Variation Between Imaging and Non-Imaging Transcranial Doppler Systems Versus a Phantom Doppler Generated Velocity

Abstract

Stroke for children with sickle cell disease (SCD) is associated with early mortality and life-long morbidity [1, 2]. In children with sickle cell anemia (SCA), high blood flow velocity in the distal internal carotid artery or middle cerebral artery, as determined by non-imaging transcranial Doppler sonography, predicts strokes in children with SCD [3].

Previous investigators have demonstrated that imaging Doppler sonography produces lower velocity measurements than non-imaging equipment [4, 5]. At the same time, different non-imaging systems used on the same patient may result in different measurements. The variation of TCD velocity from one machine to another is only one form of variation; the ultrasonographer may also have significant variation on the same child on the same day. Previously, we demonstrated that the intra-observer coefficient of variation for a trainer for TCD velocity exams for our SPIN and SPRING trials, obtained in 21 participants for the right and left MCA, was 7.6% and 12.0% [6]. These results are consistent with the experience of the STOP Trial group, where the estimate of the coefficient of variation, without a formal analysis, was 10%–15%. We now seek to determine the variation of four commercial non-imaging TCD machines from 3 companies and one imaging TCD machine. With a phantom Doppler, we produced a reference velocity of 191 cm/s, a clinically relevant TCD velocity, and a conditional outcome depicting a significant risk for a stroke in a child with SCA. We employed a CIRS 769 Doppler flow pump and 524 ATS model tissue mimicking Doppler flow phantom to obtain TAMMV from imaging and non-imaging TCD systems. The pump system comprises a validated blood-mimicking fluid, a graduated cylinder, a pulse dampener, a pack of tubing 6/16″ diameter, and control cables. The pump was initially set up and then calibrated to ensure the dispensing of a specific volume of the Doppler fluid into the graduated cylinder. Air bubbles were removed by allowing the pump to circulate the liquid for about 30 min before connection with the phantom and subsequent velocity measurements. Conversion of flow rate to average flow velocity was achieved by dividing the flow rate by the cross-sectional area of the tube. The expected average flow velocities were given in a tabular form provided by the pump manufacturers. Readings were obtained in a constant velocity flow format at a flow rate of 180 cm/s, expected to generate a corresponding TAMMV of 191 cm/s., The constant velocity flow is equivalent to peak systolic velocity as shown here: $\text{TAMMV}=\frac{{\nu }_{\max }{\int }_{T_2}^{T_1}\mathrm{dt}}{T_2-T_1}={\nu }_{\max .}$

Given that ν_max is constant, ν_max = ν_peak.

However, in children with SCD, the ν_max (also referred to as the peak systolic velocity) is not equal to the TAMMV. The phantom has 4 constant-diameter channels of 2, 4, 6, and 8 mm. The pump was connected to the 2 mm-diameter channels, which is appropriate for the diameter of the middle cerebral artery. The pump comprises urethane rubber as tissue-mimicking material with channels parallel to the scan surface at 15 mm depth. To achieve an insonation angle of about 34°, an asymmetric wedge of 50 mm maximum thickness and similar tissue-mimicking material is placed on the surface of the phantom.

We evaluated four commercial non-imaging TCD machines from 3 companies and one imaging TCD machine. The non-imaging TCD systems were two models from CareFusion Sonara/tek San Diego, USA (an originally purchased machine for the SPIN Trial in 2013 and a refurbished machine in 2020), one non-imaging machine from Lucid M1 Neuralanalytic (renamed as Neural Analytics Rebrands Itself NovaSignal (globenewswire.com)), Los Angeles, USA, and one non-imaging TCD machine from DWL multi-dop T digital system, Germany. The one imaging TCD system was a Nortek DC5-500, China.

Each machine was equipped with a 2 Mhz transducer in optimum condition, generating high-quality spectra. Coupling gel was applied to the surface of the transducer to displace air between the surfaces of the transducer and the phantom. The transducer was then placed perpendicular to the phantom wedge's short side along the phantom's 2 mm channel side, and readings were obtained. The measurement depth was adjusted until the best signal was obtained using the color of the spectrum, the loudness, and the quality of the spectrum obtained. Measurements were repeated 20 times on the same day for each machine under the same conditions: sampling depth of 50 mm, 180 cm/s, and angle of correction of 34°. All readings were split equally between two experienced radiologists who have performed TCD for at least 10 years. The TAMMV and TAMV velocities for each machine was summarized with the mean, standard deviation, and range (minimum, maximum). A one-sample t-test was used to compare the values from each machine to the target value using a two-sided p-value, including the 95% confidence interval of the difference from the target value. Following previous work, 15 cm/s was added to the imaging system data for comparability to the target value as used in an international clinical trial in children with SCA who had both imaging and non-imaging TCD velocities [7]. The trial excluded children with imaging and non-imaging TCD velocities of > 185 cm/s and > 200 cm/s, respectively.

