AJR:174, February 2000

561

Sonographic Visualization of Neonatal Posterior Fossa Abnormalities Through the Posterolateral Fontanelle

OBJECTIVE

.

This study was performed to determine whether imaging through the poste-rolateral fontanelle in addition to the anterior fontanelle during neonatal cranial sonographyimproves diagnostic accuracy or examiner confidence in the diagnosis of neonatal posteriorfossa abnormalities.

MATERIALS AND METHODS

.

In 1995 we changed our protocol of neonatal cranialsonography to include imaging through the posterolateral fontanelle in all patients. The re-ports of all sonography performed in the first 15 months of this protocol were reviewed, andtwo radiologists reviewed the images of all patients in whom a posterior fossa abnormalitywas diagnosed with posterolateral fontanelle images masked and then with posterolateral fon-tanelle images available.

RESULTS

.

In total, 1292 sonograms were obtained in 462 patients. In 200 patients, thesonographic findings were abnormal; of these 200 patients, 24 (12%) had posterior fossa ab-normalities (nine posterior fossa hemorrhages, four Arnold-Chiari malformations (type II),two posterior fossa arteriovenous malformations, and nine partial vermian defects). The pos-terolateral fontanelle images showed the posterior fossa abnormality better than the anteriorfontanelle images did in 23 (96%) of the 24 patients, increased confidence in the diagnosis of18 (75%) of the 24 patients, and was the only technique to reveal the posterior fossa abnor-mality in 11 (46%) of the 24 patients. Nearly all pathologic correlations with imaging con-firmed the posterolateral fontanelle findings

except for the diagnosis of inferior vermianagenesis, which was presumed to be a false-positive diagnosis in four patients in whom MRimaging showed no abnormalities.

CONCLUSION

.

Additional imaging through the posterolateral fontanelle during routineneonatal cranial sonography added considerable benefit. False-positive diagnosis of vermiandefects is a troubling problem but may be avoided with careful attention to the midline sagittalsonographic images of the vermis and fourth ventricle.

onography of the neonatal brain is animportant tool in the assessment ofneonates, particularly premature ne-

onates at significant risk for intracranial hemor-rhage. The procedure has shown good sensitivityand specificity in the detection of neonatal intra-cranial abnormalities, particularly in the suprat-entorial region [1–3]. When scanning through theanterior fontanelle, the most poorly evaluated re-gion is the posterior fossa. This is largely becausethe posterior fossa is farthest from the transducerand because many of its structures are parallel tothe insonating beam [4–7]. Several articles haverecommended adjunctive imaging through poste-rior and posterolateral fontanelles to improve vi-sualization of the posterior fossa [4–8].

In 1995 we added posterolateral fontanelleangled axial images of the cerebellum to the

conventional anterior fontanelle images obtainedin all cranial sonography performed at our insti-tution. The purpose of this review of our first 15months’ experience is to determine the benefit ofadding posterolateral fontanelle imaging to theconventional anterior fontanelle examination.

Materials and Methods

The studies were performed in an academic refer-ral center with a 42-bed regional intensive care unitapproved for extracorporeal membrane oxygenation(ECMO) and pediatric cardiothoracic surgery.

We reviewed the reports of all of the cranial sono-grams obtained during the first 15 months of usingroutine posterolateral fontanelle sonography in addi-tion to our standard cranial sonography (February1995 through April 1996). Each patient whose sono-graphic report indicated a posterior fossa abnormal-ity was identified.

Joseph A. Luna

1

Ruth B. Goldstein

2

Received May 14, 1999; accepted after revision July 22, 1999.

1

Kaiser Permanente, 4647 Zion Ave., San Diego, CA 92120.

2

Department of Radiology, University of California, 505 Parnassus Ave., M-396, San Francisco, CA 94143-0628. Address correspondence to R. B. Goldstein.

