This copy is for personal use only. To order printed copies, contact reprints@rsna.

org

Original Research  n  Neuroradiology

Diagnostic Accuracy of
Simulated Low-Dose Perfusion
CT to Detect Cerebral Perfusion
Impairment after Aneurysmal
Subarachnoid Hemorrhage: A
Retrospective Analysis1
Saif Afat, MD To evaluate diagnostic accuracy of low-dose volume perfusion
Purpose:
Carolin Brockmann, MD (VP) computed tomography (CT) compared with original VP CT
Omid Nikoubashman, MD regarding the detection of cerebral perfusion impairment after
Marguerite Müller, MD aneurysmal subarachnoid hemorrhage.
Kolja M. Thierfelder, MD
In this retrospective study, 85 patients (mean age, 59.6 years;
Marc A. Brockmann, MD, MSc Materials and
62 women) with aneurysmal subarachnoid hemorrhage and who
Konstantin Nikolaou, MD Methods:
were suspected of having cerebral vasospasm at unenhanced CT
Martin Wiesmann, MD and VP CT (tube voltage, 80 kVp; tube current–time product,
Jong Hyo Kim, PhD 180 mAs) were included, 37 of whom underwent digital sub-
Ahmed E. Othman, MD traction angiography (DSA) within 6 hours. Low-dose VP CT
data sets at tube current–time product of 72 mAs were retro-
spectively generated by validated realistic simulation. Perfusion
maps were generated from both data sets and reviewed by two
neuroradiologists for overall image quality, diagnostic confidence
and presence and/or severity of perfusion impairment indicat-
ing vasospasm. An interventional neuroradiologist evaluated 16
vascular segments at DSA. Diagnostic accuracy of low-dose VP
CT was calculated with original VP CT as reference standard.
Agreement between findings of both data sets was assessed by
1
 From the Department of Diagnostic and Interventional using weighted Cohen k and findings were correlated with DSA
Neuroradiology, RWTH Aachen University, Aachen, Germany by using Spearman correlation. After quantitative volumetric
(S.A., O.N., M.M., M.W., A.E.O.); Department for Diagnostic analysis, lesion volumes were compared on both VP CT data sets.
and Interventional Radiology, Eberhard Karls University
Tuebingen, University Hospital Tuebingen, Hoppe-Seyler- Low-dose VP CT yielded good ratings of image quality and di-
Strasse 3, 72076 Tübingen, Germany (S.A., K.N., A.E.O.); Results:
agnostic confidence and classified all patients correctly with
Department of Neuroradiology, University Hospital Mainz,
high diagnostic accuracy (sensitivity, 99.0%; specificity, 99.5%)
Mainz, Germany (M.A.B., C.B.); Institute for Clinical
Radiology, Ludwig-Maximilian-University Hospital Munich,
without significant differences regarding presence and/or se-
Munich, Germany (K.M.T.); Department of Transdisciplinary verity of perfusion impairment between original and low-dose
Studies, Graduate School of Convergence Science and data sets (Z = 20.447; P = .655). Findings of both data sets
Technology, Seoul National University, Suwon, South Korea correlated significantly with DSA (original, r = 0.671; low dose,
(J.H.K.); Department of Radiology, Seoul National University r = 0.667). Lesion volume was comparable for both data sets
College of Medicine, Seoul, South Korea (J.H.K.); and
(relative difference, 5.9% 6 5.1 [range, 0.2%–25.0%; median,
Center for Medical-IT Convergence Technology Research,
4.0%]) with strong correlation (r = 0.955).
Advanced Institute of Convergence Technology, Suwon,
South Korea (J.H.K.). From the 2016 RSNA Annual Meeting.
Conclusion: The results suggest that radiation dose reduction to 40% of
Received December 5, 2016; revision requested January
30, 2017; revision received September 24; accepted original dose levels (tube current–time product, 72 mAs) may
October 24; final version accepted November 3. Address be performed in VP CT imaging of patients with aneurysmal
correspondence to A.E.O. (e-mail: ahmed.e.othman@ subarachnoid hemorrhage without compromising the diagnostic
googlemail.com). accuracy regarding detection of cerebral perfusion impairment
Study supported by Korea government (MSIP) (2017-0- indicating vasospasm.
01329) and the Ministry of Trade, Industry and Energy
(MOTIE) of Korea (10051357).
q
 RSNA, 2018

q
 RSNA, 2018 Online supplemental material is available for this article.

