Downloaded from www.ajronline.org by Tulane University on 09/05/14 from IP address 129.81.226.78. Copyright ARRS.

For personal use only; all rights reserved

Soft-tissue Hemangiomas in Infants

and Children: DiagnosisUsingDoppler
Sonography

Josée Dubois1 OBJECTIVE. We describethe sonographicappearanceand vascularizationof hemangio

HeidiB. Patriquin1 mas and determine if vessel density and peak systolic Doppler shifts distinguish hemangiomas
LaurentGarel1 from other superficial soft-tissue masses.
Julie Powell2 SUBJECTSAND METHODS. Ourpilotstudyincluded20 infantsandchildrenwithhe
mangiomas who were to undergo biopsy before treatment with interferon alpha-2b. We used
DenisFiliatrault1
Doppler sonography to determine the number of vesselsper square centimeter, peak arterial
MichèleDavid3 Doppler shift, resistive index, and signs of arteriovenous shunting. All hemangiomas showed
Andrée Grignon1 high vesseldensity (more than five per squarecentimeter) and high Doppler shifts (more than
2 kHz), and these two factors became our diagnostic criteria. A prospective study of I 16 pa
tients was then carried out. One hundred sixteen consecutive pediatric patients with superficial
soft-tissue masses were examined using Doppler sonography: sonographic findings were
compared with the final diagnoses that were established by biopsy, CT, or clinical follow-up.
RESULTS. The final diagnosesincluded70 hemangiomas,
20 venousmalformations,
three arteriovenous malformations, three arteriolocapillary malformations, and 20 other
masses. Fifty-nine lesions showing high vessel density (more than five per square centimeter)
and a peak arterial Doppler shift exceeding 2 kHz were correctly diagnosed as hemangiomas
(sensitivity, 84%; specificity, 98%). One artenovenous malformation showed high vesselden
sity and high Doppler shifts, but none of the other masses that were not hemangiomas did.
Eleven patients with hemangiomas who were being treated with interferon at the time of the
study fulfilled only one of the two diagnostic criteria.
CONCLUSION. HighvesseldensityandhighpeakarterialDopplershiftcanbeusedto
distinguish hemangiomas from other soft-tissue masses.

H emangiomas,which areamongthe high-frequency gray-scale and Doppler sonog

most common soft-tissue tumors raphy and whether they can be distinguished
in infants [1—3),
typically appearas from other superficial soft-tissuemassesin in
slightly raised,bluish red subcutaneousmasses fantsandchildren.
that resemblethe surface of a strawberry and
regressasthe child grows older. Someheman
giomasdo not havethis typical appearancebe Subjects and Methods
cause part or all of the lesion is deep in the soft PilotStudy

ReceivedOctober27,1997;acceptedafterrevision tissue and the overlying skin appears normal. To determine the general appearanceand vascu
February
5,1998. These lesions are difficult to distinguish clini larization characteristicsof soft-tissue hemangiomas
1Department of Medical Imaging, HôpitalSainte-Justine, cally from more suspicioussoft-tissuemasses, in infants, we performedhigh-resolutiongray-scale
3175Câte-Sainte-Catherine,
Montréal,
Québec,
Canada such as vascularmalformations, soft-tissueUi and Doppler sonography and CT with IV contrast
H3T1C5.
Addresscorrespondence
toJ.Dubois. materialon 20 infantsand children with hemangio
mors (e.g., metastasesfrom neuroblastomaor
2Departmentof Dermatology,
Hôpital
Sainte-Justine, rhabdomyosarcoma),and infantile myofibro maswhowereaboutto undergobiopsybeforecom
Montréal,
Québec,CanadaH3T1C5. mencingtherapywith interferonalpha-2b(IntronA,
matosis.Children with suchlesionsareusually
3Department
ofHematology
, Hôpital
Sainte-Justine, Schering,Canada).We useda Mark9 HDIscanner
referredfor imaging studies or biopsy.
Montréal,
Québec,
CanadaH3T1C5. (AlL, Seattle,WA) with a 10-MHz linear array
Blood flow in superficial vessels is readily transducer.Color Dopplersonogramswere obtained
AJR1998;171:247—252
discernible using Doppler sonography. We with low-pulse repetition frequency and a wall filter.
0361—803X/98/171
1—247 soughtto ascertainwhetherhemangiomashave Pulse-repetition frequency was increased only if
©AmericanRoentgen
RaySociety characteristic features that can be seen using aliasing occurred. The area of greatest vasculariza

