Interpretation of Embolic Phenomena During Carotid Endarterectomy
Background and Purpose Air and particulate emboli are a major source of morbidity during carotid endarterectomy (CEA); however, amplitude overload and poor time resolution have restricted the ability of transcranial Doppler ultrasound to differentiate between the two.
Methods We have now overcome these two limitations by (1) rerouting embolic signals away from the audio frequency amplifier to avoid amplitude overload and (2) substituting the Wigner distribution function for the fast Fourier transform to improve time and frequency resolution. Thus, we can now accurately determine embolic duration and embolic velocity, the product of which is the sample volume length (SVL). This measurement represents the physical distance over which an embolic signal can be detected. The underlying hypothesis was that air reflected more ultrasound and would therefore be detected over a greater SVL.
Results The median SVL (interquartile range) for 75 in vitro air emboli was 1.97 cm (range, 1.70 to 2.35) compared with 0.27 cm (range, 0.16 to 0.43) for 185 particulate emboli detected during the dissection phase of CEA. Off-line analysis on an additional 560 embolic signals detected during different phases of CEA suggested that 46 of 143 (32%) of emboli immediately after shunt insertion were particulate, as were 19 of 33 (58%) occurring during shunting, 28 of 78 (36%) after restoration of flow in the external carotid artery, 23 of 251 (9%) after restoration of flow in the internal carotid artery, and 55 of 55 (100%) of those emboli detected during the early recovery phase.
Conclusions This development provides objective physical criteria upon which embolus characterization (particulate/air) can be based. This could have major implications for future patient monitoring with respect to modification of surgical technique and pharmacological intervention.
Cerebral emboli occurring during CEA can easily be detected with the use of TCD, now the accepted method for diagnosing intracranial emboli.1 Until recently the clinical significance of such emboli was unknown since it had been impossible to conclusively differentiate between insignificant air emboli and potentially damaging particulate emboli. There is evidence, however, that persistent particulate embolization during the dissection phase of CEA may be associated with a significant decline in cognitive function,2 while emboli detected during the early recovery phase are associated with an increased risk of perioperative thrombosis and cerebral infarction.2 Air emboli, however, rarely cause significant morbidity.2
To date, amplitude overload and poor time resolution have consistently limited the ability of current TCD technology to differentiate between air and particulate emboli, while sizing still remains a distant objective. Early reports noted that embolic signals from air bubbles were characteristically of high amplitude and short duration and appeared in both the forward and reverse flow channels.3 4 These features resulted from the highly reflective nature of air giving rise to reflected signals of 40 to 50 dB above the background Doppler blood signal. This is well above the dynamic range of most TCD systems (20 to 40 dB), and the effect of amplitude overload (Fig 1⇓, top panel) needs to be eliminated before any sensible interpretations of embolic signals can be attempted.5
The second problem concerns the limited temporal resolution of FFT analyzers that are used in all commercial TCD machines. Embolic signals are very transient in nature, often only 10 milliseconds in duration. To provide an adequate time resolution of at least 1 millisecond to accurately measure embolic duration, the frequency resolution would have to be reduced to 1 kHz. The latter becomes impractical, however, because the normal Doppler-shifted frequencies from the basal cerebral arteries are of the order of 1 kHz. Thus, FFT analysis yields itself almost redundant when required to analyze the specific properties of emboli. To improve this situation, an alternative method of analysis must be found. We have overcome these two constraints5 with one hardware change (to improve the dynamic range) and one software change (substituting Wigner analysis for FFT to improve temporal resolution6 ). The new system was then applied in a series of in vitro and in vivo studies aimed at differentiating air from particulate microemboli.