Radiology: Volume 287: Number 2—May 2018  n  radiology.rsna.org 643

NEURORADIOLOGY: Simulated Low-Dose Perfusion CT to Detect Cerebral Perfusion Impairment Afat et al

A
neurysmal subarachnoid hem- been evaluated for various body parts [mean age, 58.2 years]; significant sex
orrhage is a relevant cause of including brain imaging (12,13). Despite differences [P = .048]). For additional
mortality and morbidity (1). Half the high radiation dose, cerebral VP CT characteristics, see the Table.
of the patients who survive the initial received little attention. Few technical Thirty-seven patients suspected of
phase suffer from persistent neurologic studies showed that reduced dose VP CT having severe angiographic vasospasm
deficits (1–3), mainly caused by delayed with low tube-current time product might underwent DSA within 6 hours after VP
cerebral ischemia, which is strongly yield sufficient image quality for the diag- CT, forming a subgroup (see Standards
associated with cerebral vasospasm. nosis of acute stroke (13–16). However, for Reporting of Diagnostic Accuracy
Early aggressive management leads to perfusion impairment in patients with flowchart; Fig 1).
improved functional outcome (1). Ce- cerebral vasospasm is often more subtle,
rebral vasospasm usually manifests as detectability might be compromised at Data Acquisition
a new neurologic deficit or decrease in lower dose levels (17). Dedicated evalu- Noncontrast CT and VP CT imaging was
consciousness (1,3,4). Patients with an- ation of diagnostic accuracy is crucial be- performed on a 40-section CT imager
eurysmal subarachnoid hemorrhage are fore clinical application. We hypothesize (Somatom AS; Siemens Healthineers,
under intensive neurologic surveillance, that low-dose VP CT yields similar results Erlangen, Germany). Noncontrast CT
oftentimes requiring sedation, poten- to original VP CT with high diagnostic was performed by using a sequential
tially delaying detection, and resulting confidence and high diagnostic accuracy scan (tube current–time product, 385
in poor functional outcome (5). for the detection of cerebral perfusion mAs, fixed; tube voltage, 120 kVp; vol-
Device-related monitoring methods impairment. In this study, we aimed to ume CT dose index, 64.1 mGy).
include transcranial Doppler, digital sub- evaluate the diagnostic accuracy of low- VP CT images were acquired with a
traction angiography (DSA) and volume dose VP CT compared with original VP 40-mL intravenous contrast bolus injec-
perfusion (VP) computed tomography CT regarding the detection of cerebral tion at a flow rate of 5.0 mL/sec followed
(CT) (1). Vasospasm occurs at multiple perfusion impairment after aneurysmal by a saline flush (84.0-mm scan length;
levels, affecting large arteries and small subarachnoid hemorrhage. tube current–time product, 180 mAs;
arterioles (6). Macroscopic vasospasm, tube voltage, 80 kVp; acquisition dura-
appearing as large artery narrowing at tion, 45 seconds; temporal resolution,
DSA or increased blood velocity at tran- Materials and Methods 1.5 seconds; section thickness, 10.0
scranial Doppler, represents only 50% The local institutional ethics committee mm; volume CT dose index, 301.2 mGy;
in delayed cerebral ischemia (1,6). Mi- approved this retrospective study and dose-length product, 2702 mGy ∙ cm).
crovascular vasospasm at the arterio- waived informed patient consent. Three- or four-vessel DSA exam-
lar level is more relevant for prediction inations were performed by using a
of delayed cerebral ischemia (1,3,6). Patient Characteristics biplane flat-panel angiography system
Therefore, regional perfusion impair- Between January 2012 and December (Artis Zee; Siemens Healthineers).
ment as observed at VP CT seems to be 2014, we identified 111 consecutive pa-
more accurate for the detection of rele- tients who underwent any brain imaging
vant perfusion impairment and delayed (ie, noncontrast agent–enhanced CT, VP
cerebral ischemia (1). CT, or DSA) in the context of vasospasm. https://doi.org/10.1148/radiol.2017162707
VP CT is acquired with continuous Patients suspected of having vasospasm Content code:
gantry rotation during intravenous injec- at our institution usually undergo non-
Radiology 2018; 287:643–650
tion of iodinated contrast material with contrast CT and VP CT. If substantial
an acquisition duration of around 45 sec- vasospasm (relevant for therapy) were Abbreviations:
onds (7,8), causing high radiation doses suspected, patients would undergo addi- DSA = digital subtraction angiography
(9,10). The U.S. Food and Drug Admin- tional DSA. All 111 patients underwent at VP = volume perfusion