AJR:171,July 1998 247

 @ @i
‘¿p.
@ J_@s@
Downloaded from www.ajronline.org by Tulane University on 09/05/14 from IP address 129.81.226.78. Copyright ARRS. For personal use only; all rights reserved
@ :-
@ ‘¿: :_. -@@“
@ . .
@1@t
@,
@ @-;J@
@ Fig.1.—Hemangioma
ofrightcheekin2-month-old
girl(normal
overly
ing skin).
@ A, Gray-scalesonogramrevealsheterogeneousmass(arrows).
B, ColorDopplersonogramshows highvessel density(morethan five
visible vessels per square centimeter).
C,Spectroscopyshows peak arterialDopplershiftof5 kHz.
tion was selected, and the number of vessels per
square centimeter outlined on color Doppler imag
ing was counted. Peak arterial Doppler shifts and
the resistiveindex(RI) were ascertainedwith pulsed
Doppler sonography.All of these lesionswere yen
fled as hemangioma.sby biopsy.They had variable
appearance on gray-scale sonography but were
highly vascular,containing at least five vessels per
squarecentimeter;thepeakvelocitywashigherthan
2 kHz. On CT, the hemangiomas showed rapid up
take and intense, persistentretention of contrast ma
terial.These findingswere used as the criteriafor the
diagnoses of hemangiomas in the subsequentpro
spectivestudy.
Prospective
ClinicalStudy
All infantsand childrenreferredto our dermatol
ogy—vascular
malformations clinic with clinically
248
Duboiset al.
@. -
.: , •¿
@@--.
@‘¿
A B
@i
-,...,@
.,,
-,@
I
diagnosedhemangiomasor other superficialsoft-tis
sue masses were screened by the attending dermatol
ogist. Patients were referred for sonographic and CT
examinationif the lesiondid not havethe typical
physical appearance of a hemangioma or appeared
to have a subcutaneous extension that was difficult to
assess clinically. Between June 1991 and January
1996. all such children were examined by sonogra
phy, and the sonographic diagnoses were compared
with the finaldiagnoses.
Sonogramswere obtained by one of four pediat
nc radiologists trained in Doppler sonography to
whom the details of the sonographic technique for
the study had been carefully explained. Either a
QuantumII (Siemens,Issaquah,WA) with a 7.5-
MHz linear gray-scale and Doppler transduceror a
Mark 9 HDI (ATL) with a 10-MHz or 6-MHz
transducerwas used.A 1-cm-thick standoffpad
was used when necessary. The length, width, and
-
@.@iI:@.
U!@JHf@J F@EV
I E@.J
1 @—¿
. -
-2—
C
depth of the lesions were measured using elec
troniccalipers.Echogenicitywasassessedashyp
echoic or hyperechoic. Gray-scale sonography was
used to look for visible vesselsaround or within the
lesion. Internal architecture was classified as homo
geneousor heterogeneous.Dopplersonographywas
performed using the lowest pulse-repetition fre
quencyandDopplergainsettingsthatdid notcause
aliasing (pulse-repetition frequency. 2—16.6kHz;
wall filter, 50—100kHz). Color Doppler sono
graphy was used to scan the entire lesion with a
restrictedfield of view.Theareaof greatestvascu
laity was retained for analysis.The fieldof interest
was reduced to I cm2. Vessels were identified by
color-flow signals, and vessel density was defined
semiquantitativelyas sparse or low density (fewer
than two vessels per square centimeter), moderate
or medium (two to four vesselsper square centime
ter), and numerous or high (more than five vessels
AJR:171,July 1998
 @ . .@

Diagnosing Childhood Hemangiomas with Doppler Sonography

@ @1 10 OCT
@J.
12.J
I @i!JE@COP
R RE@IE
AL
Downloaded from www.ajronline.org by Tulane University on 09/05/14 from IP address 129.81.226.78. Copyright ARRS. For personal use only; all rights reserved

