Power Transcranial Doppler Ultrasound in the Detection of Intracranial Aneurysms
Background and Purpose—We sought to perform a large, prospective, multicenter, blinded study comparing power transcranial color duplex sonography (power TCDS) with intra-arterial digital subtraction angiography (IADSA) in the detection of intracranial aneurysms.
Methods—Contemporaneous TCDS and IADSA examinations were performed in 171 subjects with suspected intracranial aneurysm. Via the temporal bone window, a 2-dimensional hand-held noncontrast transcranial duplex ultrasound imaging system was used operating in power and spectral modes. Sonographers were blinded to clinical history and results of brain CT and IADSA.
Results—We found that 157 subjects (92%) had an adequate bone window. Sensitivity per patient was 0.78 (95% CI, 0.66 to 0.87) and 0.46 (95% CI, 0.36 to 0.56) for any anterior circulation aneurysms. Sensitivity was 0.35 (95% CI, 0.24 to 0.46) for aneurysms ≤5 mm and 0.81 (95% CI, 0.62 to 0.94) for aneurysms >5 mm. Accuracy was lower for aneurysms on the cavernous and terminal internal carotid arteries, including posterior communicating artery origin (0.71; 95% CI, 0.63 to 0.79), than for those on the anterior (0.82; 95% CI, 0.74 to 0.89) or the middle cerebral arteries (0.79; 95% CI, 0.71 to 0.86).
Conclusions—Power TCDS is a promising, inexpensive, noninvasive test for anterior circulation intracranial aneurysms but is less sensitive per aneurysm than alternatives such as CT angiography or MR angiography. Sensitivity is poor for aneurysms ≤5 mm in diameter. The internal carotid artery is the most difficult segment to interpret.
The accepted reference standard for the detection of intracranial aneurysms is intra-arterial digital subtraction angiography (IADSA).1 2 3 This has a permanent neurological complication rate in patients investigated for a suspected aneurysm of 0.07%4 and is invasive, time consuming, and relatively expensive. As a consequence, there has been considerable interest in whether noninvasive imaging methods can detect intracranial aneurysms reliably in both symptomatic and asymptomatic patients.5 Numerous studies have compared noninvasive techniques such as MR angiography (MRA) or CT angiography (CTA) with IADSA and found similar overall accuracy of approximately 90%, but there are relatively few data from prospective, blinded studies comparing transcranial color duplex sonography (TCDS) with IADSA.6
TCDS became available in the early 1990s, with successful identification of intracranial aneurysms soon reported.7 8 Ultrasound has the advantage of lower capital cost and greater mobility than IADSA, CTA, or MRA; additionally, there are no contraindications and no exposure to ionizing radiation. A more recent technological development of color Doppler termed color Doppler energy or “power Doppler” offers significantly greater sensitivity to flowing blood than standard color flow imaging.9 Power Doppler can be readily combined with a spectral ultrasound examination to obtain additional velocity and waveform information. With this technique, overall sensitivity for detection of aneurysms between 0.4710 and 0.8911 and specificity between 0.3312 and 0.899 have been reported in relatively small studies, predominantly in patients with recent subarachnoid hemorrhage (SAH). Unfortunately, up to 10% of patients will not have an adequate bone window, and TCDS is operator dependent. To define the accuracy and limitations of power TCDS in a broader range of patients, we undertook a large, prospective, multicenter, blinded study comparing power TCDS with IADSA in the detection of intracranial aneurysms. Many of these subjects were also examined with MRA and CTA, and these results have been reported separately.13
Subjects and Methods
Any patient undergoing cerebral angiography for the detection of an intracranial aneurysm in either of the 2 participating regional neuroscience centers (total catchment population, 4 million) during the 18-month period of the study (November 1997 to April 1999) was eligible. Therefore, asymptomatic patients undergoing angiography to exclude an aneurysm because of a strong family history of aneurysmal SAH or patients with symptoms suggestive of an aneurysm (eg, late presentation after severe sudden occipital headache or a third nerve palsy) were eligible as well as patients with proven recent SAH. Exclusion criteria were patients with a poor grade (World Federation of Neurosurgeons [WFNS] grade 3 or worse) of subarachnoid hemorrhage (because ethically obtaining informed patient consent was not possible), age <18 or >75 years, or pregnancy. Approval for the study was obtained from the appropriate hospital ethics committees, and written informed consent was obtained from participants.