AJR

2000;174:561–567

0361–803X/00/1742–561

© American Roentgen Ray Society

S

562

AJR:174, February 2000

Luna and Goldstein

Scanning was performed portably in the neonatalintensive care unit by experienced sonographers usingphased array multiple foci 5.0- or 7.0-MHz sectortransducers (XP; Acuson, Mountain View, CA). Dur-ing each cranial sonogram six to eight images were ob-tained through the anterior fontanelle in both thecoronal and parasagittal planes for a total of 14–16 im-ages, including images of the posterior fossa. After thisimaging, additional focused images of the posteriorfossa were obtained through the posterolateral fonta-nelle. The transducer was placed over the postero-lateral fontanelle (using the most accessible upsidefontanelle) by gently bending the superior portion ofthe pinna down to find the acoustic window of the pos-terolateral fontanelle. The transducer was then rotatedslightly using real-time imaging to obtain the appropri-ate angled axial image depicting the cisterna magna,cerebellar hemispheres, vermis cerebelli, fourth ventri-cle, tectal lamina, and pons. Four to six images were ob-tained in this plane from superior to inferior, generallyimaging from the incisura tentorii cerebelli through theinferior vermis. A normal cerebellum imaged throughthe posterolateral fontanelle shows the biconvex dorsalmargin of the cerebellar hemispheres, normal cerebellarvermis, and the fourth ventricle (Fig. 1A).

Almost all neonates were imaged through theposterolateral fontanelle during this study period.The only exceptions were some neonates on ECMO(vascular access catheter placement would not allowslight turning of the neck). The precise number ofECMO neonates not examined through the postero-lateral fontanelle could not be determined from thisretrospective review. No patient in our study with arecognized posterior fossa abnormality was exam-ined only through the anterior fontanelle.

The cranial sonograms of the patients with poste-rior fossa abnormalities were independently reviewedby two radiologists, unaware of clinical diagnosis, inthe following fashion: first, with the posterolateralfontanelle images masked, the conventional coronaland parasagittal images obtained through the anteriorfontanelle were reviewed. Note was made of the pres-

ence or absence of a posterior fossa abnormality; andwhen a posterior fossa abnormality was present, theexaminer was asked to make a specific diagnosis ifpossible (e.g., cerebellar hemorrhage, intraventricularor extraaxial hemorrhage, ventricular dilatation, ver-mian agenesis, or other abnormality). Next, the poste-rolateral fontanelle images were unmasked and theentire sonogram was rereviewed independently bytwo observers. Again, the presence or absence of pos-terior fossa abnormality was noted and a specificdiagnosis was sought. Finally, with all the imagesavailable, each examiner was asked to make an inde-pendent subjective assessment of which images betterdisplayed the posterior fossa abnormality and whetheradding the posterolateral fontanelle images enhancedthe confidence with which he or she made the diagno-sis of posterior fossa abnormality.

Clinical follow-up was obtained through discus-sion with the neonatologists, review of clinical charts,study of any additional imaging, and review of avail-able autopsy results to determine if findings on theposterolateral fontanelle images significantly alteredthe clinical diagnosis or prompted additional imagingstudies on the neonate (including sonography, MRimaging, or CT). Confirmatory studies or follow-upfor each patient were noted when available.

Results

During the 15-month review period, 1292 cra-nial sonograms were obtained in 462 patients.Gestational ages at birth in this patient populationranged from 24 to 42 weeks. In 262 patients, thecranial sonographic findings were reported to benormal and no further follow-up was obtained. In200 patients, the sonographic findings were inter-preted to be abnormal, and 24 patients (12%) hadsonographic abnormalities of the posterior fossa.Of these 24 patients, 11 were term neonates and13 were preterm (<36 weeks’ gestational age atbirth). Eight (33%) of the 24 neonates were less

than 32 weeks’ gestational age at birth. Birthweights in the neonates with posterior fossa ab-normalities ranged from 915 to 4000 g.

Posterior Fossa Abnormalities

Of the 24 patients with posterior fossa abnor-malities, nine had posterior fossa hemorrhage:six had isolated posterior fossa hemorrhage, onehad hemorrhage as a result of posterior fossa du-ral arteriovenous malformation, and two hadfourth ventricular and extraaxial blood associatedwith marked supratentorial subependymal hem-orrhage (SEH) and intraventricular hemorrhage(grade III SEH). Of the remaining patients, fourhad findings related to the Arnold-Chiari malfor-mation II (small posterior fossa, effaced cisternamagna), two had posterior fossa arteriovenousmalformations without hemorrhage, and nineneonates were thought to have partial (inferior)vermian defects, or inferior vermian agenesis.