@ .k@ , ‘¿@1

@lL

ii
A B

Fig.2.—Hemangioma
oforbitalregionin2-month-oldgirl.
A, ColorDopplersonogramshowshighvesseldensity.
B,Spectroscopyshows highsystolicflowwithpeakof4 kHz.

per square centimeter).A pulsed Dopplerexamina Results sonography.We correctly diagnosedthese le

tion was then pertomied.placingthe Dopplersample sions as hemangiomas. Three other small le
intoarteriesand veins.PeakarterialsystolicDoppler One hundredsixteeninfantsand children
who were 2 days to 7 years old (mean age, I8 sionshad many sitesof arteriovenousshunting.
shifts were noted and classified as low (<I kHz),
months old) were examined. The final diag Several dilated vessels were noted on gray
moderate ( 1—2
kHz), or high (>2 kHz). The RI was
scale sonography and identified as feeding ar
calculated (systolic —¿
diastolic I systolic Doppler noses included 70 hemangiomas, 20 venous
teries and veins on Doppler sonography. One
shills). Evidence of arteriovenous shunting was malformations, three arteriovenous malfor
sought.includinghighdiastolicflow.RI lessthan0.5. mations, three arteriolocapillary malforma lesion had high vessel density and peak sys
and pulsatilevenousflow. tions, and 20 other masses. tolic shift greater than 2 kHz. The other two
Eighty-five percent of these examinations were The sizeof the hemangiomasrangedfrom had moderatevesseldensity (three per square
performed without sedation of the patient. The re 2 to 640 cm3. The gray-scale sonographic ap centimeter)with a highDopplershift(2 kHz).
mainder were performed immediately after CT, We therefore made the diagnosis of “¿lesion
pearancewas nonspecific. One lesion showed
taking advantage of the sedation necessary for that nonhemangioma,probably arteriovenousmal
examination. Informed consent was obtained from
a small calcification. In four lesions, one or
two vessels were visible on gray-scale sonog
formation― (Fig. 4) in thesethreepatients.This
patients' parents. The study was approved by the
raphy. In the remainder of patients,no vessels diagnosiswasconfirmed by angiography.
hospital committee on human experimentation.
After Doppler sonograms were obtained, the ex were seen on gray-scale sonography. The49 nonhemangioma
soft-tissue
masses
amining radiologist reported the diagnostic im Doppler sonographyrevealedhigh vessel had zero to five vesselsper squarecentimeter
pression to the clinician. At the end of the study density (five vesselsor more per squarecenti and a maximum systolic Doppler shift ofO.8 to
period. all sonograms were reviewed by the princi meter) in 65 (93%) of 70 hemangiomas and a 6.5 kHz, with a mean of 1.32 (SD = I .4). The
pal investigator. maximum systolic Doppler shift greaterthan 2 RI rangedfrom0.2 to 1.00,witha meanof 0.45
The diagnosis of hemangioma was made if the kHz in 63 (90%) of 70 (Figs. 1—3). Using (SD = 0.35) (Figs.5 and6). Nonefulfilledthe
Doppler examination showed high vessel density two criteria for the diagnosis of hemangioma.
thesetwo criteria establishedin the pilot study,
(more than five vessels per centimeter) and a high Some lesions had only venous flow. Three
Doppler shift (>2 kHz). All masses that did not
we were able to make diagnosesof hemangi
atypical massesthat were clearly not hemangi
masses oma in 59 (84%) of 70 hemangiomas.Maxi
fulfill thesecriteria were called “¿other
nonhemangioma.― Lesions with several sites of mum Doppler shifts ranged from I .1 to 10 omas were difficult to diagnose on Doppler
arteriovenous shunting were called “¿nonhemangi kHz with a mean ±SD of 3.8 ±2.0 kHz. RI sonography. They did not fulfill the two sono
oma—likely arteriovenous malformation.― Biopsy ranged from 0.25 to 0.89 with a mean of 0.59 graphic criteria used for the diagnosis of he
was performed if a lesion did not fulfill the criteria ±0. 13 kHz). Eleven hemangiomas fulfilled mangioma and were therefore labeled “¿other
for hemangioma on clinical follow-up or on CT. only one of the two diagnostic criteria and masses.―In one mass, vessel density was mod
The finaldiagnosis was established by clinical fol were therefore diagnosed as “¿mass—nonheerate(threeper squarecentimeter),with a high
low-up (regression or stability of the mass after mangioma― on sonography.These I I masses peaksystolic Doppler shift(6.5 kHz). The diag
age 2 years) in 49 patients: angiography or phle
were clinically stable, and the children were nosis of “¿mass—nonhemangioma―was made.
bography in 23 patients: and CT and biopsy in 44
patients. The sonographic findings were compared receiving interferon treatment at the time. The masswas very hardon palpation.Biopsy
with the final diagnoses. Sensitivity, specificity, Three hemangiomaswith high vesselden showed a rhabdomyosarcoma.Another mass,
positivepredictive value, negativepredictivevalue, sity and peak systolic shifts greaterthan 2 kHz which hadappearedlate in childhood, had high
and diagnostic accuracy of the sonographicexami had one or two sites of arteriovenous shunting. vessel density (more than five per square centi
nations were calculated. No feeding vessels were seen on gray-scale meter) and moderateDoppler shift ( 1.3 kHz). It

