245

Intracranial Hemorrhage in
Premature Infants: Accuracy of
Sonographic Evaluation

Laurence A. Mack1 The real-time high resolution mechanical sonographic sector scanner is a convenient
Kelly Wright2 and useful instrument for the detection of intracranial hemorrhage in premature infants.
Experience with 27 infants with intracranial hemorrhage detected by sonography and
Jack H. Hirsch1
confirmed by computed tomography (CT) or by autopsy is analyzed. The severity of the
Ellsworth C. Alvord3
hemorrhage shown by those methods was graded by an accepted classification for
American Journal of Roentgenology 1981.137:245-250.

Robert 0. Guthrie2
standardized reporting. The extent of intraparenchymal and intraventricular hemor-
William P. Shuma& rhage was accurately assessed by sonography in all cases except for small amounts of
James V. Rogers1 blood in normal sized ventricles in five of 1 2 instances. Sonography also failed to
Nicole F. Bolender1 detect subarachnoid hemorrhage in each of 1 3 cases. There were no known false-
positive sonograms. From this experience the authors believe sonographic sector
scanning should be the initial examination in all infants at high risk for intracranial
hemorrhage. When the ventricles are of normal size, CT scanning is recommended to
search for small intraventricular hemorrhage that may not be detected by sonography.
For subarachnoid bleeding, CT is preferable to sonography.

Sonography has been demonstrated to be effective in evaluating ventricular

size and in defining cystic areas within the neonatal brain [1 -4]. While identifi-
cation of intraventricular blood clot has been reported by several authors [5],
localization of parenchymal brain hemorrhage has only been recently described
[6, 7]. We report our experience with 27 infants who sustained intracranial
hemorrhage that was detected and graded by sonography and confirmed by
computed tomography (CT) or autopsy. Our purpose was to study the accuracy
of sonographic detection and localization.

Materials and Methods

A total of 31 4 cranial sonographic examinations was performed in 1 40 premature infants
in the Neonatal Intensive Care Unit at the University of Washington Hospital during an 11
month period. Infants with a high clinical suspicion of intracranial hemorrhage were selected
Received October 16, 1980; accepted after for examination. Intracranial hemorrhage was diagnosed by sonography in 54 cases. The
revision April 13, 1981. sonographic findings were confirmed by CT within 24 hr in 1 5 infants, by autopsy in 10,
Department of Radiology, University Hospital and by both CT and autopsy in two. These 27 patients, on whom a total of 78 studies was
5B-05, University of Washington. Seattle, WA performed, form the basis of this analysis. In the other 27 infants, timely CT or pathologic
98195. Address reprint requests to L. A. Mack,
correlation was not available and these patients were excluded from analysis. Of the 86
Harborview Medical Center, 325 Ninth Ave., Se-
attle. WA 98104. patients with normal sonograms, 1 1 had confirmatory normal CT or autopsy findings. We
2 Department of Pediatrics, University Hospital have no proof of bleeding in the other 75.
RD-20, University of Washington, Seattle, WA The 27 infants under consideration had a mean gestational age of 29.7 weeks (range,
98195. 27-36 weeks). Mean birth weight was 1 31 0 g (range, 81 0-2040 g). Mean Apgar score at
Department of Pathology, University Hospital 1 mm was 3.4 (range, 1-7), and at 5 mm, 5.7 (range, 1-9). Hyaline membrane disease
RJ-05, University of Washington, Seattle, WA developed in 22 and required respiratory therapy in 21 . Of the 27 infants, 16 had lumbar
98195. punctures and in 15 of these both red blood cell counts and total protein were elevated.
Three infants developed hydrocephalus, requiring intraventricular shunt placement in two.
AJR 137:245-250, August 1981
0361 -8o3x/81 /1372-0245 $00.00 Sonography was performed using a commercially available real-time sector scanner with
© American Roentgen Ray Society a 5 MHz, medium internal-focused transducer (Advanced Technology Labs., Bellevue,
 246 MACK ET AL. AJR:137, August 1981

Wash.). Examinations were performed by direct contact with the lateral ventricle and cerebral tissue to permit examination of the
anterior fontanelle in sagittal and coronal planes [8]. All the exami- floor of the ventricle for small, subependymal hemorrhages.
nations were pertormed by one of the authors (L. A. M., J. H. H., W.
P. S. , or J. V. R.) and images were recorded on Polaroid or 70 mm
film. Interpretations made at the time of the procedure were used in Results
this analysis and confirmed by a retrospective review of scans by
Hemorrhage in the 27 infants was graded using the clas-
the author operators without knowledge of clinical information.
CT was performed on an EMI 1 01 0 (LTS, Northbrook, Ill.) (160 sification proposed by Burstein et al. [9]: grade I, hemor-