Neuroradiologists performed all the IADSA examinations using Advantx DX angiographic equipment (IGE Medical Systems). IADSA studies were 3- or 4-vessel selective angiograms with multiple projections obtained for each vessel. Power and angle-corrected spectral TCDS examinations were performed contemporaneously on Acuson 128XP machines with the use of 2- to 2.5-MHz multihertz linear phased array transducers (Acuson). Identical imaging parameters were used in each center. A more detailed description of the TCDS scanning technique, optimization parameters, and interpretation criteria is provided in Appendix 2 (which can be found, with Appendix 1, at http://stroke.ahajournals.org).
TCDS examinations were performed via the temporal bone window to insonate the circle of Willis in the axial and coronal planes.7 9 11 The transnuchal and transorbital routes were not routinely used (some of the patients found these difficult to tolerate), and intravenous echo contrast was not used. Each major intracranial vessel segment was examined systematically with power and angle-corrected spectral Doppler ultrasound. We did not use nonpower color-coded Doppler ultrasound. Aneurysm size was determined on a frozen image with the use of electronic calipers in the standard Acuson measurement software. A video record of each examination was made, and a standard result pro forma sheet was completed at the end of each examination by the sonographer. This recorded the opinion of normal or abnormal (abnormality specified) in each major artery/branching point, including the sonographer’s degree of confidence.
Ultrasonographers comprised 2 neuroradiologists and 3 neuroradiographers. Two sonographers had several years of experience in using power TCDS as well as spectral transcranial ultrasound. Three sonographers were experienced in using spectral transcranial ultrasound in the detection of vasospasm but less experienced in power TCDS and had undergone 3 half-day training sessions in its use to detect aneurysms (as described above) before the start of the study. Sonographers were fully blinded to clinical data and the results of all other imaging investigations, including plain CT and IADSA results.
IADSA images were presented as anonymous, randomly numbered studies with no clinical details or results of other imaging for independent review by 2 consultant neuroradiologists. Disagreements were resolved by consensus review. For TCD, the report completed at the end of each examination was used for the comparison with IADSA. Aneurysm site(s) were recorded as follows: 1=middle cerebral artery (MCA) main stem; 2=MCA bifurcation; 3=distal MCA; 4=anterior communicating artery complex; 5=pericallosal segment; 6=terminal internal carotid artery (ICA) segment; 7=posterior communicating artery (PCoA); 8=ophthalmic artery; 9=other ICA; 10=basilar artery; 11=posterior inferior cerebellar artery; and 12=other (or unspecified). Aneurysm size was recorded as follows: (1) <3 mm maximum angiographic dimension, (2) 3 to 5 mm, (3) 5.1 to 10 mm, and (4) >10 mm. The confidence of the ultrasonographer in the report was assessed on a simple 5-point scale as reported by Atlas et al14 : 5=aneurysm definitely absent, 4=aneurysm probably absent, 3=uncertain, 2=aneurysm probably present, and 1=aneurysm definitely present. An assessment was also made of the visibility of the major arterial segments comprising the circle of Willis. We aimed to have 2 ultrasonographers examine a reasonable proportion of the same patients to assess interobserver variability. This proved difficult to achieve in practice in a 2-center study in which patients were frequently treated or discharged before a second sonographer was available to perform the power TCDS study; sometimes, even if a second sonographer was available, the second sonographer was not blinded to the patient’s CT/IADSA findings.
We constructed 2×2 tables of true positives, false-positives, false-negatives, and true negatives compared with IADSA. Sensitivity, specificity, positive and negative predictive values, accuracy, and likelihood ratios were calculated and compared on a per patient and per aneurysm basis. Exact 95% CIs based on binomial probabilities were calculated.15 The unweighted κ statistic was used to assess the level of interobserver agreement for a small number of cases in which it was possible to have power TCDS performed by 2 operators.16 CIs for the difference between 2 proportions were constructed to show whether there was a difference between the proportions interpreted correctly at different levels of observer confidence. Sensitivity analyses were used to examine the effect of aneurysm size and site, clinical presentation, and observer experience on the accuracy of TCD.