Materials and Methods
The problem of an insufficient dynamic range results in many air emboli producing signals with a high-velocity bidirectional signature; this is simply due to overload within the TCD unit. The dynamic range of our TCD system is 20 dB. This is more than adequate to measure cerebral blood velocities but not to analyze embolic signals. To effectively increase the dynamic range to 60 dB, we performed a hardware modification to the Doppler unit so that embolic signals no longer overload the system.5 In summary, forward and reverse flows are conventionally displayed on two channels via the audio outputs. However, because flow in the MCA is unidirectional, one of the audio outputs becomes redundant. Overload occurs as a result of the circuit amplifiers within the Doppler unit; thus, we used the redundant channel to display the normal forward flow signal before it reached the amplification stage and recorded this new output directly onto a DAT recorder.5 It is from this new attenuated channel that the correct measurements of embolic signals can be made (Fig 1⇑, bottom panel). Although no background Doppler blood signal can be detected in this new attenuated channel, there is a known 40-dB attenuation between the two channels, and thus we can measure the amplitude of the embolic signal relative to the blood signal in the nonattenuated channel.
TCD systems almost exclusively use a method of analysis known as FFT to convert the Doppler audio signal (time domain) into the standard time-frequency spectrum display (frequency domain). One intrinsic property of the FFT is that there is an inversely proportional relationship between time and frequency resolution. Thus, increasing the time resolution to cope with very short duration embolic signals decreases the frequency resolution to a level that yields no useful information on embolic velocity.
One alternative to FFT that does not possess this inverse relationship is the Wigner distribution function (Fig 2⇓, top panel). Although complex in its mathematics, it can simply be viewed as an alternative processing algorithm to FFT and allows a time resolution of 80 microseconds while retaining a frequency resolution of 48 Hz. This compares with only 10 milliseconds and 100 Hz, respectively, with the FFT method. Thus, we have effectively increased the time resolution 125-fold as well as improving the frequency resolution by a factor of 2.
Sample Volume Length
The effective SVL is simply the product of embolic duration and velocity and represents the length of artery over which the embolic signal can be detected. All embolic signals were analyzed off-line from DAT recordings. Once an embolus was located in real time, the temporal resolution was increased to its finest resolution mode (Fig 2⇑, bottom panel). Embolus velocity was obtained by measuring the peak value of the mean blood flow velocity. This is displayed as a red line in the bottom panel of Fig 2⇑, with a peak value easily apparent. The embolus duration was defined as the period of time that the amplitude of the backscattered signal was 10 dB or greater than the background blood signal. We hypothesized that because air reflects a large proportion of the incident ultrasound,7 the backscattered signal should be detected over a longer sample volume than particulate emboli, which are not as reflective5 and therefore only detected around the center of the sample volume, where sensitivity is at a maximum. The use of Wigner analysis instead of FFT analysis enables the SVL to be calculated with up to 250 times greater accuracy.
In Vitro and In Vivo Studies
It had been intended to undertake validatory studies using an in vitro flow rig model into which controlled volumes of either air or particulate debris could be introduced. However, while it was possible to inject micro air bubbles into the flow rig with ease, in common with other investigators an injection of particulate emboli inevitably introduced micro air bubbles, thereby confounding meaningful interpretation of data.8 A pure source of particulate emboli was therefore obtained during the dissection phase of CEA when no air had entered the arterial system. Conversely, a pure source of air emboli is not obtainable in vivo since it is impossible to exclude particulate debris breaking off the carotid artery. Thus, air was acquired in vitro and particulate debris in vivo.
To obtain in vitro data, a flow rig was constructed and pulsatile flow was simulated with the use of a peristaltic pump. Silicon tubing was used except at the insonation site, where a 30-cm length of 3-mm-ID heat-shrunk tubing was inserted to provide a more realistic model of arterial wall. A synthetic blood analogue9 was used with the same backscatter properties as blood. Both fluids gave the same background Doppler signal from the test section of the rig to within 1 dB. The density, viscosity, size of scatterers, and “hematocrit” of the blood analogue were equal to that of whole blood. Discrete volumes of air (1 to 5 μL) emboli were introduced through a rubber-sealed side port at a distance of 20 cm proximal to the insonation site. This distance is similar to that between the carotid bifurcation and the MCA. Seventy-five air emboli were injected into the system and the backscattered Doppler signal recorded onto a DAT recorder for off-line analysis.