AJR:171,July 1998 249

 Dubois et al.

Fig. 3.—Hemangioma with normal

,et overlying skin on back of 6-month
oldboy.
A, ColorDopplersonogram shows
highvesseldensity(morethanfive
.9-- persquarecentimeter).
B, Spectroscopyshowspeaksys
Downloaded from www.ajronline.org by Tulane University on 09/05/14 from IP address 129.81.226.78. Copyright ARRS. For personal use only; all rights reserved

tolic Dopplershift of 5.7kHz.

Fig.4.—Arteriovenous
malformationoffacein2-year-oldgirl.
A,Gray-scale sonogramshowsseveral vessels (arrow).
B,Spectroscopy showslargefeedingartery(arrow).Bloodflowhaslowresistanceandhighdiastolicflow,sug
gestingarteriovenous
shunting.
C,Spectroscopy showslargedrainingvein(arrow).Bloodflowispulsatile,compatible
withvenousflowdistalto
arteniovenousmalformation.
. ; - .@

@: . -

. .‘@ .

250 AJR:171,July 1998

 @ ‘¿@ . @@:: . -

Diagnosing Childhood Hemangiomas with Doppler Sonography

,@ --. 9'
Downloaded from www.ajronline.org by Tulane University on 09/05/14 from IP address 129.81.226.78. Copyright ARRS. For personal use only; all rights reserved

@ V9 9

;@
A

Fig.5.—Infantile
myofibromatosis
onbackof1-month-old
girl.
A,Gray-scalesonogramrevealsnonspecific
mass(arrows).
B,ColorDopplerspectroscopy
showslowvesseldensitywithsystolicpeakof 1kHz.

was identified as an angiofollicular hamartoma Discussion bach-Merritt syndrome). Microscopically, the

on biopsy.A lesion in a neonatehad low vessel
density (fewer than two per square centimeter)
with very high Doppler shifts (6.5 kH.z). No
specific sonographicdiagnosis was made. Bi
opsy showeda sarcoma.
Using the two criteria of high vessel den
sity (more than five per square centimeter)
and peak arterial Doppler shift greater than 2
kHz for the diagnosis of hemangioma, we
found the sensitivity of the Doppler sonogra
phy to be 84% (59/70); specificity, 98% (45/
46); positive predictive value, 97% (59/61);
and negative predictive value, 82% (45/55).
Infantile hemangiomasare benign vascular
lesions characterized by a phase of initial
growth andangiogenesis,a plateauphaseof in
activity, slow resolution of angiogenesis,and
regressionof both the size and vascularity of
the tumors [4-6]. Their severity varies; heman
giomas range from an asymptomatic, discol
oredspoton the skin to large,highly disfiguring
massesthat can be life-threatening when they
occur near vital structuressuch as the airway.
Their high vascularity can cause high output
heart failure. Sequestration of platelets in the
lesion may lead to thrombocytopenia (Kasa
hallmark of a growing hemangiomais a prolif
eration of endothelial cells, forming syncytial
massescomposedof numerousvesselsor cords
of endothelial cells. The vessel lumens are of
ten compressed. Occasional mitotic figures are
seen. However, the nuclei have a benign ap
pearance and are not pleomorphic. As the pro
liferative phase progresses,vascular channels
are less compressed and capillary lumens lined
by plump endothelial cells may be seen [7].
Mast cells arealsoabundantin proliferating he
mangiomatissue[8]. Most hemangiomashave
a typical “¿strawberry―
appearancewith well