Two hundred patients meeting the inclusion criteria and who agreed to participate were recruited prospectively over an 18-month period. During the period of the study, 520 patients were admitted to the 2 centers with suspected SAH, of whom almost two thirds (ie, approximately 345) were eligible for the study, ie, were aged 18 to 75 years, not pregnant, and were classified with a good WFNS grade. Major reasons for the noninclusion of the other 155 patients in the study included inadequate time before aneurysm treatment (which was often performed within 24 hours of presentation) and patient refusal. Twenty-nine (15%) of the 200 patients recruited did not undergo power TCDS because of lack of availability of a blinded trained ultrasonographer before treatment or discharge but did undergo CTA and/or MRA (these results are reported separately13 ). The final total was therefore 171 patients who underwent contemporaneous TCDS and IADSA (84 men, 87 women) with a median age of 43 years (range, 19 to 71 years). Fourteen of the 171 patients (8%) (9 men and 5 women) had an inadequate bone window and therefore were excluded from further analysis. Patients were grouped into 3 categories on the basis of the clinical indication for cerebral angiography. Group 1 comprised 72 patients with a known aneurysm undergoing further assessment (n=5) or patients with proven SAH (n=67); group 2, 64 patients with symptoms that might be due to an aneurysm (eg, painful oculomotor nerve palsy); and group 3, 21 asymptomatic patients at risk of harboring an aneurysm because of a strong family history of aneurysmal SAH.
On IADSA, 122 aneurysms were present in 43% of patients (67/157); these findings are detailed with a breakdown of the site and size of aneurysms, plus the corresponding TCDS result, in Appendix 2. Details of the false-negative and -positive results for TCD are also included. Twenty-two (18%) were aneurysms of the anterior cerebral artery (ACA) circulation, 32 (26%) of the MCA, 51 (42%) of the ICA (including PCoA aneurysms), and 17 (14%) of the vertebrobasilar system. These latter would not be expected to be detected by the power TCDS protocol used in the study since aneurysms in the posterior fossa distant from the circle of Willis would not be visualized. Twenty-seven (25.7%) of the anterior circulation aneurysms were very small (<3 mm maximum angiographic dimension), 51 (48.6%) were small (3 to 5 mm), and 27 (25.7%) were >5 mm. Of the 17 vertebrobasilar aneurysms in the patients, 8 were 3 to 5 mm and 9 were >5 mm. Anterior circulation aneurysm prevalence was 60% (43/72) in patient group 1, 17% (11/64) in group 2, and 38% (8/21) in group 3.
The overall comparative diagnostic performance of TCDS to IADSA is given (with 95% CIs) in Table 1⇓. The accuracy on a per patient basis (the ability to correctly discriminate a patient as true positive or true negative for possession of an intracranial aneurysm) of 0.85 was much better than the accuracy on a per aneurysm basis (the ability to detect correctly the precise site of all aneurysms in each patient) of 0.60. This difference was largely due to sonographers failing to identify a second (often smaller) aneurysm in patients in whom they had already identified an aneurysm (Appendix 2).
The 2 sonographers more experienced in power TCD performed 29% (45/157) of the examinations. There was a trend for the more experienced sonographers to be more accurate than the less experienced sonographers: accuracy per patient 0.89 (95% CI, 0.76 to 0.96) versus 0.83 (0.75 to 0.89) and per aneurysm 0.62 (95% CI, 0.49 to 0.74) versus 0.59 (95% CI, 0.51 to 0.67), respectively. The differences were quite small and not statistically significant (P>0.05). However, this was not a comparison of like entities because it was not possible for logistical reasons for each patient to be examined by both an experienced and a less experienced sonographer. In a small number of cases (n=12), 2 blinded, less experienced sonographers were able to independently perform TCDS examinations. Interobserver agreement in this small sample was good, with a κ statistic of 0.76, but the sample was too small to be a true test of observer variability.