In vivo particulate data were acquired during routine CEA surgery. This procedure was performed in the standard manner with the use of normotensive, normocarbic general anesthesia, systemic heparinization (5000 IU IV), and carotid sinus nerve blockade (1 mL 1% lidocaine). Sixty-seven consecutive patients underwent continuous intraoperative monitoring of the MCA blood flow velocity during CEA. Sixteen (24%) patients had TCD evidence of one or more emboli during the initial dissection phase of CEA before cross-clamping. In total, 185 embolic signals were detected. In addition, a further 560 embolic signals were detected in these 16 patients after shunt insertion (n=143), during shunting (n=33), after restoration of flow in the external carotid artery (n=78), after restoration of flow in the internal carotid artery (n=251), and in the recovery phase (n=55). The latter refers to the first 3 hours after restoration of flow. We calculated the SVL of all 635 emboli using Wigner analysis.
Our data were not normally distributed, and accordingly all the results in the text and figures refer to median values and their interquartile ranges. Nonparametric (Mann-Whitney) tests were used. Significance was assumed at a value of P<.05.
The median SVL (interquartile range) for 75 air emboli introduced into the flow rig was 1.97 cm (range, 1.7 to 2.35) but was only 0.27 cm (range, 0.16 to 0.43) for 185 particulate emboli detected during dissection (Fig 3⇓). Assuming an upper range SVL limit of 1.28 cm for the diagnosis of particulate emboli, this would mean that 97.6% of the air emboli would have been correctly interpreted by the new method, ie, two air emboli would have been incorrectly identified as being particulate (Fig 3⇓).
For the purpose of analyzing the remaining 560 emboli, we therefore assumed that they were particulate if the SVL was 1.28 cm or less and gaseous if the SVL was greater than 1.28 cm. Fig 4⇓ details the calculated SVLs for the remaining phases of the operation. As can be seen, there was a wide distribution of SVLs for each operative phase, but certain important factors became apparent. Immediately after shunt insertion and restoration of flow, 46 of 143 emboli (32%) were probably particulate, while 97 (68%) were probably air. Interestingly, 19 of 33 emboli (58%) detected during shunting (ie, during the actual CEA) fell into the particulate category, as did 28 of 78 (36%) when flow was first restored in the external carotid artery. However, only 23 of 251 emboli (9%) after restoration of flow in the internal carotid artery were classified as particulate. Finally, in a small number of patients emboli continued into the early postoperative (recovery) period. This phase was defined as that period after final restoration of flow and after all manipulations to the carotid artery had ceased. Off-line analysis suggested that without exception, all (55 of 55) were particulate in nature.
In general, previous research into classifying emboli has used a direct measurement of either the backscattered signal from an embolus or the embolic velocity to predict embolic composition. Unfortunately, neither method has given rise to a distinct separation between air and particulate emboli because of either signal overload10 or poor time resolution.11
Our work represents an entirely new approach to analyzing embolic phenomena and has implications for both vascular and cardiac surgeons. We collected data from pure sources of both air and particulate emboli, namely, air in vitro and particulate in vivo, ensuring that all the ultrasonic properties of the blood substitute were analogous to whole blood. Although the attenuation of the ultrasound in the in vivo and in vitro studies (and indeed between patients) may be very different, the measurement of the period for which the embolic signal exceeds the level of the blood/blood substitute signal by 10 dB removes this potentially confounding factor. The angle of insonation in the in vitro study was 50°, yielding smaller sample volume dimensions than in vivo; this would tend to underestimate the SVLs for air, thereby decreasing the separation of the two groups. Even with this possible effect of underestimation of the SVL of air, there is still a clearly significant separation of the two types of emboli. This being so, we find that the signals obtained after initial shunt opening and final clamp release represent a mixture of air (the majority) and particulate (the minority). The unexpectedly high proportion of particulate emboli on shunt opening may be accounted for by the act of shunt insertion, which can in fact dislodge atheromatous debris.12
We have also shown that the SVL is not a constant because of the sensitivity distribution and the shape of the sample volume of the transducer.5 This has not been recognized previously. At the center of the sample volume, sensitivity is at a maximum, but there is no sharp cutoff in sensitivity. Instead, there is a gradual decrease in sensitivity to far beyond the stated value (usually 1 to 1.5 cm) with any pulsed Doppler system. As a result, very strong signals, such as those from air bubbles, are detected even in the very-low-sensitivity regions of the sample volume, leading to SVL values greater than that stated by the manufacturers. Weaker signals, such as those from particulate emboli, yield smaller SVLs. This is in contrast to other investigators who have stated or simply assumed that the SVL remained constant for all embolic signals, which can lead to misinterpretation of data.11 This phenomenon has only become so apparent because of the increase in accuracy of measuring duration and velocity of embolic signals with the Wigner analyzer integrated into our system.