istration raised concerns and called for least noncontrast CT for suspected vaso- Author contributions:
alternative reduced-dose protocols (10), spasm and 87 patients underwent both Guarantors of integrity of entire study, S.A., A.E.O.;
making dose reduction a central subject noncontrast CT and VP CT. The remain- study concepts/study design or data acquisition or data
of radiologic research (11). Because of ing 24 patients who did not undergo VP analysis/interpretation, all authors; manuscript drafting or
manuscript revision for important intellectual content, all
young age and multiple imaging examina- CT were either already undergoing max-
authors; approval of final version of submitted manuscript,
tions, cumulative radiation risk is high for imal vasospasm therapy or had specific all authors; agrees to ensure any questions related to the
patients with aneurysmal subarachnoid contraindications for VP CT. work are appropriately resolved, all authors; literature
hemorrhage at risk of perfusion impair- Of the 87 patients with available VP research, S.A., C.B., O.N., M.M., K.T., M.A.B., J.H.K., A.E.O.;
ment and delayed cerebral ischemia (1). CT, patients with severe motion artifacts clinical studies, S.A., C.B., M.W., A.E.O.; experimental
Acquisition (improved detector technol- were excluded (n = 2), resulting in a final studies, J.H.K., A.E.O.; statistical analysis, S.A., M.A.B.,
A.E.O.; and manuscript editing, S.A., C.B., O.N., M.M., K.T.,
ogies, low tube voltage, low tube current sample size of 85 patients (mean age,
M.A.B., K.N., M.W., A.E.O.
imaging) and image reconstruction tech- 59.6 years; age range, 34–86 years; 23
niques (iterative reconstruction) have men [mean age, 63.5 years]; 62 women Conflicts of interest are listed at the end of this article.

644 radiology.rsna.org  n Radiology: Volume 287: Number 2—May 2018

NEURORADIOLOGY: Simulated Low-Dose Perfusion CT to Detect Cerebral Perfusion Impairment Afat et al

50–80 mAs produces acceptable image

quality (14–16). We therefore gener-
Patient Characteristics ated data sets with 40% of the original
All Eligible Patients Included Patients Subgroup with Available milliampere-seconds (tube current–
Parameter (n = 111) (n = 85) DSA (n = 37) time product, 72 mAs), resulting in
40% of the original dose (CT dose in-
Mean age (y)* 58.3 (20–86) 59.6 (34–86) 60.3 (34–77) dex, 120.5 mGy; dose-length product,
Mean time after aneurysmal 6.2 (1–17) 6.4 (1–17) 6.6 (3–13)
1080 mGy ∙ cm).
subarachnoid hemorrhage (d)*
Sex Data Postprocessing
 Male 27 (24.3) 23 (27.1) 12 (32.4)
VP CT image postprocessing was
 Female 84 (75.7) 62 (72.9) 25 (67.6)
performed with a commercial per-
  Total no. of patients 111 (100) 85 (100) 37 (100)
Aneurysm location
fusion package (VPCT Neuro; Sie-
  Anterior communicating artery 45 35 19 mens Healthineers). Cerebral blood
  Middle cerebral artery 33 25 9 flow, cerebral blood volume, mean
  Basilar artery 10 7 4 transit time, and time-to-drain maps
  Posterior communicating artery 6 6 2 were generated by using a deconvo-
  Internal carotid artery 26 11 3 lution-based algorithm (Appendix E1
 Other 7 5 0 [online]).
Initial Hunt and Hess scale score
 I 22 (19.8) 17 (20) 7 (18.9) Qualitative Analyses of Perfusion Maps
 II 13 (11.7) 12 (14.1) 10 (27) and Noncontrast CT Images
 III 35 (31.5) 25 (29.4) 12 (32.4) Two radiologists (A.E.O. and S.A.,
 IV 30 (27) 21 (24.7) 7 (18.9) with 5 and 2 years of experience in di-
 V 11 (9.9) 10 (11.8) 1 (2.7) agnostic neuroradiology, respectively)
Modified Rankin scale score were blinded for clinical data, radia-
at discharge tion dose, findings at DSA, and final
 0 10 (9) 6 (7.1) 3 (8.1) diagnosis, and separately assessed
 1 8 (7.2) 5 (5.9) 0 (0)
original and low-dose perfusion maps
 2 10 (9) 7 (8.2) 0 (0)
regarding overall image quality and di-
 3 13 (11.7) 12 (14.1) 7 (18.9)
agnostic confidence on a five-point Lik-
 4 35 (31.5) 27 (31.8) 11 (29.7)
ert scale (Appendix E1 [online]). They
 5 5 (4.5) 3 (3.5) 2 (5.4)
 6 30 (27) 25 (29.4) 14 (37.8)
visually evaluated anterior cerebral
Modified Rankin scale
artery, middle cerebral artery, and
score 3 months after posterior cerebral artery territories
discharge (n = 62) on the perfusion maps (six territories
 0 21 (25.9) 13 (21.7) 6 (26.1) per patient) regarding presence or ab-
 1 14 (17.3) 13 (21.7) 6 (26.1) sence of perfusion impairment that in-
 2 8 (9.9) 6 (10) 2 (8.7) dicated vasospasm.
 3 28 (34.6) 20 (33.3) 7 (30.4) Because temporal perfusion maps
 4 9 (11.1) 7 (11.7) 2 (8.7) (particularly mean transit time) are
 5 1 (1.2) 1 (1.7) 0 (0) most accurate for assessment of vaso-
 6 0 (0) 0 (0) 0 () spasm, territories were evaluated on
Aneurysm treatment mean transit time maps with a three-
  No treatment 11 (9.9) 8 (9.4) 0 (0) point Likert scale: score of 0, no im-
 Coiling 61 (55) 45 (52.9) 22 (59.5) pairment; score of 1, impairment af-
 Clipping 39 (35.1) 32 (37) 15 (40.5) fecting less than 50% of territory; and
Note.—Unless otherwise indicated, data are number of patients and data in parentheses are percentages.
score of 2, impairment affecting 50%
* Data in parentheses are range. or more of territory (Fig E1 [online]).
Discordant findings were assessed in
consensus.
Low-Dose Simulation reconstruction to generate realistic To compare perfusion maps with
Realistic reduced-dose CT simulation noise patterns on low-dose VP CT im- DSA in the subgroup, each vascular seg-
was applied on original source VP ages by using a validated model (18). ment on DSA images (16 segments per
CT images. The simulation is based Previous studies showed that dose re- patient) was assigned to a predefined
on sinogram synthesis and image duction to tube current–time product territory (14 segments per patient) on