Fig.6.—Atypical
massonforehead
of 1-week-oldboy.
A, ColorDopplersonogramshows
low vessel density. .@,
B, Spectroscopy
showspeakDop
pIer shift of 1.5kHz.Rhabdomyosar
comawasdiagnosedbybiopsy.
B

AJR:171,
July1998 251
 Dubois et al.

defined borders. This type presents little diag
nostic difficulty on clinical examination.Ding
nostic problemsarisewhen lesionsareatypical
and when all or part of a lesion lies deep in the
subcutaneous tissues and therefore cannot be
clearly defined on physical examination. This
Downloaded from www.ajronline.org by Tulane University on 09/05/14 from IP address 129.81.226.78. Copyright ARRS. For personal use only; all rights reserved
et al. [9], who found that arteriovenousmalfor high density ofvessels throughout its mass that
mations can be differentiated from hemangio is seen in hemangiomas. The 11 hemangiomas
man by the absence of a soft-tissue mass. Only in our seriesthat had fewer than five vessels
one of 70 hemangiomas showed one vessel per square centimeter or less than 2 kHz Dop
large enough to be identified on gray-scale pler shift were clinically stable lesions in chil
sonography, and no pulsatile draining veins dren undergoing interferon alpha-2b treatment.

type of lesion is difficult to distinguish from were identified. Some hemangiomashad one It remains to be seen whether the diminishing
vascularmalformations and tumors such as in or two sites of artenovenous shunting that angiogenic stimulation in treated or spontane
fantile myofibromatosis, neuroblastomas, me were not of the same density as arteriovenous ously involuting hemangiomas is reflected in
tastases, and sarcomas. Children with such malformations. Two sarcomasand one anglo changingDopplerpatterns.
lesionsareusually referredfor imaging studies. fofficular hamartoma had the high Doppler
Because modern sonographic equipment is shifts characteristic of hemangiomas, but their
ideally suited to examining superficial lesions, vessels were sparse in number and at the pe
References
this noninvasivemethod is useful in examining riphery ofthe lesions [10, 111.
1. Pratt AG. Birthmarksin infants.Arch Dermatol
children with such lesions. Feeding arteries Why do hemangiomas(benign lesions) and
Syphilol 1953;67:302—305
and veins are more easily identified by Dop sarcomassometimessharesimilar blood flow 2. Jacobs
AH. WaitonRG.Theincidence
of birth
pler sonography than by CT because the le characteristicsas seen on Doppler sonogra marks in the neonate. Pediatrics 1976;58:218—
sions can be examined in many planes. phy? The answer probably lies in the concept 222
In our pilot study of biopsy proven heman of angiogenesis [12, 13]. Some lesions are ca 3. PowellTG, WestCR, PhamahP0, CoatieRW.Epi
giomas, we found that all hemangiomas had pable of stimulating the host to create new demiologyof strawberryhemangiomas in low
birthweightinfants.Br J Dermatol1987;I16:635—
blood flow through many vessels too small to vessels, usually from existing endotheial
641
be seenon gray-scalesonography. We devised cells, by secreting angiogenesis factors. The 4. Mulliken JB, Glowacki J. Hemangioma and yes
a method of estimating vessel density by molecular strecture of several angiogenesis cular malformations in infants and children: a
counting the number of vesselsper squarecen factors has been defined [14, 15]. Angiogene classification
basedonendotheialcharacteristics.
timetershownoncolorDopplersonographyin sis factors have been isolated from many tis Plast ReconsrrSurg 19$3;69:412—420
the areaof greatestvascularization in eachle sues including fibroblasts, wound fluids, 5. Burrows PE, Mulliken JB, Fellows ICE,Strand RD.
sion. Technical factors were kept constant. The placentas, numerous malignant tumors, he Cbildhoodheman@ and vascularmalforma