Images of the Lesion

The examiners thought the posterolateral fon-tanelle images displayed the posterior fossa ab-normality better than the anterior fontanelleimages in 23 of the 24 patients. The single abnor-mality better seen with anterior fontanelle imageswas an extremely small and distorted posteriorfossa related to an Arnold-Chiari malformationII. The severely distorted posterior fossa ob-scured the usual posterior fossa landmarks, andthe conventional anterior fontanelle images weremore easily interpreted in this patient.

The posterolateral fontanelle images in-creased confidence in the diagnosis of abnor-malities in 18 (75%) of the 24 unhealthypatients (Figs. 1–3). These 18 neonates in-cluded eight of the nine patients with posteriorfossa hemorrhage (Figs. 1 and 2), including

C

Fig. 1.—5-day-old premature male neonate born at 25 weeks’ gestation.A, Sonogram obtained through posterolateral fontanelle shows normal cerebellar hemispheres (c), echogenic cerebellar vermis (curved arrow), and fourth ventricle (straightarrow). L = left.B, Sonogram obtained 3 days later through anterior fontanelle approach shows possible left hemorrhage (arrow).C, Posterolateral fontanelle image obtained on same day as B shows unequivocal left cerebellar hemorrhage (arrows). lt = left, rt = right.

BA

Sonography of Neonatal Posterior Fossa Abnormalities

AJR:174, February 2000

563

Fig. 2.—3-day-old premature femaleneonate born at 27 weeks’ gestation.Posterolateral fontanelle imagesmore sensitively detect cerebellarhemorrhage. A, Anterior fontanelle images do notshow posterior fossa hemorrhage.R = right. B, Posterolateral fontanelle imagesin same patient, same day, showunequivocal echogenic unilateralhemorrhage (arrow) in cerebellum.L = left, R = right.

BA

C

Fig. 3.—8-day-old male neonate in whom cerebellar hemorrhage is better seen on posterolateral fontanelle images.A, Coronal image obtained through anterior fontanelle fails to show cerebellar abnormality (arrow).B, Midline sagittal image obtained through anterior fontanelle shows questionable cerebellar abnormality (arrow).C, Posterolateral fontanelle image clearly shows bilateral cerebellar hemorrhages (arrows). Neonate died ofcomplications of necrotizing enterocolitis and did not undergo either autopsy or MR imaging.

BA

564

AJR:174, February 2000

Luna and Goldstein

both patients with fourth ventricle hemor-rhage associated with Grade III SEH (Fig.4), one patient with Arnold-Chiari malforma-tion II, and all nine neonates with suspectedinferior vermian agenesis (Figs. 5 and 6).

In five (21%) of the 24 patients, althoughthe abnormality was better seen on the poster-olateral fontanelle images, the improved vi-sualization was not considered to increaseexaminer confidence in diagnosis becausethe lesion was already adequately displayedby the anterior fontanelle images. These fiveabnormalities included both posterior fossaarteriovenous malformations, two small pos-terior fossae related to the Arnold-Chiarimalformation II, and one 5-cm posteriorfossa hemorrhage. In one of the 24 patients,the abnormality of the posterior fossa wasbetter visualized using conventional anteriorfontanelle imaging.

In 11 (46%) of the 24 patients, one or bothexaminers saw significant abnormalities exclu-sively on the posterolateral fontanelle images.These abnormalities included three cases of cer-ebellar hemorrhage, both cases of fourth ven-tricular and extraaxial hemorrhage, and six ofthe nine cases of suspected inferior vermianagenesis. Confirmatory studies were obtained in

six of the 11 cases: two of the three posteriorfossa hemorrhages were confirmed with au-topsy or MR imaging; the other was monitoredwith serial sonograms showing evolution andgradual resolution of the lesion. In two cases(posterior fossa extraaxial and intraventricularhemorrhage) the abnormalities were not con-firmed with additional imaging but were associ-ated with large supraventricular subependymaland intraventricular hemorrhages (grade IIISEH) (Fig. 4). Of the six cases of possible orprobable inferior vermian agenesis detectedonly on the posterolateral fontanelle images(Figs. 5 and 6), four were evaluated with MRimaging. Three were shown to have a normalvermis cerebelli (false-positive sonographic di-agnoses) on MR imaging and one showed infe-rior vermian agenesis.

