1559
Differentiation Between Gaseous and Formed
Embolic Materials In Vivo
Application in Prosthetic Heart Valve Patients
D. Georgiadis, MD; T.G. Mackay, PhD; A.W. Kelman, PhD; D.G. Grosset, MD;
D.J. Wheatley, MD; K.R. Lees, MD
Downloaded from http://stroke.ahajournals.org/ by guest on June 18, 2017
Background and Purpose Doppler emboli detection is an
established technique, but the nature of the underlying embolic material remains unclear. The intensity and spectral
distribution of emboli signals could help to distinguish between signals arising from formed and gaseous emboli. We
undertook this study to develop and evaluate a differentiation
algorithm based on the spectral characteristics of emboli
signals. Subsequently the algorithm was applied to patients
with mechanical prosthetic cardiac valves.
Methods Emboli signals detected in patients with carotid
disease, acute stroke, and atrial fibrillation were used as
formed emboli data, and signals detected in patients undergoing cardiac catheterization studies were used as gaseous
emboli data. For each embolus signal, the maximal amplitude,
the sum of amplitudes, and the spectral intensity distribution
were calculated. Two hundred emboli signals from each category were used to develop a differentiation algorithm, which
was subsequently evaluated on 501 additional solid and 995
gaseous emboli signals. The same algorithm was used to
analyze 5958 emboli signals detected in 60 patients with
mechanical prosthetic valves.
Results The best results were obtained with an algorithm
based on both the maximal amplitude and the sum of amplitudes (sensitivity, 99%; specificity, 96.5%). On subsequent
evaluation, the sensitivity and specificity of the algorithm were
99.6% and 89.8%, respectively. Of the 5958 emboli signals
detected in prosthetic valve patients, 92.4% were classified as
gaseous.
Conclusions Differentiation between gaseous and formed
emboli signals, as detected by transcranial Doppler in vivo, is
feasible by means of spectral analysis. Application of the
differentiation algorithm in prosthetic valve patients suggests
that the embolic material in these patients is gaseous. The
possibility of distinguishing between different formed embolic
materials with this technique remains to be evaluated. (Stroke.
1994^5:1559-1563.)
Key Words • embolism • heart valve prosthesis •
ultrasonics
E
emboli signals from patients with carotid disease, acute
stroke, and atrial fibrillation, which are presumed to be
caused by solid emboli, were analyzed and compared
with emboli signals from patients undergoing cardiac
catheterization studies, caused primarily by gaseous
emboli. Subsequently, the differentiation technique was
applied to emboli signals from patients with mechanical
prosthetic cardiac valves in an attempt to identify the
embolic material in these patients.
mbolism is a major cause of stroke. Current
therapeutic approaches depend on the embolic
source. However, as Caplan1 pointed out in a
recent review, the nature of the embolic material is
more important, particularly in deciding the appropriate antihemostatic treatment.
Transcranial Doppler ultrasonography (TCD) can be
used to detect intra-arterial emboli in several patient
groups. 26 The reflected signal amplitude depends on
both the size and density of the scattering particles. 78
Several in vitro 910 and animal7-81112 studies have used
amplitude of reflected Doppler signal in an attempt to
distinguish between embolic materials. A recent study
suggested that emboli signals arising from patients with
carotid stenosis and patients with prosthetic valves can
be differentiated on the basis of power of reflected
Doppler signal.13
This study was undertaken to develop and evaluate an
algorithm to differentiate formed and gaseous emboli
signals as detected by TCD in vivo. For this purpose
Received January 20, 1994; final revision received March 22,
1994; accepted March 24, 1994.
From the University Departments of Medicine and Therapeutics (D.G., A.W.K., D.G.G., K.R.L.) and Cardiac Surgery
(T.G.M., DJ.W.), University of Glasgow (United Kingdom).
Correspondence to Dr D. Georgiadis, Department of Neurology, University of Munster, Albert-Schweitzer-Str 33, 48129 Munster, FRG.
O 1994 American Heart Association, Inc.