Radiology: Volume 287: Number 2—May 2018  n  radiology.rsna.org 645

NEURORADIOLOGY: Simulated Low-Dose Perfusion CT to Detect Cerebral Perfusion Impairment Afat et al

Figure 1 present impairment [positive]). Terri-

tories were considered to be true-pos-
itive findings when readers correctly
identified any perfusion impairment.
Segment-based sensitivity and speci-
ficity of low-dose VP CT for detection
of vasospasm-related perfusion abnor-
malities were calculated with original
VP CT as standard of reference. To
account for clustered data, 95% con-
fidence intervals were adjusted by us-
ing a variance inflation factor (21,22).
Findings regarding presence and se-
verity of vasospasm were compared
by using Wilcoxon signed-rank tests.
Weighted Cohen k value was calcu-
lated for interrater agreement.
In the subgroup, segment-based
Spearman correlation coefficients were
calculated for DSA and both VP CT find-
ings and compared by using Fisher r-to-z
transformation. For quantitative analyses,
we calculated absolute and relative mea-
surement errors as a measure of accu-
racy of low-dose data sets. Measurement
errors were defined as the difference be-
tween the measured affected volume on
low-dose VP CT maps and original VP CT
maps (Appendix E1 [online]).
Interclass correlation coefficient was
Figure 1:  Standards for Reporting of Diagnostic Accuracy flowchart of patient inclusion. DSA = digital calculated for interreader agreement.
subtraction angiography; NCT = noncontrast CT; VPCT = vascular perfusion CT. Spearman correlation coefficient was
calculated between volumes on both data
sets and between lesion size and relative
mean transit time maps (Appendix E1 The two blinded readers (A.E.O. measurement error. The latter aimed to
[online]). and S.A.) performed segmentation of detect possible effects of lesion size on
Territories with demarcated cere- the mean transit time maps by using measurement accuracy. To account for
bral infarction on noncontrast CT images an open-source Digital Imaging and the clustered data, we used the boot-
(rated in consensus) were excluded. Communications in Medicine software strap approach with patients as clusters
(OsiriX V.7.3; Pixmeo, Geneva, Swit- (sampling with replacement, 1000 rep-
Evaluation of Cerebral Angiography zerland). The volumetric segmentation lications) and calculated bias-corrected
Images was performed by using the closed 95% confidence intervals. P values less
An interventional neuroradiologist polygon region of interest similar to than .05 were considered to indicate sta-
with (C.B., with 5 years of experience) Thierfelder et al (20) (Fig E2 [online]). tistically significant differences.
evaluated the 16 segments regarding
presence or absence and severity of va- Statistical Analyses
sospasm on a three-point Likert scale Statistical analyses were performed by Results
(Appendix E1 [online]). using software (SPSS version 22; IBM,
Armonk, NY). Overall image quality Overall Image Quality and Diagnostic
Quantitative Analyses of Perfusion Maps and diagnostic confidence of VP CT Confidence
Because of the variety of reported data sets were compared by using Wil- Both readers rated the overall image
thresholds and the lack of a definite coxon signed-rank tests. quality of original data sets as excel-
quantitative threshold for vasospasm on Findings regarding presence of lent (median score, 5; range, 3–5) and
perfusion maps (19), we performed vol- perfusion abnormalities and vaso- of low-dose data sets as good (median
umetric analysis of impaired regions in spasm were transformed into binary score, 4; range, 3–5) with significant
each territory. data (0, no impairment [negative]; 1, differences between both data sets

646 radiology.rsna.org  n Radiology: Volume 287: Number 2—May 2018

NEURORADIOLOGY: Simulated Low-Dose Perfusion CT to Detect Cerebral Perfusion Impairment Afat et al

Figure 2

Figure 2:  Original and low-dose vascular perfusion ( VP) CT ( VPCT ) maps in a 76-year-old male patient with cerebral vasospasm 6 days after aneurysmal sub-
arachnoid hemorrhage originating from a basilar artery aneurysm. Areas of hypoperfusion in both hemispheres indicating cerebral vasospasm can be identified on
both original and low-dose VP CT images. CBV = cerebral blood volume; CBF = cerebral blood flow; MT T = mean transit time; T TD = time to drain.

