Note: This copy is for your personal non-commercial use only.

To order presentation-ready
copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

Original Research  n  Neuroradiology

Multiphase CT Angiography: A
New Tool for the Imaging Triage of
Patients with Acute Ischemic Stroke1

Bijoy K. Menon, MD
Purpose: To describe the use of an imaging selection tool, mul-
Christopher D. d’Esterre, PhD
tiphase computed tomographic (CT) angiography, in
Emmad M. Qazi, BSc
patients with acute ischemic stroke (AIS) and to dem-
Mohammed Almekhlafi, MD2 onstrate its interrater reliability and ability to help deter-
Leszek Hahn, PhD mine clinical outcome.
Andrew M. Demchuk, MD
Mayank Goyal, MD Materials and The local ethics board approved this study. Data are from
Methods: the pilot phase of PRoveIT, a prospective observational
study analyzing utility of multimodal imaging in the triage
of patients with AIS. Patients underwent baseline unen-
hanced CT, single-phase CT angiography of the head and
neck, multiphase CT angiography, and perfusion CT. Mul-
tiphase CT angiography generates time-resolved images of
pial arteries. Pial arterial filling was scored on a six-point
ordinal scale, and interrater reliability was tested. Clinical
outcomes included a 50% or greater decrease in National
Institutes of Health Stroke Scale (NIHSS) over 24 hours
and 90-day modified Rankin Scale (mRS) score of 0–2.
The ability to predict clinical outcomes was compared
between single-phase CT angiography, multiphase CT an-
giography, and perfusion CT by using receiver operating
curve analysis, Akaike information criterion (AIC), and
Bayesian information criterion (BIC).

Results: A total of 147 patients were included. Interrater reliability

for multiphase CT angiography is excellent (n = 30, k =
0.81, P , .001). At receiver operating characteristic curve
analysis, the ability to predict clinical outcome is modest
(C statistic = 0.56, 95% confidence interval [CI]: 0.52,
0.63 for 50% decrease in NIHSS over 24 hours; C sta-
tistic = 0.6, 95% CI: 0.53, 0.68 for 90-day mRS score of
0–2) but better than that of models using single-phase CT
1
 From the Calgary Stroke Program, Department of Clinical angiography and perfusion CT (P , .05 overall). With AIC
Neurosciences (B.K.M., C.D.d.E., E.M.Q., M.A., A.M.D., and BIC, models that use multiphase CT angiography are
M.G.), Department of Radiology (B.K.M., C.D.d.E., M.A.,
better than models that use single-phase CT angiography
L.H., A.M.D., M.G.), Department of Community Health
Sciences (B.K.M.), Hotchkiss Brain Institute (B.K.M., A.M.D.,
and perfusion CT for a decrease of 50% or more in NIHSS
M.G.); and Seaman Family MR Research Centre, Foothills over 24 hours (AIC = 166, BIC = 171.7; values were lowest
Medical Centre (B.K.M., C.D.d.E., A.M.D., M.G.), University for multiphase CT angiography) and a 90-day mRS score
of Calgary, 1403-29th St NW, Calgary, AB, Canada T2N 2T9. of 0–2 (AIC = 132.1, BIC = 137.4; values were lowest for
Received September 22, 2014; revision requested October multiphase CT angiography).
22; revision received December 17; accepted December
19; final version accepted January 12, 2015. B.K.M. and
Conclusion: Multiphase CT angiography is a reliable tool for imaging
M.G. supported by a Canadian Institute of Health Research
Operating Grant and bridge funding from the Faculty of selection in patients with AIS.
Medicine, University of Calgary. Address correspondence
to M.G. (e-mail: [email protected]). q
 RSNA, 2015
2
 Current address: Department of Internal Medicine, King
Abdulaziz University, Jeddah, Saudi Arabia. Online supplemental material is available for this article.

