- Top of page
Technical developments in the field of magnetic resonance imaging (MRI), particularly the possibility of obtaining ultrafast sequences, have made the study of the fetal brain possible and studies on the normal anatomy and different pathologies have been published1–5.
Most of these papers claim that MRI offers new diagnostic capabilities beyond ultrasound resolution and therefore it is more accurate in the evaluation of central nervous system (CNS) anomalies. The only study that attempted to statistically analyze the results obtained by MRI vs. ultrasound, in a relatively small number of patients, did not demonstrate any significant advantage of one technique over the other6.
We have recently commented on the possible biases of some of the published papers comparing ultrasound and MRI7. In the present study we tried to overcome some of these biases in order to evaluate accurately whether brain MRI adds useful clinical information to dedicated neurosonography, in fetuses with suspected brain anomalies.
- Top of page
In the 2-year period between January 2000 and January 2002, 42 patients underwent concomitant neurosonographic and MRI examinations of the fetal brain because of suspected anomalies. The neurosonographic examinations were performed in the setting of the Fetal Neurology Clinic and the MRI studies were performed at the Pediatric Radiology Unit in 39 fetuses; three MRI studies were performed in another institute. Some of these patients have been previously reported8–10.
All patients were referred after an examination by a senior ultrasonographer (in most of the cases by the director of the Obstetrics–Gynecology Ultrasound Unit of the referring facility). The referral indications were: asymmetric ventriculomegaly (13), ventriculomegaly (7), periventricular cysts (2), suspected midline findings (7), agenesis of the corpus callosum (3), infratentorial pathology (3), cytomegalovirus (CMV) infection (2) and miscellaneous indications (5).
Ventriculomegaly was defined as ventricular width at the atrium between 10 and 15 mm. Asymmetric ventriculomegaly was defined as one ventricle smaller than 10 mm and the second between 10 and 15 mm.
The ultrasound examination was performed at a mean pregnancy age of 29.3 (range, 22–37.5) weeks. The MRI examination was performed at a mean pregnancy age of 30.2 (range, 23–37) weeks.
In 30 patients the neurosonographic examination preceded the MRI and in the remaining 12 it followed it. The elapsed time between the examinations was less than 1 week in 15 patients, between 1 and 2 weeks in 12 and more than 2 weeks in 15 (in five patients the MRI was postponed, due to the lack of significant findings in the neurosonographic examination).
All examinations were performed with prior knowledge of the referral ultrasonographic findings and the results of either neurosonography or MRI.
The ultrasonographic examinations were performed with an HDI 3000 (Advanced Technology Laboratories, Bothell, WA, USA) machine with 2–4- and 5–7-MHz transabdominal probes and a 4–8-MHz transvaginal probe. The same examiner (G.M.) performed all the examinations and the Fetal Neurology Clinic team provided the final diagnosis and counseling.
All patients underwent a transabdominal targeted ultrasound examination. The brain was studied using transabdominal and/or transvaginal axial, coronal and sagittal planes as previously described11, 12. The detailed study of the fetal brain included visualization of: the cerebral cortex, including sulci and gyri, germinal matrix, basal ganglia, corpus callosum and cavum septi pellucidi, lateral ventricle size and shape, third ventricle, cerebellar hemispheres and vermis, fourth ventricle and cisterna magna, and extra-axial spaces.
Fetal MRI studies were performed using a 1.5T system (General Electric Medical Systems, Milwaukee, WI, USA). Following a localizing gradient-echo sequence, ultra T2-weighted, single-shot, fast-spin echo MRI images were collected according to fetal position in the axial, coronal and sagittal planes (TR/TE, infinite/90; bandwidth 32 KHz; field of view, 16 × 28 cm; matrix, 256 × 192; slice thickness, 3–5 mm; gap 0–1 mm; number of excitations, 0.5). A torso-phased array coil was used. An experienced pediatric radiologist (L.B.S.) analyzed the MRI images and a description of the same structures observed by ultrasound was obtained.
The referral ultrasonographic diagnosis was compared to the diagnoses obtained by neurosonography and MRI. The neurosonographic and MRI diagnoses were compared to the final diagnoses as obtained by autopsy (following termination of pregnancy (TOP), eight patients; following intrauterine fetal death, one patient), neurological examination combined with neuro-imaging (ultrasound, 29 patients; computed tomography (CT), one patient; MRI, one patient) or neurological examination alone (one patient). In one patient TOP was performed but an autopsy was not done. Postnatal CT and MRI were performed only when clinically indicated.
The data were analyzed using SPSS 9.0 Analytical Software (SPSS Inc., Chicago, IL, USA, 1999). Sensitivity, specificity and positive (PPV) and negative (NPV) predictive values were calculated for each method. Intermethod agreement was assessed by calculating the kappa value. Cohen's kappa measures the agreement between the evaluations of two raters (or two methods) when both are rating the same object. A negative kappa value indicates that the raters are actually rating in the opposite direction while a positive kappa value indicates that the raters are concordant. All tests were considered significant at P < 0.05.