Subjects and Methods
All emboli signals were detected during TCD monitoring
with a pulsed ultrasound Doppler machine (EME TC-2000)
with a 2-MHz probe. Monitoring was performed over the
middle cerebral artery (MCA), using an elasticated band to
stabilize the probe. The monitoring time varied between 0.5
and 3 hours per patient. The machine settings were kept
constant in all studied patients (power, 58%; sweep speed, 6
seconds; depth of insonation, 46 to 56 mm; sample volume, 9
mm). After identifying the MCA and adjusting the probe, the
gain was reduced to produce a homogeneous background
signal, with an intensity of 4 dB. Standard criteria were used to
differentiate between emboli signals and artifact both during
the recording14 and in off-line analysis15 of digital data stored
to hard disk, as defined in detail elsewhere.3
The following patients were examined. Group 1 included 58
patients (age, 28 to 82 years; median, 66 years) with carotid
disease, which was unilateral in 36 and bilateral in 22 cases.
Eight patients had a unilateral carotid occlusion and a contralateral severe (>70%, n=5) or moderate (50% to 70%, n=3)
stenosis. Of the remaining 50 patients, 38 had a severe (>70%)
1560
Stroke
Vol 25, No 8 August 1994
Downloaded from http://stroke.ahajournals.org/ by guest on June 18, 2017
stenosis, 11 a moderate (30% to 70%) stenosis, and 1 a
non-hemodynamicaJly significant plaque. In 24 patients additional cardiac abnormalities were diagnosed (atriaJ fibrillation
[n=14], valvular lesion [n=5], and dilated cardiomyopathy
[n=l]). Fifty-two patients were receiving aspirin (75 to 300
mg), 1 warfarin, and 5 were on no antihemostatic treatment.
All subjects were outpatients at the time of examination.
Monitoring time in these patients was 30 minutes over each
MCA. Group 2 included 20 patients with atrial fibrillation
(age, 61 to 70 years; median, 66 years) who were recruited
from the anticoagulant clinic and were stabilized on warfarin
at the time of study (international normalized ratio, 2.0 to 3.0).
Monitoring time for these patients was 30 minutes. Group 3
included patients presenting with acute ischemic stroke
(n=33) or transient ischemic attack (n=8) (age, 32 to 76 years;
median, 67 years). All of these patients were hospitalized at
the time of TCD study. None of them had an arterial line, but
21 had a venous line. No intravenous injections were performed during TCD monitoring, which was performed for 30
minutes over each MCA. Group 4 included 32 patients undergoing cardiac catheterization studies for ischemic heart disease
(n=30) or valvular lesions (n=2) (age, 32 to 68 years; median,
58 years). Only emboli signals that were associated with the
introduction of gas in the left heart or in the aorta (emboli
signals detected 5 to 20 seconds after contrast or saline
injection in the ventricle or supra-aortic region, during ventriculography, and during flushing of the catheter) were analyzed. Emboli signals detected during catheter insertion or
manipulation were not further evaluated in this study. Monitoring time in these patients varied between 20 and 80
minutes, depending on the duration of the procedure.
The screen display of the TCD consists of 512 spectral lines
in the* axis and 128 lines in they axis. The amplitude of each
of the 128 points along each vertical line (a,, a2, a3, . . . a128)
containing an embolus signal was calculated using software
developed in our department. The 128 amplitudes were then
compressed to 64 by addition of consecutive amplitude pairs
(a* 1 =a 1 +a 2 /2, a* 2 =a 3 +a4/2,... a*M=a127+a128/2). For emboli
signals extending over more than one spectral line, the average
amplitude was calculated at each of the 64 points of the y axis
(sum of amplitudes [a*0+b*0+c*a+d*a+ .. .]/number of spectral lines containing the embolus) (Fig 1). All subsequent data
manipulations were based on these average values. The maximal amplitude and the sum of amplitudes of each embolus
signal were then calculated. Only unidirectional emboli signals
arising from emboli in the MCA were studied.
Emboli signals detected in patients with carotid disease,
atrial fibrillation, and acute stroke were analyzed together,
forming the group of solid emboli signals. Emboli signals
detected during cardiac catheterization formed the group of
air emboli signals.
Two hundred emboli signals detected in patients with
carotid disease (n=102 emboli), atrial fibrillation (n=15 emboli), and acute stroke (n=83 emboli) and 200 emboli detected
during coronary angiography were randomly selected and used
as training data. Ideal cutoff limits based on the maximal
amplitude or the sum of amplitudes were calculated using
receiver operating characteristic (ROC) curves. Based on
statistical decision theory,16 these curves were developed in
the context of electronic signal detection and have been
applied to decision making in diagnostic systems in clinical
medicine. "• ls The ROC curve shows the various trade-offs
existing between proportions of true-positive and false-positive
responses, as the decision criterion is systematically varied, for
a given capacity to discriminate between positive and negative
cases.18 Each point of the curve represents the detection of
true positives, depending on the decision criteria. The calculated cutoff limits were then explored to determine sensitivity
and specificity in differentiating gaseous from formed emboli
signals. Afterward a combination of the cutoff values of
maximal amplitude and sum of amplitudes was used in the
same way. This algorithm was then evaluated on a test set of
501 formed (carotid disease patients, n=281 emboli; atrial
fibrillation patients, n=24 emboli; acute stroke patients,
n=196 emboli) and 995 gaseous emboli signals.