q
 RSNA, 2015

Radiology: Volume 000: Number 0—   2015  n  radiology.rsna.org 1

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

I
n the past few years, the treatment (CT), single-phase CT angiography, per- of patients with AIS. Herein, we will de-
of acute ischemic stroke (AIS) has fusion CT, and magnetic resonance (MR) scribe the tool, its interrater reliability,
changed dramatically (1). Newer imaging. Unenhanced CT has moderate and its utility for making clinical decisions
mechanical devices offer rapid and suc- interrater reliability, even among experts in patients with AIS.
cessful recanalization in the majority of (10–13). Reliability in interpreting early
patients who undergo treatment (2–4). ischemic changes is less in patients who
Even with this progress, many patients present within 90 minutes after stroke Materials and Methods
who have undergone treatment do not symptom onset and in those who are Inclusion criteria for the study are as
do well clinically (5,6). Nonetheless, aged, and it is affected by patient motion follows: (a) patient presented to the
data from previous trials show that (14). Single-phase CT angiography does emergency department with symptoms
clinical outcome improves if patients not have temporal resolution; therefore, consistent with ischemic stroke, (b) pa-
(a) have a salvageable brain at presen- collateral status is mislabeled in many tients older than 18 years, and (c) base-
tation and (b) undergo early recanali- patients (15). Both perfusion CT and MR line imaging included multiphase CT
zation (6–8). Every 30-minute delay in imaging are susceptible to patient motion angiography performed within 12 hours
treatment could increase the risk of and require trained personnel to process of stroke symptom onset and initiated
poor clinical outcome by around 14% the data (16,17). Dynamic CT angiogra- before recanalization therapy. Exclusion
(9). Thus, an ideal imaging selection phy is a technique that derives time-re- criteria were as follows: (a) intracranial
tool should enable one to detect a sal- solved images of pial arterial filling from hemorrhage identified at baseline CT;
vageable brain quickly and reliably and perfusion CT images; however, it needs (b) previous moderate to large stroke
should be widely available. postprocessing and whole-brain perfu- in the ipsilesional hemisphere; (c) mod-
Current imaging techniques include sion CT (18,19). Conventional angiogra- ified Rankin scale (mRS) score greater
unenhanced computed tomography phy is invasive, resource intensive, and than 2 at baseline; (d) patient unable
not feasible as a fast diagnostic tool (20). to undergo CT angiography because of
Advances in Knowledge Thus, we developed an imaging tool, recent estimated creatinine clearance of
nn Multiphase CT angiography is an multiphase CT angiography, that gives cli- less than 60 mL/min, contrast material
imaging tool that provides three nicians information on degree and extent allergy, or other reasons; (e) participa-
time-resolved images of pial arte- of pial arterial filling in the whole brain tion in another study that results in the
rial filling in the whole brain, in a time-resolved manner. Furthermore, patient receiving an investigational drug
unlike conventional single-phase this technique is quick to perform and
CT angiography. yields images that are easy to acquire and
interpret. In this study, we used pilot data Published online before print
nn Interrater reliability for multi- from the PRoveIT (Precise and Rapid as- 10.1148/radiol.15142256  Content codes:
phase CT angiography is excel- sessment of collaterals using multi-phase
lent (n = 30, k= 0.81, P , .001). Radiology 2015; 000:1–11
CTA in the triage of patients with acute
nn At receiver operating curve ischemic stroke for IA Therapy) study, an Abbreviations:
analysis, the ability to predict ongoing prospective observational study AIC = Akaike information criterion
clinical outcome (50% decrease AIS = acute ischemic stroke
that seeks to understand the utility of
BIC = Bayesian information criterion
in National Institutes of Health multimodal imaging in the imaging triage CBF = cerebral blood flow
Stroke Scale from baseline to 24 CI = confidence interval
hours) based on assessment of MCA = middle cerebral artery
pial arterial filling in the ischemic Implications for Patient Care mRS = modified Rankin Scale
region is modest (C statistic = NIHSS = National Institutes of Health Stroke Scale
0.56; 95% confidence interval nn Multiphase CT angiography is an TICI = Thrombolysis in Cerebral Infarction
[CI]: 0.52, 0.63) but higher than imaging tool with excellent inter-
Author contributions:
that for single-phase CT angiog- rater reliability that can be used Guarantors of integrity of entire study, B.K.M., M.G.;
raphy (C statistic = 0.55; 95% to predict clinical outcomes in study concepts/study design or data acquisition or data
CI: 0.49, 0.6), perfusion CT mis- patients with acute ischemic analysis/interpretation, all authors; manuscript drafting or
match ratio greater than 1.8 (C stroke. manuscript revision for important intellectual content, all
authors; approval of final version of submitted manuscript,
statistic = 0.49; 95% CI: 0.45, nn Unlike perfusion CT, multiphase
all authors; agrees to ensure any questions related to the
0.52), mismatch ratio greater CT angiography does not need work are appropriately resolved, all authors; literature
than 3 (C statistic = 0.47; 95% any mathematical algorithm or research, B.K.M., C.D.d.E., E.M.Q., A.M.D.; clinical studies,
CI: 0.41, 0.53), and perfusion CT complex postprocessing at an B.K.M., E.M.Q., M.A., A.M.D., M.G.; experimental studies,
infarct volume greater than 80 independent workstation; it also B.K.M., E.M.Q., A.M.D., M.G.; statistical analysis, B.K.M.,
mL (C statistic = 0.46; 95% CI: requires a lower radiation dose C.D.d.E., E.M.Q.; and manuscript editing, B.K.M., C.D.d.E.,
E.M.Q., A.M.D., M.G.
0.4, 0.5) (P = .01 for comparison and no additional contrast
of C statistic). material. Conflicts of interest are listed at the end of this article.

2 radiology.rsna.org  n Radiology: Volume 000: Number 0—   2015

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

Figure 1 Figure 2

Figure 1:  Multiphase CT angiography image, with

each phase represented by an arrow. The first phase
(long solid arrow) is conventional arch-to-vertex
CT angiography. The next two phases (short solid
arrows) are sequential skull base–to-vertex acqui-
sitions performed in the midvenous and late venous
phases. Dashed arrows indicate movement of the
scanner in between image acquisitions.

or therapy; and (f) any terminal illness

(patient not expected to survive longer Figure 2:  Multiphase CT angiography images. Top row: Images in a patient with a left M1 MCA occlusion
than 1 year). We analyzed two clinical (arrow) and good collaterals (backfilling arteries). Middle row: Images in a patient with a left M1 MCA occlu-
outcomes in this study, namely (a) ma- sion (arrow) and intermediate collaterals. Bottom row: Images in a patient with a right M1 MCA occlusion
jor neurologic improvement at 24-hour (arrow) and poor collaterals (minimal backfilling arteries).
follow-up, defined as a 50% decrease in
National Institutes of Health Stroke Scale
(NIHSS) over 24 hours and (b) mRS are from the skull base to the vertex in for axial, coronal, and sagittal images
score of 0–2 at 90 days. The local institu- the equilibrium/peak venous and late of the circle of Willis were performed
tional review board approved the study. venous phases in the healthy brain. Im- with 3-mm thickness at 1-mm intervals.
ages were acquired with a 0.625-mm Thick-section axial maximum intensity
Imaging Protocol and Analysis section thickness. The first phase of CT projections at 24-mm thickness and
All patients underwent standard unen- angiography from the arch to the ver- 4-mm intervals were also reconstructed.
hanced CT with 5-mm section thickness tex was acquired in less than 7 seconds, An important feature of our imaging
followed by head and neck CT angiogra- with an average dose length product of protocol was that the two additional
phy, including multiphase CT angiogra- 700–800 mGy·cm. The second phase phases of multiphase CT angiography
phy and perfusion CT. was acquired after a delay of 4 seconds use no additional contrast material; the
Multiphase CT angiography.—This that allows for table repositioning to the total radiation dose as per our imaging
technique generates time-resolved ce- skull base. Scanning duration for each protocol was less than that in many es-
rebral angiograms of brain vasculature additional phase was 3.4 seconds. Thus, tablished stroke centers (21) (Table 1).
from the skull base to the vertex in three the three phases were each 8 seconds In addition, we used an AccuProbe me-
phases after contrast material injection apart. A total of 80 mL of contrast ter (RadCal, Monrovia, Calif) equipped
(Figs 1, 2). Aortic arch vertex CT an- material (68% ioversol, Optiray 320; with a 20X5-3 ion chamber and a hu-
giography performed with a multidetec- Mallinckrodt, St Louis, Mo) was injected man head phantom to measure the total
tor CT scanner made up the first phase. at a rate of 5 mL/sec and followed by absorbed radiation dose (in milligrays)
Image acquisition was timed to occur a 50-mL normal saline chase at a rate for the eye with single-phase CT angiog-
during the peak arterial phase in the of 6 mL/sec. The axial images were re- raphy, multiphase CT angiography, and
healthy brain and was triggered by bolus constructed at 1-mm overlapping sec- perfusion CT (Table 1). This analysis re-
monitoring. The remaining two phases tions, and multiplanar reconstructions vealed that the total radiation dose for