In total, follow-up MR images, autopsy, orfollow-up sonography was available for 16 ofthe 24 unhealthy neonates. Six neonates withposterior fossa hemorrhage underwent confir-matory studies (MR imaging, three; autopsy,one; and serial sonograms, two). One neonatewith posterior fossa hemorrhage died ofcomplications of necrotizing enterocolitis anddisseminated intravascular coagulation anddid not undergo either MR imaging or au-

topsy (Fig. 3). The two patients with posteriorfossa dural arteriovenous malformations hadconfirmed diagnoses with MR imaging. Onepatient

died at 13 days of life from complica-tions related to a large myelomeningocele anddid not undergo MR imaging. Nine neonateswere thought to have inferior vermian agene-sis as a result of an unusually large communi-cation between the fourth ventricle and thecisterna magna. MR imaging was performedin six of these neonates and showed inferiorvermian agenesis in two neonates and normalcerebellum in the other four. Neither MR im-age nor autopsy was performed in the otherthree neonates thought to have inferior ver-mian agenesis.

Altered Diagnosis

Adding the posterolateral fontanelle im-ages to our standard cranial sonograms al-tered diagnosis or prompted additionalevaluation (including imaging studies) in 12of the 24 patients identified with posteriorfossa abnormalities. Three posterior fossahemorrhages and five suspected inferior ver-mian ageneses were visible only on the pos-terolateral fontanelle images. In three otherposterior fossa hemorrhages and one sus-

Fig. 4.—Large intraventricular and extraaxial hemorrhage in 7-day-old premature male neonate born at 32 weeks’ gestation.A, Posterolateral fontanelle image shows extraaxial hemorrhage (thick arrow and cursors) and clot in fourth ventricle (thin arrow). B, Anterior fontanelle image shows neither of the findings in posterior fossa (arrow ) seen in A.

BA

Sonography of Neonatal Posterior Fossa Abnormalities

AJR:174, February 2000

565

Fig. 5.—False-positive sonographic diagnosis of inferior vermian agenesis in 4-day-old male neonate.A, False-positive inferior vermian agenesis (arrow) shown on posterolateral fontanelle image. B, Midline sagittal image obtained through anterior fontanelle shows intact vermis (curved arrows), retrospectively appreciated, covering fourth ventricle (straight arrow).MR imaging showed normal cerebellum.

BA

Fig. 6.—1-day-old male neonate with inferior vermian agenesis.A, Posterolateral fontanelle image shows inferior vermian defect (arrow) that was associated with agenesis of corpus callosum (not shown). B, Midline sagittal anterior fontanelle image reveals possible inferior vermian agenesis. Note poor covering of fourth ventricle (long arrow) by vermis (curved arrows). Also notethat partial voluming of small part of contralateral cerebellar hemisphere (small arrow) can be mistaken for intact vermis. Inferior vermian defect was confirmed on MR imaging.

BA

566

AJR:174, February 2000

Luna and Goldstein

pected inferior vermian agenesis the conven-tional images showed an equivocal area thatwas interpreted as definitely abnormal on theposterolateral fontanelle images.

Interobserver Agreement

Good interobserver agreement was seen. Theexaminers agreed with the diagnosis and “im-provement” or “no improvement” of the postero-lateral fontanelle image for the diagnosis in 23 ofthe 24 neonates. The examiners differed in theconfidence of the findings on the anterior fon-tanelle images in only one patient. One examinerinterpreted the images as having normal findings;the other interpreted the image as showing a“possible abnormal area.” Both examiners inter-preted this patient’s posterolateral fontanelle im-ages as showing posterior fossa hemorrhage.

Discussion

Cranial sonography of the neonate is awidely accepted technique for evaluating theneonatal brain. Initial reports describing thetechnique stressed imaging through the anteriorfontanelle [9, 10]. Imaging through the anteriorfontanelle allows excellent evaluation of thecommon sites of germinal matrix hemorrhageand the cerebral ventricles, but a weakness ofanterior fontanelle imaging is its evaluation ofthe posterior fossa [4, 5, 7]. Taylor et al. [2] andBabcock et al. [3] described neonates in whomsignificant posterior fossa hemorrhage wasmissed on sonography performed exclusivelythrough the anterior fontanelle.

