6 gas via the lung is reported to be, even in patients with obstructive pulmonary disease and pulmonary fibrosis, in the same range as in healthy volunteers with a 80-90% clearance within 11 minutes . The phospholipid monolayer is metabolized in the liver. These monolayers are commonly used in the formulation and manufacturing of liposomes, a drug delivery system that is approved by the United States Food and Drug Administration (FDA) and considered biologically safe . Even in blunt abdominal trauma patients, the use of SonoVue® was proven to be safe . Only in patients with a recent cardiac infarction or heart failure class III and IV, the European Medicines Agency (EMA) took precautions and these conditions are contraindicated for use of SonoVue®.
Limitations of our study relate to the thermodilution technique. Even with extended isolation of the set-up, volumes were overestimated using thermodilution. The pressure wires are normally used in coronary arteries to measure flow with thermodilution. These wires are very thin and ideal in our in-vitro set-up but not used for larger blood volume measurement. Thus, this correlation cannot be extrapolated in-vivo. Further research will be needed to compare CEUS and evaluate the margin of error with clinically used thermodilution methods for volume measurement, such as PiCCO® (Pulsion Medical Systems, Munich, Germany), and to estimate its value as a minimally-invasive and bedside-applicable technique in the ICU and operating room [3, 30].