Radiology: Volume 000: Number 0—   2015  n  radiology.rsna.org 3

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

Table 1 CT perfusion.—A total of 45 mL of

the same contrast agent was injected
Mean Effective Radiation Dose with Our Protocol for Multimodal Imaging Including at a rate of 4.5 mL/sec followed by a
Multiphase CT Angiography When Compared with a Conventional Protocol for saline chase of 40 mL at a rate of 6
Multimodal Imaging mL/sec. Axial shuttle (step-and-shoot)
Mean Estimated Effective Mean Estimated mode was used to cover an 8-cm sec-
Dose in an Established Effective Dose at Radiation Dose to tion of the brain, including the intracra-
Type of Examination Center (mSv)* Our Center (mSv) the Eye (mGy)† nial artery at a 5-mm section thickness.
Scanning began after a 5-second delay
Unenhanced head CT 2.7 6 0.3 2.0 6 0.5 NA
after contrast material injection, with
Routine head and neck 5.4 6 2.2 5.0 6 0.5 15
24 passes performed over 66 seconds.
CT angiography
Total exposure time was 19.20 seconds.
Two additional phases of NA 1.0 6 0.5 45
multiphase CT angiography
See Appendix E1 (online) for CT perfu-
CT perfusion 4.9 6 0.0 3.5 6 0.5 200 sion postprocessing details.
Contrast-enhanced head CT 2.6 6 0.2 NA NA Recanalization and reperfusion.—
Smart prep 0.1 6 0.1 0.1 6 0.1 NA Recanalization and reperfusion were
Total dose 16 12 … assessed either on conventional cere-
bral angiograms obtained at the end of
Note.—NA = not applicable. the intraarterial procedure (by using
* Center data is from Mnyusiwalla et al (21).
†
Measured by using the aforementioned meter equipped with a 20X5–3 ion-chamber and a human head phantom. These the Thrombolysis in Cerebral Infarction
radiation doses conform to the 2012 International Commission on Radiological Protection guidelines (22). [TICI] score) or on CT angiograms of
the circle of Willis obtained within 2–4
hours after baseline imaging in patients
who underwent only intravenous tissue
Figure 3
plasminogen activator by using the TICI
CT angiography score. Successful re-
canalization was defined as a modified
TICI score of 2b/3 or a TICI CT angiog-
raphy score of 2b/3 (23).

Testing Multiphase CT Angiography

The various steps in testing multiphase
CT angiography are described in Figure
E1 (online).
Interrater reliability.—Two raters
(B.K.M., M.A.) independently assessed
Figure 3:  Pial arterial filling at single-phase CT angiography (CTA) and multiphase CT angiography shows multiphase CT angiography in 30 ran-
incongruence between the scores and mislabeling of many patients with better pial arterial filling at multi- domly chosen subjects by using the
phase CT angiography as having a poor score at single-phase CT angiography. six-point ordinal scale that was then re-
classified into three clinically relevant
categories (ie, good, intermediate, and
the eye with multiphase CT angiography by using a six-point scale (Fig 3). If no poor pial arterial filling) (Table 2). Inter-
was well within the acceptable limits (as occlusion was evident, pial filling of the rater reliability was measured by using
per the International Commission on symptomatic hemisphere was compared unweighted k values. (Details on the sta-
Radiological Protection Guidelines 2012) with that of the contralateral side. The tistical method are given in Appendix E1
and significantly less than that with per- six-point scale was then trichotomized [online]).
fusion CT (22). into three categories of pial arterial fill- Agreement on clinical decision mak-
All postprocessing with multiphase ing. The method for scoring pial arterial ing.—Since we did not have a reference
CT angiography is automated and avail- filling at single-phase and multiphase CT standard with which to assess concur-
able for review within 2–3 minutes of CT angiography is described in Table 2. Two rent validity (agreement) of multiphase
angiography. Pial arterial filling in the is- raters (B.K.M., M.A.) who were blinded CT angiography vis-á-vis other imaging
chemic territory was measured in the first to treatment decisions, perfusion CT, tools, we compared its ability to assist
phase of CT angiography (single-phase and follow-up data scored unenhanced in making a clinical decision with that
CT angiography) and during multiphase CT, single-phase CT angiography, and of other available imaging tools. We did
CT angiography by comparing it to simi- multiphase CT angiography images by this by designing an imaging experiment
lar arteries in the unaffected hemisphere consensus. in which two authors (B.K.M., M.A.)

4 radiology.rsna.org  n Radiology: Volume 000: Number 0—   2015

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

Table 2
Pial Arterial Filling Score within the Symptomatic Ischemic Territory Using Single- and Multiphase CT Angiography
Score Single-Phase CT Angiography Multiphase CT Angiography