(P , .001) and substantial interrater in whom infarcts were detected on non- No significant differences regarding
agreement (k  0.771). Diagnostic contrast CT images (distribution: one presence and severity of perfusion im-
confidence of both data sets was rated territory in 12 patients, two territories pairment were detected between origi-
as excellent by both readers (original: in one patient, and four territories in nal and low-dose data sets (Z = 20.447;
median score, 5 [range, 4–5]; low-dose: one patient). These territories were ex- P = .655). Interreader agreement was al-
median score, 5 [range, 4–5]) without cluded from further analyses. most perfect for both original (k = 0.971;
significant differences between both After exclusion of the correspond- bias-corrected 95% confidence interval:
data sets (P = .083). The interreader ing territories, 492 territories on VP 0.942, 0.994) and low-dose data sets (k
agreement regarding diagnostic confi- CT images were included. On origi- = 0.983; bias-corrected 95% confidence
dence was substantial (k  0.799). nal VP CT data sets, 101 territories interval: 0.961, 1.00).
(20.5%) showed perfusion abnor-
Prevalence and Detectability of malities that indicated cerebral va- Subgroup Analyses
Vasospasm sospasm. Low-dose data sets helped In the patient-based subgroup analysis
Patient-based analysis.—Of the 85 pa- to identify 100 of the 101 positive of the 37 patients in whom DSA was
tients, 41 patients (48.2%) were negative territories correctly and yielded one available within 6 hours, low-dose VP
for vasospasm at both original VP CT and false-negative territory (ie, negative CT data sets were 100% concordant
angiography. In 14 patients, infarcts were for perfusion impairment on low-dose to original VP CT data sets and to
detected on noncontrast CT images. perfusion maps). Among the 391 neg- DSA regarding the prevalence of vaso-
No discordant findings were ob- ative territories, low-dose data sets spasm in cases without false-positive
served between the low-dose VP CT revealed 389 true-negative and two or false-negative findings (Appendix E1
data sets (100% agreement). Image false-positive findings. This resulted in [online]).
examples are given in Figures 2 and 3. high diagnostic accuracy with a sen- Both original and low-dose perfu-
Segment-based analysis.—The seg- sitivity of 99.0% (adjusted 95% con- sion data showed significant levels of
ment-based analysis of noncontrast CT fidence interval: 97.1%, 100%) and correlations with DSA regarding pres-
data sets showed 18 territories with a specificity of 99.5% (adjusted 95% ence and severity of vasospasm (orig-
demarcated infarcts in the 14 patients confidence interval: 98.8%, 100%). inal: r = 0.671 [bias-corrected 95%

Radiology: Volume 287: Number 2—May 2018  n  radiology.rsna.org 647

NEURORADIOLOGY: Simulated Low-Dose Perfusion CT to Detect Cerebral Perfusion Impairment Afat et al

Figure 3

Figure 3:  Original and low-dose vascular perfusion ( VP) CT ( VPCT ) maps in a 67-year-old female patient clinically suspected of having cerebral vasospasm 5
days after aneurysmal subarachnoid hemorrhage originating from a basilar artery aneurysm. Original and low-dose VP CT images showing normal perfusion of both
hemispheres were negative for vasospasm. CBV = cerebral blood volume; CBF = cerebral blood flow; MT T = mean transit time; T TD = time to drain.