x 160 matrix, 80 sec scan time, 30 mA and 200_300 to Reid rhage restricted to the subependymal regions; grade II,

baseline). CT images were recorded on film using a multiformat subependymal hemorrhage with extension of hemorrhage
camera. As with the sonograms, the initial written report was used into normal sized ventricles; grade III, subependymal hem-
in the analysis and confirmed by a retrospective analysis without orrhage with extension of hemorrhage into dilated ventri-
clinical information. des; and grade IV, grade III hemorrhage with extension of
Pathologic examinations were pertormed using standard tech- subependymal hemorrhage into adjacent brain parenchyma.
niques by one of the authors (E. C. A.). After at least 2 weeks of Sonographic detection and grading of hemorrhage is shown
fixation in 10% neutral buffered formalin, coronal sections at the
in table 1.
level of the head of the caudate nucleus were scanned through the

Grade I Hemorrhage
TABLE 1: Sonographic Detection of Hemorrhage

No. Detected No. Confirmed

Five infants had hemorrhage restricted to the subepen-
Grade of Hemorrhage
by Sonography by cT/Autopsy dymal region adjacent to the head of the caudate nucleus.
I .. 10 5 It appeared as a highly echogenic region, with convex
American Journal of Roentgenology 1981.137:245-250.

II 7 12 borders in each instance (fig. 1). The echo levels of hem-

III 7 7 orrhage were far greater than the medium-level echoes of
IV 3 3
normal caudate nucleus and were best seen on direct cor-
Subarachnoid 0 13
onal scans. On sagittal views, the regions of subependymal

Fig. 1 -Grade I hemorrhage. A, Coronal sonogram at level of head of caudate nuclei. Subepen-
dymal hemorrhage on right (arrows). B. Angled sagittal image. Subependymal hemorrhage with
characteristic relation to thalamocaudate notch. c. Next day. CT scan at level of caudate nuclei. Right-
sided subependymal hemorrhage appears as area of increased density (long arrow). Subarachnoid
hemorrhage (not seen on sonogram) in left temporoparietal region (short arrows). CT scan 4 days
later was unchanged. D, 4 days later. Section of brain in same plane as A. Subependymal hemorrhage
(long arrows) and area of subarachnoid hemorrhage (short arrows). v = lateral ventricle, 3 = third
ventricle, C = cavum septum pellucidum, T = thalamus, Lv = lateral ventricle, A = anterior, P =
posterior, H = hemorrhage.

D
 AJR:137, August 1981 INTRACRANIAL HEMORRHAGE IN PREMATURE INFANTS 247

Fig. 2.-Grade i hemorrhage. A,

Coronal sonogram at level of head of
caudate nucleus. Small left subependy-
mal hemorrhage (arrows). B, Pathologic
specimen at same level 9 days later.
Small left subependymal hemorrhage
(arrows). c, Sagittal sonogram in an-
other patient. Clot in occipital pole of
lateral ventricle (arrowheads). D, Sec-
tion of brain. Intraventricular clot in oc-
cipital horn of lateral ventricle. IC = in-
traventricular clot, 0 = occipital horn,
HC = head of caudate nucleus, BC =
body of caudate, T = thalamus, A =
antrum of lateral ventricle, CP = choroid
plexus, v = lateral ventricle, C = cavum
septum pellucidum.
American Journal of Roentgenology 1981.137:245-250.

C D

hemorrhage could be differentiated from choroid plexus by

its anterior and lateral location and by a characteristic
relation to the thalmocaudate notch.

Grade II Hemorrhage

In 1 2 patients subependymal hemorrhage extending into

normal sized lateral ventricles was diagnosed by CT or
pathologic examination. The intraventricular blood was den-
tified as echogenic material in the occipital horn on sagittal
scans in three cases (fig. 2), and in the frontal horns anterior
to the caudate nucleus in four cases (fig. 3).
In the other five cases, sonography demonstrated the
subependymal hemorrhage but failed to detect the intraven-
tricular blood. In three infants, CT demonstrated small Fig. 3.-Grade II hemorrhage. Coronal scan angled
anteriorly. Intraventricular clot fills frontal horn of normal
amounts of intraventricular blood layered dependently in the
sized lateral ventricles. Echogenic clot conforms to con-
occipital horns of normal sized lateral ventricles (fig. 4). In cave margin of ventricle (arrows). IC = intraventricular
one case, sonography failed to differentiate between sub- clot.

ependymal blood and intraventricular blood in normal sized

ventricles. The fifth neonate, scanned at 2 hr of age, dem- tricular clot, as well as a small left subependymal hemor-
onstrated hemorrhage in the subependymal region and nor- rhage. We presume that this represents interim extension of
mal-sized ventricles. The infant died at 9 hr of age. At the original hemorrhage, and does not constitute a sono-
autopsy there was marked ventriculomegaly with intraven- graphic error.
 248 MACK ET AL. AJR:137, August 1981

b
Fig. 4.-Grade II hemorrhage. A, Coronal sonogram at level of head of than B. Small amount of intraventricular blood layered in occipital horns of
caudate nucleus. Bilateral subependymal hemorrhages (arrows) appear as lateral ventricles (arrows), which are not seen on coronal or sagittal sono-
focal areas of high level echoes. B, CT scan same day confirms subependymal grams. v = lateral ventricles, C = cavum septum pellucidum, 0 = occipital
American Journal of Roentgenology 1981.137:245-250.

hemorrhages as areas of increased density (arrows). c, More caudal level horn of lateral ventricle.