5 When compared with asymptomatic contralateral hemisphere, there is When compared with the asymptomatic contralateral hemisphere, there
increased or normal prominence and extent of pial vessels within the is no delay and normal or increased prominence of pial vessels/normal
ischemic territory in the symptomatic hemisphere extent within the ischemic territory in the symptomatic hemisphere
4 When compared with the asymptomatic contralateral hemisphere, there When compared with the asymptomatic contralateral hemisphere, there
is slightly reduced prominence and extent of pial vessels within the is a delay of one phase in filling in of peripheral vessels, but prominence
ischemic territory in the symptomatic hemisphere and extent is the same
3 When compared with the asymptomatic contralateral hemisphere, When compared with the asymptomatic contralateral hemisphere, there
there is moderately reduced prominence and extent of pial vessels is a delay of two phases in filling in of peripheral vessels or there is a
within the ischemic territory in the symptomatic hemisphere one-phase delay and significantly reduced number of vessels in the
ischemic territory
2 When compared with the asymptomatic contralateral hemisphere, When compared with the asymptomatic contralateral hemisphere, there
there is decreased prominence and extent and regions with no is a delay of two phases in filling in of peripheral vessels and decreased
vessels within the ischemic territory in the symptomatic hemisphere prominence and extent or a one-phase delay and some ischemic regions
with no vessels
1 When compared with the asymptomatic contralateral hemisphere, there When compared with the asymptomatic contralateral hemisphere,
are just a few vessels visible in the occluded vascular territory there are just a few vessels visible in any phase within the occluded
vascular territory
0 When compared with the asymptomatic contralateral hemisphere, When compared with the asymptomatic contralateral hemisphere, there
there are no vessels visible within the ischemic territory are no vessels visible in any phase within the ischemic vascular territory

assessed unenhanced CT followed by were recorded as an uncertain response. models were compared by using the x2
single-phase CT angiography, multiphase All responses (“yes,” “no,” and “uncer- test of Gönen (28). Since comparison of
CT angiography, and perfusion CT to an- tain”) for each imaging modality were models by using receiver operating char-
swer the following two questions: Is the reported as proportions. acteristic curves may result in misclassi-
patient a candidate for intravenous tissue Predictive ability.—We compared fication errors, we also used Akaike in-
plasminogen activator? Is the patient a the ability of multiphase CT angiogra- formation criterion (AIC) and Bayesian
candidate for intraarterial therapy? All phy to enable prediction of both clini- information criterion (BIC) to compare
patients were assumed to have fulfilled cal outcomes vis-á-vis single-phase CT models (29). These latter methods have
clinical characteristics for treatment eligi- angiography and perfusion CT. For sin- the ability to express the probability that
bility when images were being read. The gle-phase CT angiography, a pial arte- each model is correct when compared
readers were not provided information rial filling score of 0–2 was considered with the best model (ie, the one with
on the time from when normal images poor; therefore, the patient was not the highest probability to minimize in-
were last obtained. Responses recorded likely to benefit from recanalization. For formation loss). A model with the low-
for each imaging modality were “yes,” multiphase CT angiography, a score of est AIC or BIC score is the best model
“no,” or “uncertain.” Uninterpretable 0–3 was considered poor and therefore (29). Each of the previously mentioned
images were classified as uncertain. The unlikely to benefit from recanalization analyses was restricted to patients in
raters used clinically relevant commonly (Table 2). Separate logistic regression whom all information on the dependent
used prespecified rules for image inter- models were developed for each diag- variable and classifier was available. We
pretation. These rules are described in nostic tool (ie, single-phase CT angiog- also performed additional sensitivity
detail in Appendix E1 (online). Salvage- raphy, multiphase CT angiography, and analyses with the previously described
able brain was measured by using pre- perfusion CT [with mismatch ratios models restricted to patients (a) with
defined perfusion thresholds and two .1.8 vs 1.8 and .3.0 vs 3.0 and in- proximal anterior circulation occlusions,
predefined mismatch ratios (ie, .1.8 and farct volume ,80 mL vs 80 mL]) as (b) who underwent revascularization
.3.0) (24,25). Baseline infarct volume of predictor variable. For each model, the therapy, and (c) who had early recana-
80 mL or more at perfusion CT was con- ability of the individual diagnostic tool to lization data.
sidered large; therefore, in such cases the aid in determining clinical outcome was
response was “no” for both intravenous assessed by using the area under the re-
tissue plasminogen activator and intraar- ceiver operating characteristic curve (or Results
terial therapy (26,27). Severe noncorrect- C statistic) derived from the receiver op- A total of 147 patients were included
able patient motion (Fig E2 [online]) and erating characteristic curves of the logis- in the present study. Mean age was
lack of appropriate arterial input function tic regression model. The C statistics of 72 years 6 13.1 (standard deviation),

Radiology: Volume 000: Number 0—   2015  n  radiology.rsna.org 5

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

Table 3
Certainty in Clinical Decision Making for Intravenous Tissue Plasminogen Activator Administration and Intraarterial Therapy with
Each Baseline Imaging Modality and Paradigm
Intravenous Tissue Plasminogen Activator Intraarterial Therapy
No. of No. of
Imaging Modality and Criteria Patients Yes (%) No (%) Uncertain (%) Patients Yes (%) No (%) Uncertain (%)

Unenhanced CT 147 89.8 7.5 2.7 147 32.6 8.8 58.5

Multiphase CT angiography and unenhanced CT 147 90.5 8.8 0.7 147 51.0 47.6 1.4
Single-phase CT angiography and unenhanced CT 147 83.0 14.3 2.7 147 44.2 51.0 4.8
Baseline infarct volume ,80 mL 145 88.9 6.9 4.1 85 37.2 58.6 4.1
Mismatch ratio 1.8 145 92.4 3.4 4.1 85 40.7 55.2 4.1
Mismatch ratio 3.0 145 82.7 13.1 4.1 85 34.5 61.4 4.1

Note.—Agreement between imaging modalities for clinical decision making is described in the text.

Figure 4
Figure 4:  Multimodal CT imaging at 2 hours
51 minutes after symptom onset in a 47-year-old
woman with NIHSS of 20 and right hemisphere
symptoms. A, Unenhanced CT shows movement
artifact; however, ASPECTS score was 7. B, A prox-
imal right M1 MCA occlusion is seen (i). Multiphase
CT angiography (three phases) maximum intensity
projection images are shown (ii, iii, iv). Pial arterial
filling is modest, with delay of two phases and some
regions indicating minimal filling when compared
with the contralateral side, thus indicating that no
treatment be performed. C, Perfusion CT Tmax and
cerebral blood flow (CBF) maps (i, ii ). Tissue with
Tmax greater than 6 seconds (pink) is superimposed
onto the CT perfusion average maps for both gray
and white matter (iii and iv, respectively). CBF
less than 10 mL·min21·100 g21 and less than 7
mL·min21·100 g21 for gray and white, respectively,
is flooded in blue on the CT perfusion average
maps (iii, iv). CBF-defined infarct core is 100 mL. A
mismatch ratio (total Tmax hypoperfusion volume/
total CBF infarct volume) of 1.7 and a large infarct
core indicates that no treatment should be per-
formed. Multiphase CT angiography and perfusion
CT imaging are congruent for treatment decision.
D, Diffusion MR images at 24 hours after admission
show the final infarct as hyperintense.