confidence interval: 0.607, 0.725], P A weak but significant negative cor- low-dose VP CT images had similar
, .001; low dose: r = 0.667 [bias-cor- relation between lesion size and the rel- correlation levels with DSA findings.
rected 95% confidence interval: 0.607, ative measurement error was detected Furthermore, the quantitative vol-
0.723], P , .001) without significant (r = 20.366; P , .001). umetric analysis of the low-dose VP
differences between the two correlation CT images showed a high accuracy
coefficients (z = 0.12; P = .905). with low measurement errors and a
Discussion strong correlation with the original
Quantitative Analyses of Perfusion Maps We evaluated the diagnostic accuracy VP CT images. The weak but signif-
The mean affected volume was 56.9 of low-dose VP CT for assessment icant negative correlation between
cm3 6 38.5 (standard deviation) at of patients with subarachnoid hem- lesion size and the relative measure-
original VP CT and 54.4 cm3 6 37.6 orrhage who are at risk for cerebral ment error, however, could indicate
at low-dose CT with small measure- vasospasm and delayed cerebral is- an association of lesion size and mea-
ment errors between the two dose chemia. We found that image qual- surement accuracy, specifically that
levels (measurement error, 2.8 cm3 ity and diagnostic confidence of low- measurement accuracy in low-dose
6 2.7 [range, 0.1–11.6 cm3; median, dose VP CT images were sufficient VP CT might be impaired in smaller
1.8 cm3]; relative measurement er- in all cases. When compared with lesions. Nonetheless, the findings in-
ror, 5.9% 6 5.1 [range, 0.2%–25.0%; original perfusion data sets (defined dicate that low-dose CT is comparable
median, 4.0%). The interreader as the reference standard), low-dose to original-dose VP CT for assessment
agreement was almost perfect for both VP CT images helped to classify the of cerebral vasospasm in the included
original and low-dose data sets (intra- patients regarding positive and nega- cohort.
class correlation coefficient, 0.995; P tive findings correctly in all cases. The We hypothesize that our findings
, .001). Original and low-dose perfu- segment-based diagnostic accuracy are caused by a relative insensitivity
sion data sets showed strong correla- was also high for low-dose data sets of perfusion images to the added im-
tion regarding affected volumes (r = without significant differences com- age noise. A recent study (14) on dose
0.955; bias-corrected 95% confidence pared with original data sets. In the dependence of quantitative VP CT pa-
interval: 0.935, 0.972; P , .001). subgroup analysis, both original and rameters showed that a radiation dose

648 radiology.rsna.org  n Radiology: Volume 287: Number 2—May 2018

NEURORADIOLOGY: Simulated Low-Dose Perfusion CT to Detect Cerebral Perfusion Impairment Afat et al