Analyses of the diagnostic performance by size and site of aneurysm are given in Tables 2⇓ and 3⇓. Sensitivity was substantially better for aneurysms >5 mm, at 0.81 (22/27), than for aneurysms ≤5 mm in size, at 0.35 (27/78). This magnitude of difference was similar for both the experienced and less experienced sonographers. Overall, within the anterior circulation, accuracy was poorer for ICA complex aneurysms because of poorer sensitivity and slightly poorer specificity than for the ACA and MCA complexes (Table 3⇓). The poorer sensitivity for ICA complex aneurysms was concentrated in the less experienced sonographers: 14 of 38 aneurysms detected compared with 8 of 13 for the more experienced sonographers. Whereas for the more experienced observers the MCA complex had the poorest sensitivity (3 of 11 aneurysms detected compared with 12 of 21 for the less experienced sonographers), sensitivity, specificity, and accuracy for the ACA complex were very similar for all sonographers.
Diagnostic performance was also analyzed by clinical group, and the results are presented in Table 4⇓. Similar accuracy was found for the 3 clinical groups on a per patient basis, although the CIs are wide because of the smaller numbers engendered by a subgroup analysis. On a per aneurysm basis, TCDS had poorer specificity in the subjects in group 1 (known aneurysm or proven SAH) and group 3 (positive family history) than for subjects in group 2 (Table 4⇓).
TCDS performance was related to the level of ultrasonographer confidence for the less experienced observers; this was available in 106 of 112 examinations. In cases in which the sonographer was confident (56 of 106 patients had a confidence score of 1 or 5), the accuracy per patient was 0.89 (50/56) compared with 0.75 (30/40) for moderate confidence (40 of 106 patients had a score of 2 or 4) and 0.70 (7/10) if confidence was low (10 of 106 patients had a score of 3).
This study is only the second large prospective study comparing power TCDS with IADSA, and it is also one of the largest prospective studies of any noninvasive imaging technology (eg, CTA, MRA, and TCDS) to date.6 17 In a systematic review of the literature on noninvasive imaging of intracranial aneurysms, it was clear that data on TCDS were very limited compared with CTA and MRA, reflecting the more mature state of those technologies. Few conclusions could be drawn about the diagnostic performance of TCDS from the data available before the present study.
Diagnostic Performance by Site and Size of Aneurysm
Unsurprisingly, we found diagnostic performance was much poorer in aneurysms ≤5 mm in size. For anterior circulation aneurysms >5 mm in size, sensitivity was 0.81, which is in agreement with the best of previous 2-dimensional noncontrast TCDS reports. Although large aneurysms are more readily detected by TCDS, small aneurysms can be detected in patients with a good bone window, as demonstrated in Figures 1⇓ and 2⇓, respectively. This size-related performance effect applies equally to CTA and MRA.6 13
It is reassuring that the differences in sensitivity and specificity between the less and more experienced TCDS sonographers were quite small (difference overall per patient: 9% for sensitivity, 3% for specificity; difference per aneurysm: 5% for sensitivity, 2% for specificity). These results indicate that diagnostic performance is not highly dependent on the experience of the sonographer and that satisfactory performance in the technique can be achieved without prolonged, extensive training if a sonographer is already competent in spectral transcranial Doppler ultrasound. Three half-day training sessions in power TCDS were provided for the less experienced sonographers in this study. Interobserver agreement (albeit on a small sample) was also good. The practical logistics of patient admission, investigation, and treatment meant that we were unable systematically to have 2 blinded sonographers examine subjects. In particular, it would have been valuable if an experienced and a less experienced sonographer could have done this, but this proved to be impractical within the available resources.
Because of the methodology of this study, in which only the temporal bone window was routinely insonated, it is not possible to comment on performance in the basilar and vertebral areas. Regarding the anterior circulation, overall accuracy was poorer for ICA complex aneurysms than the MCA or ACA complexes, largely because of the relatively greater number of false-negative readings for the less experienced sonographers (24 of 38 versus 5 of 13 for the more experienced sonographers; Table 3⇑). The ICA complex was also a difficult area for detection of aneurysms with CTA and MRA.6 13 It is interesting but difficult to explain why the experienced observers had a substantially lower detection rate for MCA aneurysms (0.27) than the less experienced observers (0.57), although the CIs overlap widely.