49.7% were male, median baseline NI-

HSS was 9 (interquartile range, 13),
median Alberta Stroke Program Early
CT Score at unenhanced CT was 9 (in-
terquartile range, 4), and median time
from stroke symptom onset to base-
line CT was 133 minutes (interquartile
range, 188 minutes). Distribution of oc-
clusions was as follows: internal carotid

6 radiology.rsna.org  n Radiology: Volume 000: Number 0—   2015

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

artery (six of 147 patients), M1 middle Figure 5

cerebral artery (MCA) (60 of 147 pa-
tients), M2 MCA (21 of 147 patients),
posterior cerebral artery (three of 147
patients), distal occlusions (24 of 147
patients), and no occlusions (33 of
147 patients). Distribution of pial arte-
rial filling with single- and multiphase
CT angiography is shown in Figure 3.
Single-phase CT angiography consis-
tently resulted in underestimation of
pial arterial filling when compared with
multiphase CT angiography; thus, many
patients with moderate pial arterial fill-
ing at multiphase CT angiography were
labeled as having a poor score. When
we used a priori thresholds for infarct
and penumbra for both gray and white
matter at perfusion CT, median mis-
match ratio was 6.6 (range, 1.2–319.6),
while mean baseline infarct volume was
18.9 mL 6 31.1. Fifty-one patients un-
derwent intravenous thrombolysis, 24
underwent intravenous and intraarterial
therapy, seven underwent intraarterial
therapy alone, and 44 underwent conser-
vative treatment. Early reperfusion data
were available in 71 patients; 42 (59%)
achieved reperfusion. Fifty-six (38.1%)
patients achieved the primary clinical
outcome (50% decrease in NIHSS over
24 hours), while 72 of 119 (60.5%) had
an mRS score of 0–2 at 90 days.

Interrater Reliability
Interrater reliability for pial arterial fill-
ing with multiphase CT angiography was
excellent (n = 30, k = 0.81, P , .001).

Agreement on Clinical Decision Making Figure 5:  Multimodal CT images obtained 2 hours 18 minutes after symptom onset in an 87-year-old
Table 3 describes “yes,” “no,” and “un- woman with an NIHSS of 15 and left hemisphere symptoms. A, Unenhanced CT ASPECTS score was 6.
certain” for intravenous tissue plasmino- B, A proximal left M1 MCA occlusion (i). Multiphase CT angiography (three phases) maximum intensity
gen activator and intraarterial treatment projection images (ii, iii, iv) are indicative of one phase delay at worst, with similar extent and prominence
with each imaging modality. Detailed when compared with the contralateral side. These indicate of a score of 4 and suggest the patient should
results are described in Appendix E1 undergo treatment. C, CT perfusion Tmax and CBF maps (i, ii). A CBF-defined infarct core is 1 mL (iii, no blue
(online). For intravenous tissue plasmin- regions). A mismatch ratio of 106 and a small infarct core suggests the patient should undergo treatment.
Multiphase CT angiography and perfusion CT imaging are congruent for treatment decision. D, MR diffusion
ogen activator decision making, maxi-
images at 26 hours after admission show the final infarct as hyperintense. This patient did not attain recana-
mal agreement (92.5%, k = 0.68) was
lization with endovascular therapy.
seen between single- and multiphase CT
angiography. The next best agreement
was between unenhanced CT and multi- treatment decision, maximal agreement k = 0.46) and between multiphase CT
phase CT angiography (89.1%, k = 0.4) (89.8%, k = 0.8) was seen between sin- angiography and perfusion CT mismatch
and then between unenhanced CT and gle- and multiphase CT angiography. The ratio greater than 1.8 (72.1%, k = 0.45).
single-phase CT angiography (85.7%, k next best agreement was between multi- Agreement for all other pairs was 45%
= 0.41). Agreement for all other pairs phase CT angiography and perfusion CT or less. Figures 4–7 show various combi-
was 70.1% or less. For intraarterial mismatch ratio greater than 3 (72.5%, nations of congruence or incongruence

Radiology: Volume 000: Number 0—   2015  n  radiology.rsna.org 7

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

Figure 6 in determination of primary clinical out-

come. The model with baseline infarct
volume greater than or equal to 80 mL
versus that with baseline infarct volume
of less than 80 mL at perfusion CT is
next best; it is 0.41 times as probable to
minimize information loss as the model
with multiphase CT angiography. Other
models with diminishing probability of
minimizing information loss when com-
pared with the best model (multiphase
CT angiography) are as follows: sin-
gle-phase CT angiography (0.16 times);
mismatch ratio greater than 1.8 (0.06
times), and mismatch ratio greater
than 3 (0.061 times). Results with BIC
are similar to those with AIC (Table
4). Similar results were seen when we
compared models with receiver oper-
ating characteristic analysis, AIC, and
BIC, with an mRS score of 0–2 at 90
days as the clinical outcome (n = 102)
(Table 5).