reduction down to half of the original with aneurysmal subarachnoid hemor- tations. Other relationships: disclosed no rele-
vant relationships. A.E.O. disclosed no relevant
dose does not have relevant effects on rhage. Our findings indicate a high diag-
relationships.
quantitative perfusion parameters. nostic accuracy of low-dose VP CT for
We deliberately chose original VP detection of perfusion impairment after References
CT images over DSA data sets as refer- aneurysmal subarachnoid hemorrhage, 1. Connolly ES Jr, Rabinstein AA, Carhua-
ence standard for calculating diagnostic and therefore encourage further pro- poma JR, et al. Guidelines for the man-
accuracy despite the lack of established spective use and clinical evaluation of agement of aneurysmal subarachnoid
metrics for VP CT. We chose not to use low-dose VP CT in such patients. hemorrhage: a guideline for healthcare pro-
fessionals from the American Heart Asso-
DSA as reference standard because it This study has limitations. The ret-
ciation/american Stroke Association. Stroke
only depicts macrovascular vasospasm rospective design of this study is associ- 2012;43(6):1711–1737 .
(23), whereas perfusion impairments ated with a selection bias. The relatively
on VP CT images reflect both macro- small sample size is also a limitation. 2. Al-Khindi T, Macdonald RL, Schweizer TA.
Cognitive and functional outcome after an-
vascular and arteriolar vasospasm (23). Our low-dose images were simulated,
eurysmal subarachnoid hemorrhage. Stroke
Therefore, the use of DSA as a refer- and thus the lack of real low-dose im- 2010;41(8):e519–e536 .
ence standard would be inappropriate ages is a limitation of this study. The
3. Macdonald RL, Pluta RM, Zhang JH. Ce-
because perfusion impairments, which applied low-dose simulation method
rebral vasospasm after subarachnoid hem-
are solely because of arteriolar vaso- has recently been validated and applied orrhage: the emerging revolution. Nat Clin
spasm, would be incorrectly classified in prior studies with satisfactory results Pract Neurol 2007;3(5):256–263 .
as false-positive findings (24). (18,31). Finally, the section thickness
4. Vergouwen MD, Etminan N, Ilodigwe D,
Many studies (1,17,25,26) reported of CT and particularly of VP CT affects Macdonald RL. Lower incidence of cere-
that VP CT is an appropriate method image noise, with higher noise levels bral infarction correlates with improved
to assess perfusion abnormalities as- in thinner sections (34). In this study, functional outcome after aneurysmal sub-
sociated with cerebral vasospasm and we used a section thickness of 10.0 arachnoid hemorrhage. J Cereb Blood Flow
delayed cerebral ischemia. In fact, mm. Therefore, our findings should be Metab 2011;31(7):1545–1553 .
these are emerging data indicating taken with caution because they cannot 5. Heros RC, Zervas NT, Varsos V. Cerebral va-
that perfusion imaging is more accu- easily be transferred to data acquired sospasm after subarachnoid hemorrhage: an
rate to assess vasospasm and delayed with thinner sections; further studies update. Ann Neurol 1983;14(6):599–608 .
cerebral ischemia compared with an- regarding the interference of section 6. Vergouwen MD, Vermeulen M, van Gijn J,
atomic imaging of arterial narrow- thickness and radiation dose reduction et al. Definition of delayed cerebral ische-
ing (ie, DSA and CT angiography) or should be performed. mia after aneurysmal subarachnoid hemor-
changes of blood velocity (ie, transcra- To conclude, the results of this study rhage as an outcome event in clinical tri-
als and observational studies: proposal of
nial Doppler) (25,27,28). Connolly et indicate that radiation dose reduction
a multidisciplinary research group. Stroke
al (1) recently recommended VP CT to 40% of original dose levels may be 2010;41(10):2391–2395 .
imaging in patients suspected of having performed in perfusion CT imaging of
7. Hoeffner EG, Case I, Jain R, et al. Cerebral
vasospasm. Temporal perfusion maps, patients with aneurysmal subarachnoid
perfusion CT: technique and clinical applica-
particularly mean transit time maps, hemorrhage without compromising the tions. Radiology 2004;231(3):632–644 .
are known to be most sensitive for ce- diagnostic accuracy regarding detection
rebral vasospasm (17,29). Until now, of cerebral perfusion impairment. 8. Hirata M, Sugawara Y, Murase K, Miki H,
Mochizuki T. Evaluation of optimal scan du-
data were lacking on the applicability of
ration and end time in cerebral CT perfusion
low-dose VP CT for detection of vaso- Disclosures of Conflicts of Interest: S.A. dis- study. Radiat Med 2005;23(5):351–363.
spasm and delayed cerebral ischemia. closed no relevant relationships. C.B. disclosed
no relevant relationships. O.N. disclosed no rel- 9. Hoang JK, Wang C, Frush DP, et al. Esti-
Perfusion abnormalities in patients with
evant relationships. M.M. disclosed no relevant mation of radiation exposure for brain per-
vasospasm are often more subtle com- relationships. K.M.T. disclosed no relevant rela- fusion CT: standard protocol compared with
pared with patients with acute stroke tionships. M.A.B. disclosed no relevant relation- deviations in protocol. AJR Am J Roentgen-
because of vessel occlusion (17,25,30). ships. K.N. disclosed no relevant relationships. ol 2013;201(5):W730–W734 .
Therefore, results of low-dose VP CT in M.W. J.H.K. Activities related to the present
article: disclosed no relevant relationships. Ac- 10. Wintermark M, Lev MH. FDA investigates
stroke patients cannot be transferred to tivities not related to the present article: dis- the safety of brain perfusion CT. AJNR Am
this patient group (13–15,31–33), and closed money paid by Stryker Neurovascular for J Neuroradiol 2010;31(1):2–3 .
the effects of radiation dose reduction a consultancy; disclosed money paid by Bracco
Imaging, Medtronic, Siemens Healthcare, and 11. Sodickson A, Baeyens PF, Andriole KP, et
in these patients have to be evaluated al. Recurrent CT, cumulative radiation ex-
Stryker Neurovascular for payments for lec-
carefully before prospective application. tures; disclosed royalties paid by Springer; and posure, and associated radiation-induced
Given the lack of data on this topic, we disclosed money paid by Abbott, AB Medica, cancer risks from CT of adults. Radiology
chose a retrospective low-dose simula- Acandis, Bayer, Bracco, B. Braun, Codman Neu- 2009;251(1):175–184 .
rovascular, Medtronic, Dahlhausen, Microven-
tion study design as an initial step to- tion, Penumbra, Phenox, Philips Healthcare, 12. Silva AC, Lawder HJ, Hara A, Kujak J, Pav-
ward preparation and application of a Silk Road Medical, St Jude, and Stryker Neuro- licek W. Innovations in CT dose reduction
low-dose VP CT protocol for patients vascular for development of educational presen- strategy: application of the adaptive statis-

Radiology: Volume 287: Number 2—May 2018  n  radiology.rsna.org 649

NEURORADIOLOGY: Simulated Low-Dose Perfusion CT to Detect Cerebral Perfusion Impairment Afat et al

tical iterative reconstruction algorithm. AJR with acute ischemic stroke. Neuroradiology different CT modalities in patients with sub-
Am J Roentgenol 2010;194(1):191–199 . 2013;55(7):827–835 . arachnoid hemorrhage with clinical deterio-
ration. Stroke 2009;40(11):3493–3498 .
13. Othman AE, Afat S, Brockmann MA, et al. 21. Genders TS, Spronk S, Stijnen T, Steyer-
Radiation dose reduction in perfusion CT berg EW, Lesaffre E, Hunink MG. Methods 29. Sanelli PC, Ugorec I, Johnson CE, et al. Us-
imaging of the brain: A review of the litera- for calculating sensitivity and specificity ing quantitative CT perfusion for evaluation
ture. J Neuroradiol 2016;43(1):1–5 . of clustered data: a tutorial. Radiology of delayed cerebral ischemia following an-
2012;265(3):910–916 . eurysmal subarachnoid hemorrhage. AJNR
14. Manniesing R, Oei MT, van Ginneken B,
Am J Neuroradiol 2011;32(11):2047–2053 .
Prokop M. Quantitative Dose Dependency 22. Fleiss JL, Levin B, Paik MC. Statistical