The same algorithm was applied to 5958 emboli signals
detected in 60 patients (age, 30 to 74 years; median, 59 years)
with mechanical prosthetic heart valves (Bjork-Shiley monostrut, n=38 patients, 5777 emboli; Medtronic-Hall, n=22
patients, 181 emboli) (duration since valve insertion, 21 ±5
months). The monitoring time in this group was 30 minutes,
but some of the patients underwent prolonged monitoring (up
to 3 hours).
Results
Emboli signals were detected in 94% of carotid
disease patients, 71% of acute stroke patients, and 40%
of atrial fibrillation patients. The detailed results of the
emboli signal counts in the acute stroke and carotid
disease patients have been described elsewhere.19-20
Training data consisting of 200 solid and 200 gaseous
emboli signals were used to develop the algorithm.
According to the ROC curves, the best cutoff values
were a maximal amplitude of 110 (Fig 2) and a sum of
amplitudes of 600 (Fig 3). Use of the maximal amplitude
provided a sensitivity of 85% and a specificity of 95.5%
in differentiating gaseous from solid emboli; use of the
sum of amplitudes provided a sensitivity of 97.5% and a
specificity of 91.5%. A combination of these two cutoff
values (emboli signals were classified as gaseous if their
maximal amplitude was >110 and their sum of amplitudes >600) provided a sensitivity of 83.5% and a
specificity of 90.5%.
A two-step combination was then used. Emboli signals were classified as gaseous when their maximal
amplitude was greater than 110. Remaining signals with
a smaller value were categorized depending on sum of
their amplitudes to be either gaseous, if this exceeded
600, or formed, if it did not. The sensitivity of this
algorithm was 99% and the specificity 96.5%.
On evaluating the algorithm, 991 of 995 gaseous
emboli signals and 449 of 501 solid emboli signals were
correctly identified (specificity, 89.8%; sensitivity,
99.6%).
The distribution of amplitudes, based on the combined test and evaluation data sets of gaseous and
formed emboli signals, is displayed in Fig 4. The spectral
distribution of gaseous emboli signals is highly variable.
Gaseous emboli had significantly higher maximal amplitudes (median, 234 [95% nonparametric confidence
interval {95% CI}, 228 to 250] versus 66 [95% CI, 62 to
70]; / ) <.0001, Mann-Whitney). The sum of amplitudes
was also significantly higher in gaseous than in formed
emboli signals (median, 2162 [95% CI, 1848 to 2355]
versus 308 [95% CI, 287 to 333]; P < . 0 0 0 1 ,
Mann-Whitney).
Subsequent application of the algorithm to the prosthetic valve patients resulted in 92.4% of emboli signals
(5503 of 5958) being classified as gaseous. The spectral
distribution of the amplitudes of this group is displayed
in Fig 4.
Discussion
The detection of emboli signals by TCD has been
widely reported, but the constitution and clinical significance of the underlying embolic substances have not
always been apparent. The potential exists to distin-
Georgiadis et al Differentiation Between Embolic Materials
MCG0U111
O c t 2 7 1993
1:27 p.. n.
cn/s
•
1OO-
II
11 O-
80
1561
WmmmB1
HH-128 R
15
D
Vn
Us
Ud
PI
RI
13:14:23
48 M M
4 Ch/S
O cn/s
O cn/s
o.. OO
o.. OO
<EHE TCMin>
MCG0U111
Oct 27 1993
1:27 p.n.
HH-128 R
IS
Downloaded from http://stroke.ahajournals.org/ by guest on June 18, 2017
<EME TCMin)
MCG0U111
Oct 27 1993
1:27 p.n.
HH-128 R
15
<EME TCWin>
FIG 1. A (top), Embolus signal in patient with mechanical prosthetic cardiac valve. A (bottom), Spectral analysis reveals that this
embolus signal is contained in six spectral lines (103 through 108).