Sensitivity Analyses
In sensitivity analyses restricted to pa-
tients with only intracranial artery, M1
MCA, or proximal M2 MCA occlusions,
the C statistic was highest with mul-
tiphase CT angiography (C statistic =
0.6; 95% CI: 0.54, 0.67). In sensitiv-
ity analyses restricted to the patients
who underwent revascularization ther-
apy (intravenous tissue plasminogen
activator 6 intraarterial therapy), the
Figure 6:  Multimodal CT images obtained 1 hour 28 minutes after symptom onset in a 78-year-old woman C statistic was again highest for mul-
with NIHSS of 18 and right hemisphere symptoms. A, Unenhanced CT ASPECTS score is 8. B, Proximal tiphase CT angiography (C statistic
right M1 MCA occlusion (i). Multiphase CT angiography (three phases) maximum intensity projection images = 0.57; 95% CI: 0.5, 0.65). Similarly,
(ii, iii, iv) are indicative of one phase delay, with similar extent and prominence when compared with the in sensitivity analyses restricted to pa-
contralateral side. These suggest a score of 4 and that the patient should undergo treatment. C, CT perfusion tients with early recanalization/reperfu-
Tmax and CBF maps (i, ii). A CBF-defined predicted infarct core is 113 mL (blue) and mismatch ratio (blue/ sion data, the C statistic was highest for
pink areas) (iii, iv) is 1.7; this indicates the patient should not undergo treatment. Multiphase CT angiography multiphase CT angiography (C statistic
and perfusion CT imaging are incongruent for treatment decision. D, MR diffusion images at 26 hours after = 0.57; 95% CI: 0.46, 0.67); other im-
admission show the final infarct as hyperintense. This M1 MCA clot recanalized with intraarterial therapy.
aging modalities had a lower C statistic.
Recanalization or reperfusion (TICI =
in clinical decision making between un- at 24 hours is described in Table 4. 2b/3), however, was the best predictor
enhanced CT, multiphase CT angiogra- The C statistic was highest for multi- of primary clinical outcome (C statistic
phy, and perfusion CT in our data. phase CT angiography (x2 test for model = 0.66; 95% CI: 0.54, 0.77) whenever
comparison, P = .007); nonetheless, those data were available.
Predictive Ability multiphase CT angiography has only
The C statistic for models using sin- modest discrimination, while the other
gle-phase CT angiography, multiphase imaging modalities fared worse. Model Discussion
CT angiography, and perfusion CT (with comparisons on the same data set with Multiphase CT angiography is a quick
mismatch ratios .1.2, .1.8, and .3.0 AIC and BIC are also described in Table and easy-to-use imaging tool in pa-
and infarct volume ,80 mL vs 80 mL) 4. AIC suggests that multiphase CT an- tients with AIS. Our study shows that
in determining a 50% decrease in NIHSS giography is the best imaging modality multiphase CT angiography has good

8 radiology.rsna.org  n Radiology: Volume 000: Number 0—   2015

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

Figure 7 infarct core and penumbra can be used

to select patients for intravenous tissue
plasminogen activator or endovascular
therapy 3 hours after onset of stroke
symptoms (27,30). Perfusion CT, how-
ever, requires 10–20 minutes from image
acquisition to interpretation and needs al-
gorithms for postprocessing images that
are vendor specific, not standardized, and
therefore variable across centers. Perfu-
sion CT also needs trained personnel to
process these images (16,31,32). In ad-
dition, image quality is affected to some
extent by patient motion (17). Additional
radiation dose is also a concern (21). By
acquiring temporal information at three
data points, multiphase CT angiography
is conceptually similar to perfusion CT
(and dynamic CT angiography that is de-
rived from perfusion CT images) (18,19).
Differences from perfusion CT are there-
fore in using less information and avoid-
ing the need for postprocessing. Of note,
our study shows that currently available
perfusion CT thresholds for infarct or
penumbra are not better than multiphase
CT angiography in clinical decision mak-
ing or outcome prediction. A possible ex-
planation could be that these externally
validated thresholds are not internally
valid within our own data set and that
we need to derive our own thresholds for
infarct core and penumbra (17). We plan
to derive such thresholds. Nonetheless,
the fact that neither we nor our vendors
Figure 7:  Multimodal CT images obtained 1 hour 32 minutes after symptom onset in a 67-year-old man currently have validated thresholds from
with NIHSS of 17 and right hemisphere symptoms. A, Unenhanced CT ASPECTS score is 7. B, A proxi- literature that we can apply prospectively
mal right M1 MCA occlusion extending to M2 MCA is seen (i). Multiphase CT angiography (three phases) to our data set is an inherent limitation
maximum intensity projection images (ii, iii, and iv ) are indicative of, at worst, one phase delay but with large to the widespread use of perfusion CT
regions having minimal pial arterial filling when compared with the contralateral side. The multiphase CT in the real world. However, automated
angiography score is 3, which indicates no treatment should be performed. C, Perfusion CT Tmax and CBF perfusion-based algorithms now available
maps (i, ii). A CBF-defined predicted infarct core is 19 mL (iii, iv). A mismatch ratio of 7.1 indicates treatment are capable of providing information to
should be performed. Multiphase CT angiography and CT perfusion images are incongruent for treatment clinicians in a rapid manner like multi-
decision. Fast (,60 minutes) recanalization was achieved by using endovascular thrombectomy. D, Unen- phase CT angiography (33).
hanced CT 3 days after admission shows the final infarct as hypointense. Unenhanced CT is widely used
for patient selection. Unenhanced CT,
interrater reliability. It reduces uncer- There is currently no reference however, has moderate interrater re-
tainty in clinical decision making and standard for imaging selection in pa- liability even among experts (10–13).
may be slightly better in the prediction tients with AIS. Perfusion CT, however, Single-phase CT angiography lacks
of clinical outcome than currently used is used in many centers for patient se- temporal resolution; therefore, this
techniques, such as unenhanced CT, lection. Clinical trials Echoplanar Im- modality leads to risk of mislabeling
single-phase CT angiography, and per- aging Thrombolytic Evaluation Trial pial arterial filling when compared with
fusion CT. Other advantages include (or EPITHET) and Diffusion Weighted multiphase CT angiography (18,19)
minimal additional radiation, no addi- Imaging Evaluation for Understanding (Fig 3). Unlike contrast-enhanced CT,
tional contrast material, whole-brain Stroke Evolution Study-2 (or DEFUSE-2) multiphase CT angiography provides
coverage, and no postprocessing. have shown that a strategy of delineating clinicians with three time-resolved

Radiology: Volume 000: Number 0—   2015  n  radiology.rsna.org 9

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

Table 4
Ability of Each Imaging Modality to Discriminate Clinical Outcome (50% or More Decrease in NIHSS from Baseline to 24 Hours) Using
Logistic Regression Analysis Receiver Operating Curve Analysis, AIC, and BIC
No. of
Imaging Modality and Criteria Patients Odds Ratio P Value C Statistic AIC BIC

Multiphase CT angiography (.3 vs 3) 126 4.3 (1.2, 15.7) .02 0.58 166.0 171.7
Single-phase CT angiography (.2 vs 2) 126 2.8 (0.75, 10.6) .12 0.55 169.6 175.4
Baseline infarct volume (,80 mL vs 80 mL) 126 5.2 (1.1, 27.1) .05 0.45 167.8 173.5
Mismatch ratio (1.8 vs .1.8) 126 3.2 (0.3, 36.6) .34 0.49 171.4 177.1
Mismatch ratio (3 vs .3) 126 1.6 (0.6, 4.8) .33 0.47 171.5 177.1

Note.—Data in parentheses are 95% CIs. Higher C statistics imply better models, whereas lower AIC and BIC values imply better models.