Analysis of Whole-Brain CT Perfusion Imag- methods for rates and proportions. Hobo- 30. Pham M, Johnson A, Bartsch AJ, et al.

ing. Radiology 2016;278(1):190–197 . ken, NJ: Wiley, 2013. CT perfusion predicts secondary cerebral
infarction after aneurysmal subarachnoid
15. Othman AE, Brockmann C, Yang Z, et al. 23. Nolan CP, Macdonald RL. Can angiographic hemorrhage. Neurology 2007;69(8):762–
Effects of radiation dose reduction in Vol- vasospasm be used as a surrogate marker 765 .
ume Perfusion CT imaging of acute ischemic in evaluating therapeutic interventions for
stroke. Eur Radiol 2015;25(12):3415–3422 . cerebral vasospasm? Neurosurg Focus 31. Othman AE, Afat S, Brockmann C, et al.
2006;21(3):E1 . Low-Dose Volume-Perfusion CT of the
16. Niesten JM, van der Schaaf IC, Riordan AJ, Brain: Effects of Radiation Dose Reduction
et al. Radiation dose reduction in cerebral 24. Macdonald RL. Origins of the Concept of
on Performance of Perfusion CT Algorithms.
CT perfusion imaging using iterative recon- Vasospasm. Stroke 2016;47(1):e11–e15 . Clin Neuroradiol 2017;27(3):311–318 .
struction. Eur Radiol 2014;24(2):484–493 .
25. van der Schaaf I, Wermer MJ, van der Graaf 32. Othman AE, Brockmann C, Yang Z, et al.
17. Othman AE, Afat S, Nikoubashman O, et
Y, Hoff RG, Rinkel GJ, Velthuis BK. CT after Impact of image denoising on image quality,
al. Volume perfusion CT imaging of cere- subarachnoid hemorrhage: relation of cere- quantitative parameters and sensitivity of
bral vasospasm: diagnostic performance of bral perfusion to delayed cerebral ischemia. ultra-low-dose volume perfusion CT imag-
different perfusion maps. Neuroradiology Neurology 2006;66(10):1533–1538 . ing. Eur Radiol 2016;26(1):167–174 .
2016;58(8):787–792 .
26. Sanelli PC, Pandya A, Segal AZ, et al.
33. Murphy A, So A, Lee TY, et al. Low dose CT
18. Won Kim C, Kim JH. Realistic simulation of Cost-effectiveness of CT angiography and perfusion in acute ischemic stroke. Neuro-
reduced-dose CT with noise modeling and perfusion imaging for delayed cerebral is- radiology 2014;56(12):1055–1062 .
sinogram synthesis using DICOM CT im- chemia and vasospasm in aneurysmal sub-
ages. Med Phys 2014;41(1):011901 . arachnoid hemorrhage. AJNR Am J Neuro- 34. Bennink E, Oosterbroek J, Horsch AD, et
radiol 2014;35(9):1714–1720 . al. Influence of Thin Slice Reconstruction
19. Cremers CH, van der Schaaf IC, Wensink on CT Brain Perfusion Analysis. PLoS One
E, et al. CT perfusion and delayed cere- 27. Jost SC, Diringer MN, Zazulia AR, et al.
2015;10(9):e0137766 .
bral ischemia in aneurysmal subarach- Effect of normal saline bolus on cerebral
noid hemorrhage: a systematic review and blood flow in regions with low baseline 35. Wintermark M, Ko NU, Smith WS, Liu S,
meta-analysis. J Cereb Blood Flow Metab flow in patients with vasospasm following Higashida RT, Dillon WP. Vasospasm after
2014;34(2):200–207 . subarachnoid hemorrhage. J Neurosurg subarachnoid hemorrhage: utility of perfu-
2005;103(1):25–30 . sion CT and CT angiography on diagnosis
20. Thierfelder KM, Sommer WH, Baumann
and management. AJNR Am J Neuroradiol
AB, et al. Whole-brain CT perfusion: re- 28. Dankbaar JW, de Rooij NK, Velthuis BK,
2006;27(1):26–34.
liability and reproducibility of volumetric Frijns CJ, Rinkel GJ, van der Schaaf IC.
perfusion deficit assessment in patients Diagnosing delayed cerebral ischemia with

650 radiology.rsna.org  n Radiology: Volume 287: Number 2—May 2018