It has recently been hypothesized that bring-ing the high-frequency transducer closer andmore perpendicular to many of the posteriorfossa structures should improve image clarity.Several recent studies have confirmed im-proved visualization of the normal posteriorfossa using imaging through a variety of poste-rior acoustic windows, including the foramenmagnum [4], the posterior fontanelle [5, 7],and the posterolateral fontanelle [8, 11].

The technique for obtaining imagesthrough the posterolateral fontanelle islearned easily and quickly. After only a few“learning cases” the images may be obtainedin less than 5 min, and our sonographers con-sistently obtain the images in less than 2 min,usually in less than 1 min.

On the basis of these early reports and ourown experience, we added routine posterolat-eral fontanelle imaging to our standard cranialsonography. Posterior fossa anatomy is un-equivocally better displayed using the postero-lateral fontanelle, as are posterior fossa

abnormalities. In 23 (96%) of the 24 sono-graphically diagnosed posterior fossa ab-normalities, two examiners independentlyconcluded that the abnormality was better seenusing the posterolateral fontanelle images.Even if the posterior fossa abnormality couldbe identified on the anterior fontanelle images,the examiners in this study concluded that add-ing the posterolateral fontanelle images re-sulted in increased confidence in diagnosis in18 (75%) of the 24 cases.

In 11 (46%) of the 24 patients, the postero-lateral fontanelle approach allowed detectionof significant findings that were not clearlyseen on anterior fontanelle imaging, especiallyin the seven neonates with posterior fossahemorrhage. One hemorrhage was well seenwith anterior fontanelle imaging. The other sixwere poorly seen or not detected at all usinganterior fontanelle imaging, yet confidentlyidentified using posterolateral fontanelle imag-ing. Six of these hemorrhages were confirmedby autopsy (

n

= 1), MR imaging (

n

= 3), or fol-low-up sonography (

n

= 2). There were nofalse-positive diagnoses of posterior fossa hem-orrhage. Only three of the nine neonates withposterior fossa hemorrhage had died at the timeof this writing. Whereas large and catastrophichemorrhages have generally been described inthe literature, posterolateral fontanelle imagingallowed the detection of some relatively smallcerebellar hemorrhages in this study. Two largeautopsy studies of low-birth-weight prematureneonates have reported an incidence of posteriorfossa hemorrhage between 16% and 21% eventhough this hemorrhage is less common thansupraventricular and intraventricular hemor-rhage [12, 13]. In one of these studies [13] thefindings of cerebellar hemorrhage fell into twogroups: large hemorrhages destroying one thirdor more of the cerebellar parenchyma and smallhemorrhages no larger than 5 mm.

The observation of small cerebellar hem-orrhages using the posterolateral fontanelleapproach in premature neonates has recentlybeen described [14]. The clinical signifi-cance of these otherwise undiagnosed hem-orrhages is currently under investigation.

Unfortunately, better visualization ofstructures not well visualized in the past canlead to false-positive diagnoses when oneembarks on using this new technique. In thisstudy at least four false-positive diagnosesoccurred as a result of adding the posterolat-eral fontanelle images; all were misdiag-noses of possible inferior vermian agenesis(Fig. 5) that deserve mention. These errorswere made as a result of misinterpretation of

a small apparent communication between thefourth ventricle and the cisterna magna thatwas mistakenly thought to represent an infe-rior vermian defect. The erroneous diagnoseswere all made early in our study. This imag-ing pitfall has been described in the fetalsonography literature (angled axial and coro-nal images of the posterior fossa similar toour posterolateral fontanelle images are eas-ily and commonly obtained) [15]. Angled ax-ial images of the cerebellum can make aprominent vallecula appear similar to a par-tial (inferior) vermian defect. Barkovich etal. [16], who used MR imaging to investigateDandy-Walker syndrome, found that the nor-mally formed vermis may tilt forward fromits usual position and create the impressionof an inferior vermian defect on angled axialand coronal images. We believe that scrutinyof midline sagittal images of the vermis cere-belli obtained through the anterior fontanellemay serve to arbitrate in equivocal cases(Figs. 5B and 6B). A midline sagittal imageallows display of the precise cerebellar anat-omy because the nodulus of the vermis cov-ering the inferior roof of the fourth ventriclecan be better seen, indicating whether theinferior vermis is intact or deficient. This ex-perience emphasizes the caution withwhich the diagnosis of inferior vermianagenesis should be made.