Fig. 5.-Grade Ill hemorrhage. A,

Coronal sonogram. Marked dilatation of
lateral and third ventricle. Subepenmdy-
mal hemorrhage on left has ruptured into
lateral ventricle with intraventricutar clot.
B, CT scan 1 day later. Subependymal
hemorrhage and intraventricular dilata-
tion. Lv = lateral ventricle, 3 = third
ventricle, IC = intraventricular clot, H
= hemorrhage, C = cavum septum pel-
lucidum.

Grade III Hemorrhage into the adjacent brain parenchyma (fig. 6). Parenchymal
extension was into the frontoparietal region in two patients
In seven infants, subependymal hemorrhage, as well as
and into the midparietal region in one. The parenchymal clot
hemorrhage into enlarged ventricles was diagnosed sono-
possessed the same high echo level as did blood elsewhere.
graphically and confirmed by CT or autopsy (fig. 5). The
recognition of ventricular hemorrhage was easier in these
patients because of its extension into the large ventricles. In Subarachnoid Hemorrhage
one case, a subependymal source for the ventricular hem-
orrhage was not identified. At autopsy, hemorrhage was Sonography failed to detect subarachnoid hemorrhage,
noted over the body of the caudate nucleus. which was present at autopsy in nine cases and at CT in
four cases.

Grade IV Hemorrhage Appearance of Hemorrhage over Time

In three infants subependymal hemorrhage was associ- Follow-up sonograms were available in nine infants. Hem-
ated with ventriculomegaly and extension of the hemorrhage orrhage in the subependymal region diminished very grad-
 AJR:137, August 1981 INTRACRANIAL HEMORRHAGE IN PREMATURE INFANTS 249

Fig. 6.-Grade IV hemorrhage. A,

Coronal sonogram. Bilateral subepen-
dymal hemorrhage has extended into
brain parenchyma on left (arrows). B,
Coronal image 8 days later. Slight en-
largement of hemorrhage. Cystic area
has developed in left-sided parenchymal
hemorrhage, secondary to retraction. C,
Same day as B, at level of caudate nu-
cleus. Intraventricular and bilateral sub-
ependymal hemorrhage. D, Above level
of lateral ventricle. Parenchymal hemor-
rhage in frontoparietal region. T = tem-
poral horn of lateral ventricle, R = cystic
area.
American Journal of Roentgenology 1981.137:245-250.

ually. As has been previously reported with ventricular clot sisted ventilation [1 2]. In 90% of cases examined at autopsy
[1 J, subependymal blood developed cystic spaces with time the hemorrhage occurred in the subependymal germinal
in four of our nine cases, presumably secondary to clot matrix overlying the caudate nucleus adjacent to the fora-
retraction and liquefaction. In the three infants where intra- men of Monro, with the rest occurring in the area overlying
cranial hemorrhage was followed to resolution, clot could the body of the caudate nucleus [1 0, 13].
be detected for 33 days in two cases and 60 days in one In our experience areas of hemorrhage were, in all cases,
case. In those infants lost to further follow-up, hemorrhage more hyperechoic than the medium level echoes of the head
was still seen up to 29 days of age. of the caudate nucleus. On sagittal scans, the characteristic
location of subependymal hemorrhage prevented confusion
with the more posteriorly located choroid plexus, which is
also echogenic. Ventricular blood was also echogenic, and
Discussion
could be distinguished from choroid plexus in the frontal or
Intracranial hemorrhage is a major source of morbidity occipital poles. However, intraventricular and subependy-
and mortality in premature infants and is found at autopsy mal hemorrhage in the body of the caudate nucleus could
in 50%-70% of cases [10, 11]. In prospective surveys of not be distinguished from adjacent choroid plexus. Also
infants studied with CT, 44% of all infants weighing less subependymal hemorrhages projecting into normal-sized
than 1 ,500 g sustained intracranial hemorrhage [9]. This lateral ventricles can be difficult to distinguish from intraven-
incidence increases to 70% in those infants requiring as- tricular clots.
 250 MACK ET AL. AJR:137, August 1981