Overall accuracy between the different patient groups was broadly similar on a per patient basis, suggesting that the influence of patient type on the accuracy of TCDS was not great. It is also interesting that in cases in which the less experienced sonographers felt confident about the findings in an individual patient, their accuracy was identical to that of the more experienced sonographers (0.89).
Comparison With Previous Studies
The results on a per patient basis are encouraging, with a sensitivity of 0.78, but those on a per aneurysm basis are less so, with an overall sensitivity of 0.46. However, for any modality, the performance per aneurysm will always be poorer than that per patient. The sensitivity per aneurysm in previous noncontrast TCDS studies has ranged from 0.4018 in the only previous large prospective study (n=203) to 0.89 in a smaller study of 36 patients.11 Several other small studies have also reported sensitivity in the range of 0.53 to 0.89,7 9 12 19 although all were in populations with a high prevalence of aneurysms. The difference between sensitivity in the present study and that reported in the early small studies might be due to a number of factors. There might be fewer biases in the present large, prospective, fully blinded study, eg, expectation and recall bias; the mixed patient population with a lower aneurysm prevalence (43%) than most previous studies might also have had an effect on sensitivity,20 as well as the inclusion of results from sonographers less experienced in the technique (who performed 71% of the examinations). However, probably the most significant effect is the proportion of small and very small aneurysms in this study: 74% (78/105) of the aneurysms were ≤5 mm. In 2 previous studies reporting better sensitivity per aneurysm than the present study and in which aneurysm sizes were provided, 33%12 and 0%11 were ≤5 mm, and none of the small aneurysms were detected. In the only other large, blinded study of TCDS to date, the prevalence of aneurysms was 63%.18 In that study the combined sensitivity for contrast and noncontrast TCDS was 0.28 (24/87) for small aneurysms (<6 mm) and 0.53 (43/81) for aneurysms ≥6 mm (all aneurysm sites), compared with 0.35 (≤5 mm) and 0.81 (>5 mm) for anterior circulation aneurysms in the present study.
Comparison of TCDS With Other Noninvasive Modalities
In comparison with reported studies of CTA and MRA, TCDS appears to be inferior at aneurysm detection, particularly on a per aneurysm basis, although on a per patient basis the differences between techniques are small.6 However, these comparisons have not been of like entities, ie, of CTA, MRA, and TCD in the same patients (with the same aneurysms). A proportion of the patients (n=114) in the present study underwent IADSA, CTA, MRA, and TCDS contemporaneously. These results are reported separately.13 When we use the mean of 2 observers, the sensitivity per aneurysm for CTA was 0.56 (40/72) for aneurysms ≤5 mm and 0.94 (34/36) for aneurysms >5 mm. For MRA the sensitivity per aneurysm was 0.33 (24/72) for aneurysms ≤5 mm and 0.89 (32/36) for aneurysms >5 mm versus 0.35 and 0.81, respectively, for TCDS. However, overall specificity per aneurysm was better for MRA at 0.87. For CTA it was only slightly better at 0.76 compared with 0.72 for TCDS. Therefore, in the same cohort of patients CTA and MRA were slightly more accurate overall per aneurysm at 0.72 and 0.67, respectively, than TCDS at 0.60. However, there was no difference on a per patient basis, with an accuracy of 0.85 for all 3 modalities. Although, unlike the TCDS results, the CTA and MRA results included posterior circulation aneurysms, the sensitivity and specificity were similar overall for anterior and posterior circulation aneurysms for both CTA and MRA.13
TCDS has not reached the same advanced point of technological development or maturity as CTA or MRA, and these techniques have been in much more widespread use for longer than any form of TCDS. Therefore, these first results of TCDS in a cohort of patients with CTA and MRA available for direct comparison appear promising.