With the Doppler flow pump set at a flow rate of 180 mls/min, the reference TAMMV was 191 cm/s. The mean and standard deviations of the readings are shown in Table 1. Each machine's standard deviation and range are small (high precision), demonstrating the operators' consistency in their technique. However, each machine's accuracy (the difference between the mean TCD velocity and phantom Doppler velocity of 191 cm/s) varied considerably (Table 2). All the machines read lower than the reference TCD velocity. The one-sample t-test comparison of the machine means to the target value of 191 cm/s showed all to be different (p < 0.001). The DWL machine performed the closest to the reference velocity with a mean velocity of 2 cm/s below the reference value. The two Sonara Tek non-imaging TCD machines (the originally purchased machine in 2013 and the refurbished machine purchased in 2021) provided similar mean results of 179 cm/s. The Neuroanalytic TCD machine had a mean of 169 cm/s, which was the most discrepant. The Nortek USS imaging TCD machine had a mean result of 184 cm/s after adding 15 cm/s to the TCD velocity to each reading based on the prior evidence that imaging TCD velocities were systemically underestimated compared to the non-imaging TCD velocity [4].

The relationship between TAMV for imaging and TAMMV for non-imaging TCD requires more direct assessment with definitive studies in children with SCD. Jones et al. suggested that imaging TCD TAMV readings are 10%–15% lower than TAMMV in non-imaging [4]. Others have suggested no adjustment should be made between the non-imaging and imaging TCD velocities in children with SCA [7]. We elected adding 15 cm/s to the imaging TCD velocity, which was still 7 cm/s below the reference velocity of 191 cm/s. The data from our study suggest that when the imaging TCD is performed, the clinical threshold for primary stroke prevention should not be 200 cm/s, but significantly lower with an adjustment specific to the imaging TCD machine used on location.

Our study of TCD measurements had inherent limitations. The first is the variation of TCD machines (imaging versus non-imaging). We did not test an exhaustive list of TCD machines. Thus, our results should be only considered within the context of the five machines and four manufacturers in this study. The second limitation is the intra-observer difference, which is as high as 12% and may have contributed to the variation in our study [6]; however, only two experienced radiologists performed all the evaluations with each one performing 10 assessments per machine. We believe having two radiologists perform the evaluations decreased but did not eliminate intra-observer differences.

When comparing TCD machines to a reference TCD velocity of 191 cm/s, our results indicate that pediatric radiologists and hematologists should be aware of the TCD machine variability and the influence of ultrasonographer variability that may alter the decision threshold for primary stroke prevention. In the northern Nigerian setting where the study was performed, the clinical team has two strategies to overcome the limitations of non-imaging TCD machine's velocity variations. The first strategy is to lower the threshold from ≥ 200 to ≥ 180 cm/s for all non-imaging machines to begin hydroxyurea for primary stroke prevention. After discussion with the pediatricians, there was a strong preference for a single threshold for all non-imaging machines for decision-making. The second strategy requires two ultrasonographers to agree on an abnormal velocity classification for any velocity ≥ 180 and < 220 cm/s before initiating hydroxyurea for primary stroke prevention. This decision to lower the decision threshold to 180 cm/s for all non-imaging machines is also because the Kano government has agreed to pay for hydroxyurea for primary stroke prevention. Additionally, hydroxyurea has proven benefits beyond primary stroke prevention, including a decrease in the incidence of pain, acute chest syndrome, blood transfusion therapy, and malaria [8, 9]. Ideally, all children with SCA would be treated, but this is beyond the financial resources of the local government and many families who pay out of pocket. For medical professionals implementing primary stroke prevention programs in their settings, we recommend comparing the measured velocity of their institution's TCD machines to a reference phantom Doppler of at least 180 cm/s to ensure the accuracy and precision of their specific TCD machines in their institution.

No ethical committee was involved in approving the study because the study did not involve any human participants.

TL is an employee of Sun Nuclear Corporation. MRD participated as a chair of the Novartis steering committee. This activity is not directly related to the present work.