Table 5
Ability of Each Imaging Modality to Discriminate Clinical Outcome (mRS 0–2 at 90 Days) Using Logistic Regression Analysis Receiver
Operating Curve Analysis, AIC, and BIC
No. of
Imaging Modality and Criteria Patients Odds Ratio P Value C Statistic AIC BIC

Multiphase CT angiography (.3 versus 3) 102 5.5 (1.6, 18.8) .01 0.60 132.1 137.4
single-phase CT angiography (.2 versus 2) 102 3.3 (1.0, 10.7) .05 0.57 136.4 141.7
Baseline Infarct volume ,80 mL versus 80 mL 102 1.2 (0.25, 5.5) .84 0.50 140.6 145.8
Mismatch ratio 1.8 versus .1.8 102 0.8 (0.1, 8.8) .80 0.50 140.6 145.8
Mismatch ratio 3 versus .3 102 1.4 (0.4, 4.5) .58 0.52 140.3 145.6

Note.—Data in parentheses are 95% CIs. Higher C statistics imply better models, whereas lower AIC and BIC values imply better models.

images and therefore a more nuanced Even though we did not find any such Acknowledgments: We acknowledge Ting Y.
assessment of pial arterial filling in case in the current study, this possibility Lee and Richard Frayne for their expertise and
guidance during image and data analysis. We
both the normal brain and the ische- cannot be discounted. Thus, we recom- also acknowledge the Calgary Stroke Program
mic brain. An example is the ability of mend that multiphase CT angiography fellows, nurses, and staff, including the research
multiphase CT angiography to enable images always be interpreted in con- office, for helping with the conduct, data collec-
discrimination between a one- and two- junction with head and neck CT angiog- tion, and execution of the study. We acknowl-
edge the Seaman MR Center and the Hotchkiss
phase delay whereas contrast-enhanced raphy images. Poor cardiac function can
Brain Institute for their support.
CT labels both the same. Finally, when also interfere with pial arterial filling,
compared with multiphase CT angiog- even though our data did not show this. Disclosures of Conflicts of Interest: B.K.M.
raphy, MR imaging has practical draw- A protocol that includes an additional disclosed no relevant relationships. C.D.d.E.
backs. MR imaging takes up to 30 mi- delayed fourth phase may help in such disclosed no relevant relationships. E.M.Q. dis-
closed no relevant relationships. M.A. disclosed
nutes to screen patients, perform the scenarios. Finally, multiphase CT angi-
no relevant relationships. L.H. disclosed no rele-
examination, and interpret the results ography cannot as yet be used in pa- vant relationships. A.M.D. disclosed no relevant
(16). Many patients do not tolerate it tients with posterior circulation stroke, relationships.M.G. Activities related to the pre-
well, and image quality is affected by except when involving the PCA, because sent article: received an honorarium from Covi-
patient motion. MR imaging also has of poorly understood collateral hemody- dien, has a licensing agreement with GE Health-
care, institution received a grant from Covidien.
limited availability after working hours namics of the posterior circulation. Activities not related to the present article: owns
(34). In summary, we describe multi- shares in Calgary Scientific. Other relationships:
Our tool, multiphase CT angiogra- phase CT angiography, an imaging tool none to disclose.
phy, has limitations. The presence of for clinical decision making in patients
flow-limiting proximal stenosis and cir- with AIS. In this article, we have shown
cuitous base-of-skull collaterals can re- its reliability and ability to help predict References
sult in delay in contrast material filling clinical outcome. Larger studies are 1. Menon BK, Demchuk AM. Is acute stroke
of the pial arteries, even in the healthy needed to conclusively demonstrate treatment heading towards a more endo-
hemisphere, thus potentially leading to its utility in triage and clinical decision vascular approach? Lancet Neurol 2009;
mislabeling of pial arterial filling status. making. 8(9):778–779.

10 radiology.rsna.org  n Radiology: Volume 000: Number 0—   2015

NEURORADIOLOGY: Multiphase CT Angiography Menon et al

2. Saver JL, Jahan R, Levy EI, et al. Soli- 13. Coutts SB, Hill MD, Demchuk AM, Barber 24. Campbell BC, Christensen S, Levi CR, et al.
taire flow restoration device versus the PA, Pexman JH, Buchan AM. ASPECTS read- Cerebral blood flow is the optimal CT per-
Merci Retriever in patients with acute ing requires training and experience. Stroke fusion parameter for assessing infarct core.
ischaemic stroke (SWIFT): a randomised, 2003;34(10):e179; author reply e179. Stroke 2011;42(12):3435–3440.
parallel-group, non-inferiority trial. Lancet
14. Bal S, Bhatia R, Menon BK, et al. Time
25. Bivard A, Levi C, Spratt N, Parsons M. Per-
2012;380(9849):1241–1249.
dependence of reliability of noncontrast fusion CT in acute stroke: a comprehensive
3. Nogueira RG, Lutsep HL, Gupta R, et al. computed tomography in comparison to analysis of infarct and penumbra. Radiology
Trevo versus Merci retrievers for thrombec- computed tomography angiography source 2013;267(2):543–550.
tomy revascularisation of large vessel occlu- image in acute ischemic stroke. Int J Stroke
26. Yoo AJ, Verduzco LA, Schaefer PW, Hirsch
sions in acute ischaemic stroke (TREVO 2): 2015;10(1):55–60.
JA, Rabinov JD, González RG. MRI-based
a randomised trial. Lancet 2012;380(9849):
15. Menon BK, Smith EE, Modi J, et al. Re-
selection for intra-arterial stroke therapy:
1231–1240.
gional leptomeningeal score on CT angiogra- value of pretreatment diffusion-weighted
4. Menon BK, Hill MD, Eesa M, et al. Ini- phy predicts clinical and imaging outcomes imaging lesion volume in selecting patients
tial experience with the Penumbra Stroke in patients with acute anterior circula- with acute stroke who will benefit from early
System for recanalization of large vessel tion occlusions. AJNR Am J Neuroradiol recanalization. Stroke 2009;40(6):2046–
occlusions in acute ischemic stroke. Neuro- 2011;32(9):1640–1645. 2054.
radiology 2011;53(4):261–266.
16. Sheth KN, Terry JB, Nogueira RG, et al.
27. Lansberg MG, Straka M, Kemp S, et al.