B, Detail of spectral line 107, which consists of 128 points. The
amplitude of each point is calculated (a,, a 2 , . . . a)28). Subsequently the 128 amplitudes are compressed to 64 (for example,
a*15=a29+a3o/2). The embolus signal in panel A extends over 6
spectral lines. The average amplitude for each of the 64 points is
therefore calculated from a*n(1o3)+a*n(io4)+a*n()05)-i-a*n0o6)+a*n0o7,
+a*n(io8)/6 (n=1 to 64). The maximal amplitude in this example is
a17 (continuous line).
1562
Stroke Vol 25, No 8 August 1994
150 -
125
i '
•
medi
g-100
Curve 1
If
Curv*
e
•a 50 •
2
^J-trJ
Curva
—
*
•
*
-
^
0 0
50
100
150
200
250
300
350
cut-off (max amplitude)
Spectral lines
FIG 2. Graph shows receiver operating characteristic curve for
maximal amplitude. Dashed line indicates the percentage of
solid emboli signals with an amplitude greater than the cutoff
specified on the x axis; continuous line, the percentage of
gaseous emboli signals with an amplitude less than the specified
cutoff. The lines intersect at the maximum amplitude cutoff of
110.
Downloaded from http://stroke.ahajournals.org/ by guest on June 18, 2017
guish between different embolus types on the basis of
the reflected Doppler signal. Furthermore, the vast
difference in density between gaseous and formed emboli suggests that these two emboli categories may be
particularly susceptible to differentiation by ultrasound.
To date, no definitive guidelines have been developed to
distinguish between emboli signals caused by gaseous
and formed embolic materials.
Earlier attempts to distinguish between emboli types,
by means of Doppler spectral analysis applied to in vitro
and animal models, have been unsuccessful. This may in
part be attributable to the use of embolic material of
sizes incomparable to the in vivo situation. Sizes of
biological formed emboli, prepared using microscopical
dissection and studied with TCD in previous studies,
were 0.4 to 1.5 mm (Russel et al7) and 1 to 5 mm
(Markus and Brown10). Pugsley21 was unable to distinguish between emboli signals produced by 30- to 90-fim
microspheres and air emboli. Stump et al11 compared
the emboli signals produced by three sizes of microsphere (50, 100, and 200 ^m) to air and biological
emboli signals. They suggested that differentiation between gaseous and formed emboli may be possible.
Albin et al22 reached the same conclusion. They compared the emboli signals produced by bubbles of various
sizes (0.8 to 100 mm3) and aggregates of 1- and 25-mm3
microspheres, diluted in 0.1 mL of blood in an animal
0
1500
3000
4500
6000
7500
cut-off |sum of amplitudes)
FIG 3. Graph shows receiver operating characteristic curve for
sum of amplitudes. Dashed line indicates the percentage of solid
emboli signals with an amplitude greater than the cutoff specified on the x axis; continuous line, the percentage of gaseous
emboli signals with an amplitude less than the specified cutoff.
The cutoffs intersect at a sum of amplitudes value of 600.
FIG 4. Graph shows spectral distribution of gaseous emboli
signals (curve 1), emboli signals detected in prosthetic valve
patients (curve 2), and solid emboli signals (curve 3). Cl indicates confidence interval.
model, and assumed that simple pattern recognition
techniques could be used to differentiate between gaseous and solid emboli signals. Microemboli signals
detected by TCD in asymptomatic patients are probably
generated by smaller particles than the biological material used in the above studies, as suggested by studies
performed with a canine model by Moody et al.23 After
as many as 25xlO 6 15-/xm microspheres were introduced into the left circulation the dogs appeared grossly
intact, while a much smaller number of 50-fim microspheres produced neurological injury.23
Methods previously used to analyze the intensity of
emboli signals have lacked sensitivity. Markus and
Brown10 used the maximal amplitude to differentiate
various formed embolic materials. However, the maximal amplitude has only a finite range of values (0 . . .
346), which could explain the reported absence of a
further rise in amplitude when emboli with a maximum
dimension of 2 mm or more were introduced.10 Furthermore, this analysis method does not take the embolus
signal spectral distribution into account. In particular,
air emboli signals, which cause an intensity increase
over a wide range of frequencies,12 cannot be adequately characterized by this method (Fig 4). Russell et
al7 used the color scale of the Doppler display to
estimate the intensity increase in decibels. Spencer14
determined the relative amplitude of embolic signals by
estimating the noise threshold level required to suppress the background and emboli signals. The combined
sum of intensities and maximal intensity technique
developed in this study provides a more reliable means
of characterizing emboli signals, since it takes both the
magnitude and spectral distribution of the intensity
increase into account.