In this study the threshold SVL to distinguish air from particulate emboli was taken as 1.28 cm. A very small percentage (3%) of air emboli would have been classified as particulate under this assumption; however, clinically it is preferable to overestimate rather than underestimate the number of particulate emboli.
The Wigner method is simple to use, with only two measurements needed to interpret embolic signals. The potential to differentiate between air and particulate embolization has important implications with regard to the pathophysiology of cognitive impairment after carotid surgery and cerebral angiography and could have an important role in the assessment of cerebral cytoprotective agents. More importantly, the differentiation between air and particulate embolization is the first step in developing a system capable of sizing emboli. We propose using Wigner distribution function instead of FFT to analyze Doppler signals to improve the accuracy of measuring embolic velocity and duration. This gives us the potential to investigate the clinical significance of all intraoperative emboli occurring during all procedures on patients with cerebral vascular disease.
Selected Abbreviations and Acronyms
|FFT||=||fast Fourier transform|
|MCA||=||middle cerebral artery|
|SVL||=||sample volume length|
|TCD||=||transcranial Doppler ultrasound|
This study was supported by the UK Stroke Association. J.L. Smith is a clinical research associate funded solely by the Stroke Association.
- Received July 21, 1995.
- Revision received September 11, 1995.
- Accepted September 14, 1995.
- Copyright © 1995 by American Heart Association
Spencer MP, Thomas GI, Nicholls SC, Sauvage LR. Detection of middle cerebral artery emboli during carotid endarterectomy using transcranial Doppler ultrasonography. Stroke. 1990;21:415-423.
Gaunt ME, Martin PJ, Smith JL, Rimmer T, Cherryman G, Ratliff DA, Bell PRF, Naylor AR. Clinical relevance of intra-operative embolisation detected by transcranial Doppler ultrasonography during carotid endarterectomy: a prospective study of 100 patients. Br J Surg. 1994;10:1435-1439.
Albin MS, Bunegin L, Garcia C, McKay W. The transcranial Doppler can image micro-aggregates of intracranial air and particulate matter. J Neurosurg Anesthesiol. 1989;1:134-135.
Siebler M, Sitzer M, Steinmetz H. Detection of intracranial emboli in patients with symptomatic extracranial carotid artery disease. Stroke. 1992;23:1652-1654.
Evans DH, McDicken WN, Skidmore R, Woodcock JP. Summary of the basic physics of Doppler ultrasound. In: Evans DH, McDicken WN, Skidmore R, Woodcock JP, eds. Doppler Ultrasound: Physics, Instrumentation, and Clinical Applications. Colchester, England: John Wiley & Sons; 1989:29-46.
Markus HS, Brown MM. Differentiation between different pathological cerebral embolic materials using transcranial Doppler in an in-vitro model. Stroke. 1993;24:1-5.
Oates CP. Towards an ideal blood analogue for Doppler phantoms. Phys Med Biol. 1991;11:1433-1442.
Russell D. The detection of cerebral emboli using Doppler ultrasound. In: Newell DW, Aaslid R, eds. Transcranial Doppler. New York, NY: Raven Press Publishers; 1992:207-213.
Droste DW, Markus HS, Nassiri D, Brown MM. The effect of velocity on the appearance of embolic signals studied in transcranial Doppler models. Stroke. 1994;25:986-991.