5. Goyal M, Menon BK, Coutts SB, Hill MD, Advanced modality imaging evaluation in MRI profile and response to endovascular
Demchuk AM; Penumbra Pivotal Stroke acute ischemic stroke may lead to delayed reperfusion after stroke (DEFUSE 2): a pro-
Trial Investigators, Calgary Stroke Program, endovascular reperfusion therapy without spective cohort study. Lancet Neurol 2012;
and the Seaman MR Research Center. Ef- improvement in clinical outcomes. J Neuro- 11(10):860–867.
fect of baseline CT scan appearance and interv Surg 2013;5(Suppl 1):i62–i65.
28. Gönen M. Analyzing receiver operating

time to recanalization on clinical outcomes
17. Goyal M, Menon BK, Derdeyn CP. Perfusion characteristic curves using SAS. Cary, NC:
in endovascular thrombectomy of acute is-
imaging in acute ischemic stroke: let us im- SAS Press, 2007.
chemic strokes. Stroke 2011;42(1):93–97.
prove the science before changing clinical
29. Adams NM, Hand DJ. Improving the prac-
6. Menon BK, Goyal M. Endovascular therapy practice. Radiology 2013;266(1):16–21.
tice of classifier performance assessment.
in acute ischemic stroke: where we are, the
18. Menon BK, O’Brien B, Bivard A, et al. As- Neural Comput 2000;12(2):305–311.
challenges we face and what the future holds.
sessment of leptomeningeal collaterals using
Expert Rev Cardiovasc Ther 2011;9(4): 30. Nagakane Y, Christensen S, Brekenfeld C, et
dynamic CT angiography in patients with
473–484. al. EPITHET: positive result after reanalysis
acute ischemic stroke. J Cereb Blood Flow
using baseline diffusion-weighted imaging/
7. Broderick JP, Schroth G. What the SWIFT Metab 2013;33(3):365–371.
perfusion-weighted imaging co-registration.
and TREVO II trials tell us about the role
19. Frölich AM, Wolff SL, Psychogios MN, et al. Stroke 2011;42(1):59–64.
of endovascular therapy for acute stroke.
Time-resolved assessment of collateral flow
Stroke 2013;44(6):1761–1764. 31. Kudo K, Sasaki M, Yamada K, et al. Dif-
using 4D CT angiography in large-vessel oc-
ferences in CT perfusion maps generated
8. Rha JH, Saver JL. The impact of recanali- clusion stroke. Eur Radiol 2014;24(2):390–
by different commercial software: quan-
zation on ischemic stroke outcome: a meta- 396.
titative analysis by using identical source
analysis. Stroke 2007;38(3):967–973.
20. Latchaw RE, Alberts MJ, Lev MH, et al.
data of acute stroke patients. Radiology
9. Menon BK, Almekhlafi MA, Pereira VM, et Recommendations for imaging of acute is- 2010;254(1):200–209.
al. Optimal workflow and process-based per- chemic stroke: a scientific statement from
32. Zussman BM, Boghosian G, Gorniak RJ,

formance measures for endovascular therapy the American Heart Association. Stroke
et al. The relative effect of vendor vari-
in acute ischemic stroke: analysis of the Soli- 2009;40(11):3646–3678.
ability in CT perfusion results: a method
taire FR thrombectomy for acute revascular-

21. Mnyusiwalla A, Aviv RI, Symons SP. Ra- comparison study. AJR Am J Roentgenol
ization study. Stroke 2014;45(7):2024–2029.
diation dose from multidetector row CT 2011;197(2):468–473.
10. Gupta AC, Schaefer PW, Chaudhry ZA, et imaging for acute stroke. Neuroradiology
33. Lansberg MG, Lee J, Christensen S, et al.
al. Interobserver reliability of baseline non- 2009;51(10):635–640.
RAPID automated patient selection for re-
contrast CT Alberta Stroke Program Early
22. Authors on behalf of ICRP, Stewart FA, Ak- perfusion therapy: a pooled analysis of the
CT Score for intra-arterial stroke treat-
leyev AV, et al. ICRP publication 118: ICRP Echoplanar Imaging Thrombolytic Evalua-
ment selection. AJNR Am J Neuroradiol
statement on tissue reactions and early tion Trial (EPITHET) and the Diffusion and
2012;33(6):1046–1049.
and late effects of radiation in normal tis- Perfusion Imaging Evaluation for Under-
11. Wardlaw JM, Mielke O. Early signs of brain sues and organs—threshold doses for tissue standing Stroke Evolution (DEFUSE) Study.
infarction at CT: observer reliability and out- reactions in a radiation protection context. Stroke 2011;42(6):1608–1614.
come after thrombolytic treatment—system- Ann ICRP 2012;41(1-2):1–322.
34. Hand PJ, Wardlaw JM, Rowat AM, Ha-

atic review. Radiology 2005;235(2):444–453.
23. Adatia S, Menon BK, Sohn SI et al. A novel isma JA, Lindley RI, Dennis MS. Magnetic
12. Patel SC, Levine SR, Tilley BC, et al. Lack of CT-angiogram based recanalization grading resonance brain imaging in patients with
clinical significance of early ischemic chang- system (TICI-CTA) correlates with early acute stroke: feasibility and patient related
es on computed tomography in acute stroke. neurologic response after intravenous tPA difficulties. J Neurol Neurosurg Psychiatry
JAMA 2001;286(22):2830–2838. therapy. Stroke. 2013;44:AWP52. 2005;76(11):1525–1527.

Radiology: Volume 000: Number 0—   2015  n  radiology.rsna.org 11