The major failing of our technique has been the detection 3. Skolnick ML, Rosenbaum AE, Matzuk T, Guthkelch AN, Heinz
of very small amounts of blood in normal-sized lateral yen- ER. Detection of dilated cerebral ventricles in infants: a correl-
tricles. Sagittal images, including the occipital pole of the ative study between ultrasound and computed tomography.
lateral ventricle, are essential. Axial imaging, with the in- Radiology 1979:131 :447-451
4. Haber K, Wachter RD, Christenson PC, Vaucher Y, Sahn DJ,
fant’s head in the lateral decubitus position to demonstrate
Smith LR. Ultrasonic evaluation of intracranial pathology in
a blood-cerebrospinal fluid interface has also been recom-
infants: a new technique. Radiology 1980;134:173-178
mended (Johnson ML, personal communication). Even with
5. Babcock DS, Han BK, LeQuesne GW. B-mode gray scale
these techniques, small amounts of intraventricular blood ultrasound of the head in the newborn and young infant. AJR
were not imaged in five of 1 2 cases. 1 980; 1 34 : 457-468
The recognition of intracranial bleeding has important 6. Dewbury KC, Aluwihare AP. The anterior fontanelle as an
clinical implications. Mortality is 28%-55% [1 4, 1 5]. In ultrasound window for study of the brain: a preliminary report.
survivors, hydrocephalus is present in 22%-44%. Limited Br J Radio! 198053:81-84
follow-up of survivors to age 2 years suggests that 50% will 7. London DA, Carroll BA, Enzmann DR. Sonography of ventric-
have significant neurologic sequelae, half of which are in- ular size and germinal matrix hemorrhage in premature infants.
AJNR 1980;1 :295-300
capacitating [16, 17]. While the degree of neurologic in-
8. Shuman WP, Rogers JV, Mack LA, Alvord EC Jr. Christie DP.
volvement increases with the extent of hemorrhage, further
Real-time sonographic sector scanning of the neonatal cran-
studies are required to confirm this relationship.
ium: technique and normal anatomy. AJNR 1981;2:349-356
Exclusion of intracranial bleeding by serial sonograms 9. Brustein J, Papile LA, Burstein A. Intraventricular hemorrhage
must direct clinicians to other diagnoses, such as infection, and hydrocephalus in premature newborns: a prospective
metabolic derangements, or necrotizing enterocolitis, in pre- study with CT. AJR 1979;132:631 -635
mature infants with unexplained apnea, seizures, falling 1 0. Leech RW, Kohnen P. Subependymal and intraventricular hem-
orrhages in the newborn. Am J Pathol 1974;77:465-475
American Journal of Roentgenology 1981.137:245-250.

hematocrit, or acidosis. In addition, intracranial hemor-

rhages that cannot be reliably diagnosed by sonography, 11. Volpe JJ. Neonatal intracranial hemorrhage. C/in Perinato/
such as subarachnoid hemorrhage or small extraaxial col- 1977;4 :77-102
1 2. Lee BC, Grassi AE, Schechner 5, Auld PA. Neonatal Intraven-
lections, must also be considered. In such instances, CT
tricular hemorrhage: a serial computed tomography study. J
may be indicated.
Comput Assist Tomogr 1 979;3 : 483-490
1 3. Hambleton G, Wigglesworth JS. Origin of intraventricular hae-
morrhage in the preterm infant. Arch Dis Child 1976;51 :651 -
ACKNOWLEDGMENTS 659

We thank Heio Eberhardt and Merle Robinson for technical 1 4. Papile L, Burstein J, Burstein A, et al. Incidence and evolution
assistance and Jacqueline Gilliam for assistance in manuscript of subependymal and intraventricular hemorrhage: a study of
infants with birth weights less than 1 500 gms. J Pediatr
preparation.
I 978;92 : 529-534
1 5. Ahmann PA, LaZzara A, Dykes FD, et al. Intraventricular hem-
orrhage in the high-risk preterm infant: incidence and outcome.
REFERENCES
Ann Neurol 1980;7:118-124
1. Johnson ML, Mack LA, Rumack CM, Frost M, Rashbaum C. B- 16. Papile L, Munsick G, Weaver N, et al. Cerebral intraventricular
mode echoencephalography in the normal and high risk infant. hemorrhage (CVH) in infants <1500 grams: developmental
AJR 1979; 133:375-381 follow-up at one year. Pediatr Res 1 979;1 3:528
2. Mack LA, Rumack CM, Johnson ML. Ultrasound evaluation of 17. Krishnamoorthy KS, Shannon DC, DeLong GR, et al. Neuro-
cystic intracranial lesions in the neonate. Radiology 1980;137: logic sequelae in the survivors of neonatal intraventricular
451-455 hemorrhage. Pediatrics 1 979;64 :233-237