tions:angiographic
differentiation.
AiR 1983;l41:
vessel count is semiquantitative at best, consid mangioendotheiomas, andinfantilehemangi 483-488
ering that one tortuous vessel may yield more oman. The new vessels produced through 6. Merland JJ, Rich MC, Monteil JP, Hadjean E.
than one color Doppler signal. We classified angiogenic stimulation do not have a muscu Classification
actuelledesmalformations
vascu
vessel density as low, medium, or high. Maxi lads layer, and therefore blood flow through laires.
AnnChirPlast19$O;25:l05—l
11
mum systolic Doppler shift rather than velocity the network of vessels feeding them has high 7. Mulliken JB, Young AE. Pathogenesis of hemangi
was measuredbecausethe vesselbeamangle velocity and low resistance.Flow within the omas.In: MullikenLB.YoungAE, eds.Vascular
birthmarks.Philadelpbia Saunders,1988:63-76
was rarely known. The 20 patientswith he individual microscopic vesselsis slow. Thus it
8. Glowacki J, Mulliken JB. Mast cells in hemangi
mangiomasin the pilot study had a maximum is likely that the blood flow in hemangiomas omas and vascularmalformations.Pediatrics
Doppler shift exceeding2 kHz. During the that is detected by Doppler sonography occurs 1982;70:48—5l
prospective study of 116 children with various in thevesselsfeedingtheneovasculature stim 9. MeyeriS, Hoffer FA, BarnesPD, Mulliken JB. Bi
subcutaneous masses, we found that the two ulated by angiogenesis factors. During their ologicalclassification
of softtissuevascular
anorn
criteria of high vessel density and Doppler phase of proliferation, hemangiomas receive sues:MRcorrelation.AJR 1991;157:559-564
10. Taylor KJW, Ramos I, Carter D, Morse SS,
shift greater than 2 kHz allowed us to distin maximal angiogenic stimulation, as evidenced
SnowerD, FortuneK. CorrelationofDoppler US
guish hemangiomasfrom other lesions,which by the large number of vessels and cords of tumorsignalswith neovascular
morphologicfea
had fewer vessels or lower Doppler shifts. endotheial cells seen by microscopy. Other fires.Radiology1988;l66:57—62
A numberof pitfallswereencountered. Ar benign superficial tumors of childhood do not 11. VanCampenhout I, Patriquin
H. Malignantsal
teriovenous malformations also showed more appear to manifest the intense degree of an crovasculaturein abdominal tumors in children:
than five vessels per square centimeter and giogenesis noted in hemangiomas. This may detectionwithDopplerUS.Radiology1992;l83:
445-448
Doppler shift exceeding 2 kHz. However, explain the difference in vessel density and
12. Folkman J, Watson K, In@ D, Hanahan D. In
feeding vessels large enough to be seen on peak systolic Doppler shift between hemangi duction of angiogenesisduring the transition from
gray-scale sonographywere found to be feed oman and other benign masses. hyperplasiato neoplasia.Nature 1989;339:58-6l
ing arteriesand draining veins on spectros Although malignant tumors may show in 13. Folkman J. How is blood vessel growth regulated in
copy. Multiple sites of arteriovenous shunting tenseangiogenesis, thenetworkof newvessels normalandneoplastic
tissue?
OowesMemorial
were identified by very high diastolic Doppler is usually found at the peripheiy of the mass. AwardLecture.CancerRes
1986;46:467-473
14. FolkmanJ, KlagsbrunM. Angiogenicfactors.
shifts (RI < 0.5) andabnormally pulsatiledrain Very high Doppler shifts have been detected at
Science1987;235:442-447
ing veins. Each arteriovenous malformation the peripheiy of numerous malignant tumors, 15. Folkman J. Anti-angiogenesis: new concept for
was associated with a small soft-tissue mass. especially in the abdomen [10, 11]. It is not therapy of solid tumors.Ann Surg 1972;175:
We wereunableto confirmthework of Meyer likely thata malignanttumorwouldmimicthe 409—416

252 AJR:171,
July1998