In conclusion, ours is a retrospective study,and the overall sensitivity of the posterolateralfontanelle images for posterior fossa abnor-malities is not evaluated. However, in our re-view of 15 months of experience with theseimages, the posterolateral fontanelle imagesrevealed three posterior fossa hemorrhages notdetected with standard anterior fontanelle im-aging and confirmed three more that were onlysuspected on the standard images. Nearly ev-ery abnormality observed was considered bet-ter displayed on the posterolateral fontanelleimages, and adding the posterolateral fonta-nelle images allowed increased confidence indiagnosis in 75% of the posterior fossaabnormalities. Only 1 or 2 min of additionalscan time is needed to obtain these images.The potential pitfall of overdiagnosing inferiorvermian agenesis should be anticipated andeasily avoided.

References

1. Taylor GA, Fitz CR, Kapur S, Short BL. Cere-brovascular accidents in neonates treated withextracorporeal membrane oxygenation: sonographic–pathologic correlation.

AJR

1989

;153:355–361

Sonography of Neonatal Posterior Fossa Abnormalities

AJR:174, February 2000

567

2. Taylor GA, Fiktz CR, Glass P, Short, BL. CT ofcerebrovascular injury after neonatal extracorporealmembrane oxygenation: implications for neurode-velopmental outcome.

AJR

1989

;153:121–1263. Babcock DS, Han BK, Weiss RG, Ryckman RC.

Brain abnormalities in infants on extracorporealmembrane oxygenation: sonographic and CTfindings.

AJR

1989

;153:571–5764. Sudakoff G, Montazemi M, Rifkin M. The fora-

men magnum: the underutilized acoustic windowto the posterior fossa.

J Ultrasound Med

1993

;4:205–210

5. Maertens P. Imaging through the posterior fon-tanelle.

J Child Neurol

1989

;4 [suppl]:S62–S676. Anderson N, Fulton J. Sonography through the

posterior fontanelle in diagnosing neonatal intra-ventricular hemorrhage.

AJNR

1991

;12:368–3707. Anderson NG, Hay R, Hutchings M, Whitehead

M, Darlow B. Posterior fontanelle cranial ultra-

sound: anatomic and sonographic correlation.

Early Hum Dev

1995

;42:141–1528. Buckley KM, Taylor GA, Estroff JA, Barnewolt

CE, Share JC, Paltiel HJ. Use of the mastoid fon-tanelle for improved sonographic visualization ofthe neonatal midbrain and posterior fossa.

AJR

1997

;168:1021–10259. Shuman WP, Rogers JV, Mack LA, Alvord EC,

Christie DP. Real-time sonographic sector scan-ning of the neonatal cranium: technique and nor-mal anatomy.

AJR

1981

;137:821–82810. Grant EG, Borts FT, Schellinger D, McCullough

DC, Sivasubramanian KN, Smith Y. Real-time ul-trasonography of neonatal intraventricularhemorrhage and comparison with computed to-mography.

Radiology

1981

;139:687–69111. Yousefzadeh D, Naidich T. US anatomy of the

posterior fossa in children: correlation with brainsections.

Radiology

1985

;156:353–361

12. Grunnet ML, Shields WD. Cerebellar hemor-rhage in the premature infant.

J Pediatr

1976

;88:605–608

13. Martin R, Roessmann U, Fanaroff A. Massive in-tracerebellar hemorrhage in low-birth-weight in-fants.

J Pediatr

1976

;89:290–29314. Merrill J, Piecuch RF, Fell SC, Barkovich AJ, Gold-

stein RB. A new pattern of cerebellar hemorrhagesin preterm infants.

Pediatrics

1998

;102:62–6615. Laing FC, Frates MC, Brown DL, Benson CB, Di

Salvo DN, Doubilet PM. Sonography of the fetalposterior fossa: false appearance of mega-cisternamagna and Dandy-Walker variant.

Radiology

1994

;192:247–25116. Barkovich AJ, Kjos BO, Norman D, Edwards

MS. Revised classification of posterior fossacysts and cystlike malformations based on the re-sults of multiplanar MR imaging.

AJR

1989

;153:1289–1300