Our results suggest that differentiation between emboli
signals arising from formed and gaseous emboli, as detected by TCD in vivo, is possible. It is unclear if discrimination among various types of formed emboli is feasible
using the same method. Since the constitution of formed
emboli is not apparent in any clinical group, further in
vitro and animal studies are required. However, the generation of sufficiently small emboli still proves difficult.
Cerebral embolization models, based on endogenous emboli generation, were recently described.24-25 Such models
combined with TCD monitoring may prove better than in
vitro models, since the generated microemboli correspond
to the clinical situation.
Georgiadis et al Differentiation Between Embolic Materials
The fact that the vast majority of emboli signals
detected in patients with mechanical prosthetic heart
valves correspond to the criteria set for gaseous emboli
lends further support to the hypothesis that the embolic
material in these patients is gaseous.3 The proportion of
emboli signals classified as solid is explicable by the
specificity of the algorithm of 85% to 90%. In recent
years, previously unrecognized phenomena, such as the
formation of cavitation microbubbles,26"28 and fibrin
"strands" 29 were described in normally functioning
prosthetic valves. Although the cavitation potential of
prosthetic valves has been documented, it was assumed
that such cavitation bubbles have a very short life span
and cannot be detected far from the site of creation.
However, three different types of cavitation bubbles
were described in a recent study of prosthetic valve
function, one of which is larger and has a longer life
span than previously assumed.28 The exact mechanism
of the microemboli generation associated with the valve
implant remains to be determined.
Downloaded from http://stroke.ahajournals.org/ by guest on June 18, 2017
References
1. Caplan LR. Brain embolism, revisited. Neurology. 1993;43:
1281-1287.
2. Tegeler CH, Hitchings LP, Eicke M, Leighton J, Fredericks RK,
Downes TR, Stump DA, Burke GL. Microemboli detection in
stroke associated with atrial fibrillation. J Cardiovasc TechnoL
1990;9:283-284. Abstract.
3. Georgiadis D, Grosset DG, Kelman A, Faichney A, Lees KR.
Prevalence and characteristics of intracranial microembolic signals
in patients with different types of prosthetic cardiac valves. Stroke.
1994;25:587-592.
4. Siebler M, Sitzer M, Steinmetz H. Detection of intracranial microemboli in patients with symptomatic extracranial carotid disease.
Stroke. 199203:1652-1654.
5. Padayachee TS, Parsons S, Theobald R, Lindley J, Gosling RG,
Deverall PB. The detection of microemboli in the middle cerebral
artery during cardiopulmonary bypass: a transcTanial Doppler
ultrasound investigation using membrane and bubble oxygenators.
Ann Thorac Surg. 1987;44:298-302.
6. 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.
7. Russell D, Madden KP, Clark WM, Sandset PM, Zrvin JA.
Detection of arterial emboli using Doppler ultrasound in rabbits.
Stroke, 1991^2:253-258.
8. Russell D, Brucher R, Madden KP, Clark WM, Sandset PM, Zivin
JA. The intensity of the Doppler signal caused by arterial emboli
depends on embolus size. In: Oka M, von Reutern GM, Furuhata
H, Kodaira K, eds. Recent Advances in Neurosonology. Amsterdam,
the Netherlands: Excerpta Medica; 1992:57-60.
9. Pugsley W, Klinger L, Paschalis C, Aspey B, Newman S, Harrison
M, Treasure T. Microemboli and cerebral impairment during
cardiac surgery. Vase Surg. 1990^24:34-43.
1563
10. Markus HS, Brown MM. Differentiation between different pathological cerebral embolic materials using transcranial Doppler in an
in vitro model. Stroke. 1993;24:l-5.
11. Stump DA, Tegeler CH, Hager R, Culler R, Entine G. In vivo
emboli detection: simultaneous CW Doppler monitoring and
B-mode imaging of gaseous and solid material. Stroke. 1992;23:474.
Abstract.
12. Nicholls SC, Glickerman DJ, Lam AM. Cerebral embolization:
transcranial Doppler monitoring in the swine model. Stroke. 1992;
23:473. Abstract.