After the results of the International Study of Unruptured Intracranial Aneurysms were reported,21 the role for any screening for intracranial aneurysms has remained controversial, although it can be difficult not to investigate the worried individual with a strong family history or other risk factor(s). In a screening context, it could be argued that because only aneurysms >10 mm in diameter would be considered for elective treatment, poor sensitivity for smaller aneurysms would not be clinically important, although it might be important from a medicolegal perspective. Therefore, TCDS should not yet be ignored as a method of diagnosing aneurysms. Indeed, future prospects for TCDS are encouraging, with the development of real-time 3-dimensional scanning techniques and the potential to combine these with contrast enhancement; sensitivity per aneurysm in a small study was reported as 97%.22 However, 3-dimensional and contrast techniques would increase the cost, duration, and invasiveness of TCDS examinations. Only 1 large, prospective, blinded study of contrast-enhanced power TCDS has been published (105 subjects had power TCDS before and after echo contrast). This demonstrated a significant increase in sensitivity for all aneurysm sites from 0.40 to 0.55 with the use of contrast, although at the cost of a small reduction in specificity.18
In conclusion, power TCD is a promising, quick, safe, reproducible, and inexpensive noninvasive test for anterior circulation intracranial aneurysms. At present it is less accurate on a per aneurysm basis than CTA or MRA, although identical on a per patient basis. As with other noninvasive techniques, detection of small aneurysms is particularly poor. The cavernous carotid is the most difficult site to interpret, particularly for less experienced operators. The value of newer ultrasound techniques, and in particular the role of 3-dimensional ultrasound and ultrasonic contrast agents, requires further evaluation in large, prospective, fully blinded studies preferably comparing TCDS directly with the other alternative noninvasive modalities in the same patient cohort. The role of TCDS as an adjunctive test to CTA or MRA in the diagnosis of intracranial aneurysms (analogous to the situation in many centers in the diagnosis of carotid stenosis) also merits investigation.
This study was supported by the British Brain and Spine Foundation from the Davie Cooper Scottish Aneurysm Study grant, administered by the Department of Neurosurgery, University of Glasgow (to P.M.W. and V.E.), and by the Medical Research Council under the Clinical Research Initiative in Clinical Neuroscience (to J.M.W.).
- Received November 27, 2000.
- Revision received February 14, 2001.
- Accepted February 28, 2001.
- Copyright © 2001 by American Heart Association
Cloft HJ, Joseph GJ, Dion JE. Meta-analysis of risks of cerebral angiography in patients with subarachnoid hemorrhage, intracranial aneurysm, and arteriovenous malformation. Stroke. 1999;30:317–320.
Wardlaw JM, White PM. The detection and management of unruptured intracranial aneurysms. Brain. 2000;123:205–221.
Tsuchiya T, Yasaka M, Yamaguchi T, Kimura K, Omae T. Imaging of the basal cerebral arteries and measurement of blood velocity in adults using transcranial real-time color flow Doppler sonography. AJNR Am J Neuroradiol. 1991;12:497–502.
Leftheriotis G, Pulci S, Mercier P, Abraham P, de Bray JM, Samuet JL. Imagerie des anevrysmes et malformations vasculaires cerebrales en echo-Doppler couleur transcranienne. Journal d’Echographie et de Medicine par Ultrasons. 1992;13:127–133.
White PM, Teasdale EM, Wardlaw JM, Easton V. CTA and MRA in the detection of intracranial aneurysms: a prospective, blinded comparison in a large patient cohort. Radiology. In press.
Armitage P, Berry G. Statistical Methods in Medical Research. Oxford, UK: Blackwell Scientific Publications; 1987:117–119.
Altman D. Practical Statistics for Medical Research. London, UK: Chapman & Hall; 1991:402–410.
Turner CL, Kirkpatrick PJ. Detection of intracranial aneurysms with unenhanced and echo contrast enhanced transcranial power Doppler. J Neurol Neurosurg Psychiatry. 2000;68:489–495.
Bartels E. Transcranial color-coded duplex sonography (TCCS): cerebral arteries and parenchyma. In: Color-Coded Duplex Ultrasonography of the Cerebral Vessels. Stuttgart, Germany: Schattauer; 1999:187–279.
Klotzsch C, Bozzato A, Lammers G, Mull M, Lennartz B, Noth J. Three-dimensional transcranial color-coded sonography of cerebral aneurysms. Stroke. 1999;30:2285–2290.