13. Grosset DG, Georgiadis D, Kelman AW, Lees KR. Quantification
of emboli signals in patients with cardiac and carotid disease.
Stroke. 1993^4:1922-1924.
14. Spencer MP. Detection of cerebral arterial emboli. In: Newell
DW, Aaslid R, eds. Transcranial Doppler. New York, NY: Raven
Press, Publishers; 1992:215-230.
15. Brucher R, Russell D. Spectral characteristics of emboli with
artifact suppression. Stroke, 1993;24:510. Abstract.
16. Wald A. Statistical Decision Functions. New York, NY: John Wiley
& Sons, Inc; 1950.
17. Swets JA. The relative operating characteristic in psychology.
Science. 1973;182:990-1000.
18. Swets JA, Pickett RM, Whitehead SF, Getty DJ, Schnur JA, Swets
JB, Freeman BA. Assessment of diagnostic technologies. Science.
1979;205:753-759.
19. Georgiadis D, Grosset DG, Quin R, Nicholl ARJ, Bone I, Lees
KR. Intracerebral emboli in patients with carotid disease. Eur J
Vase Surg. 1994;8:3O9-314.
20. Grosset DG, Georgiadis D, Abdullah I, Bone I, Lees KR. Doppler
emboli vary according to stroke subtype. Stroke. 1994;25:382-384.
21. Pugsley W. The use of Doppler ultrasound in the assessment of
microemboli during cardiac surgery. Perfusion. 1989;4:115-122.
22. Albin MS, Bunegin L, Garcia C, McKay W. The transcranial
Doppler can image microaggregates of intracranial air and particulate matter. Neurosurg Anesthesiol 1989;2:134-135. Abstract.
23. Moody DM, Bell MA, Challa VR, Johnston WE, Prough DS. Brain
microemboli during cardiac surgery or aortography. Ann NeuroL
1990-,28:477-486.
24. Kessler C, Kelly AB, Suggs WD, Weissman JD, Epstein CM,
Hanson SR, Harker LA. Induction of transient neurological dysfunction in baboons by platelet microemboli. Stroke. 1992;23:
697-702.
25. Dietrich DW, Dewanjee S, Prado R, Watson BD, Dewanjee MK.
Transient platelet accumulation in the rat brain after common
carotid artery thrombosis: an '"In-labeled platelet study. Stroke.
1993;24:1534-1540.
26. Graf T, Fischer H, Reul H, Rau G. Cavitation potential of
mechanical heart valve prostheses. Int J Artif Organs. 1991;14:
169-174.
27. Lamson TC, Stinebring DR, Deutsch S, Rosenberg G, Tarbell JM.
Real-time in vitro observation of cavitation in a prosthetic heart
valve. ASAIO Transactions. 1991;37:M351-M353.
28. Kingsbury C, Kafesjian R, Guo G, Adlparvar P, Unger J, Quinjano
RC, Graf T, Fisher H, Reul H, Rau G. Cavitation threshold with
respect to dP/dt: evaluation in 29 mm bileaflet, pyrolitic carbon
heart valves. Int J Artif Organs. 1993;16:515-520.
29. Stoddard MF, Dawltins PR, Longaker RA. Mobile strands are
frequently attached to the St. Jude Medical mitral valve prosthesis
as assessed by two-dimensional transesophageal echocardiography.
Am Heart J. 1992;124:671-674.
Differentiation between gaseous and formed embolic materials in vivo. Application in
prosthetic heart valve patients.
D Georgiadis, T G Mackay, A W Kelman, D G Grosset, D J Wheatley and K R Lees
Downloaded from http://stroke.ahajournals.org/ by guest on June 18, 2017
Stroke. 1994;25:1559-1563
doi: 10.1161/01.STR.25.8.1559
Stroke is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 1994 American Heart Association, Inc. All rights reserved.
Print ISSN: 0039-2499. Online ISSN: 1524-4628
The online version of this article, along with updated information and services, is located
on the World Wide Web at:
http://stroke.ahajournals.org/content/25/8/1559
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally
published in Stroke can be obtained via RightsLink, a service of the Copyright Clearance Center, not
the Editorial Office. Once the online version of the published article for which permission is being
requested is located, click Request Permissions in the middle column of the Web page under Services.
Further information about this process is available in the Permissions and Rights Question and Answer
document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Stroke is online at:
http://stroke.ahajournals.org//subscriptions/