- Top of page
Neurosonography and MRI had higher sensitivity, specificity, PPV and NPV values than the referral ultrasound. Neurosonography performed slightly better than MRI: sensitivity 96% vs. 85%, specificity 87% vs. 80%, PPV 93% vs. 88% and NPV 93% vs. 75% (Table 1).
Table 1. Sensitivity, specificity, positive (PPV) and negative (NPV) predictive values of ultrasound, neurosonography and magnetic resonance imaging (MRI) in predicting postnatal outcome
|Method||Sensitivity (%)||Specificity (%)||PPV (%)||NPV (%)|
When the referral ultrasound was compared to either neurosonography or MRI the kappa value was negative indicating that the two compared methods were not concordant. When comparing neurosonography to MRI the diagnoses showed a positive kappa value indicating excellent agreement. Excellent agreement was also found between both neurosonography and MRI to the postnatal diagnosis (Table 2).
Table 2. Measure of agreement (Cohen's kappa) between methods
|Referral ultrasound/postnatal outcome||−0.244||0.18|
|Neurosonography/postnatal outcome||0.842||< 0.0001|
|MRI/postnatal outcome||0.642||< 0.0001|
The diagnoses obtained by neurosonography and MRI were similar in 29 (69.1%) fetuses (Table 3). Both examinations were considered normal in 10 fetuses. This group included fetuses referred for evaluation because of suspected midline anomalies (4), suspected non-CNS anomalies (3), suspected incomplete vermian agenesis (2) and small head circumference (1). The suspected midline anomalies turned out to be a larger than usual cavum septi pellucidi or cavum vergae in three fetuses and a septal vein in one fetus. The 10 delivered newborns were healthy and the postnatal neurological and ultrasound examinations were normal. The pediatric neurological examination was normal at 7–18 months.
Table 3. Referral ultrasound diagnosis as compared to concordant neurosonographic and magnetic resonance imaging (MRI) diagnosis and follow-up
|1||Suspected skeletal dysplasia, poor visualization of brain||Normal||Normal ultrasound + NE|
|2||Cardiac rhabdomyoma||Normal||Normal ultrasound + NE|
|3||Incomplete vermian agenesis||Normal||Normal ultrasound + NE|
|4||Suspected microcephaly||Normal||Normal ultrasound + NE|
|5||Non-defined midline finding||Normal||Normal ultrasound + NE|
|6||Intrauterine CMV, PCR+||Normal||Normal ultrasound + NE|
|7||Incomplete vermian agenesis||Normal||Normal ultrasound + NE|
|8||Non-defined midline finding||Normal||Normal ultrasound + NE|
|9||CSP echogenicity||Normal||Normal ultrasound + NE|
|10||Large third ventricle||Normal||Normal ultrasound + NE|
|11||AV||AV||AV + normal NE|
|12||AV||AV||AV + normal NE|
|13||AV||AV||AV + normal NE|
|14||MV||MV||MV + normal NE|
|15||PVPC||PVPC||PVPC + normal NE|
|16||MV||AV||AV + normal NE|
|17||ACC||Isolated ACC||TOP, isolated ACC|
|18||AV||AV||AV + normal NE|
|19||Choroid plexus cyst||PVPC||PVPC + normal NE|
|20||Echogenic CSP||Pericallosal lipoma||Pericallosal lipoma + normal NE|
|21||AV||AV||AV + normal NE|
|22||AV||AV||AV + normal NE|
|23||MV||MV||MV + normal NE|
|24||AV||AV||AV + developmental delay|
|25||Cerebellar echogenicity||Cerebellar hemorrhage||IUD 34 weeks, PM confirmation|
|26||AV||AV||AV + normal NE|
|27||Ventriculomegaly, hypoplastic corpus callosum||Ventriculomegaly, hypoplastic corpus callosum||MV + normal NE|
|28||AV||AV||TOP, unilateral hydrocephaly|
|29||MV||AV||AV + normal NE|
Isolated asymmetric ventriculomegaly was present in 11 fetuses. Ten were delivered and the neurological examination was considered normal in nine of them at 12–24 months. One child suffers from severe developmental delay with dysmorphic features at 18 months. TOP was performed in one case because of progressive unilateral hydrocephaly identified in a follow-up scan at the referring center.
Mild ventriculomegaly was diagnosed in three fetuses; in one of them hypoplastic corpus callosum was suspected but this was ruled out after delivery. The three children had normal neurological examinations at ages 7, 11 and 12 months. Periventricular cysts, located in the caudo-thalamic groove, were found in two cases and were confirmed postnatally. Isolated agenesis of the corpus callosum was diagnosed in one fetus and the diagnosis was confirmed after the TOP. Another two fetuses, one with a pericallosal lipoma (normal neurological examination at 12 months) and one with a small cerebellar hemorrhage were diagnosed accurately by both techniques.
In seven (16.7%) patients neurosonography provided more accurate information than MRI (Table 4). In the first patient, MRI depicted an abnormally deep medial occipital sulcus suggesting a parenchymal anomaly, which was not corroborated by ultrasound or autopsy. In the second patient, MRI suggested abnormal focal periventricular signals that were not confirmed by ultrasound; neonatal ultrasound and neurological examination at 12 months of age were normal. In the third patient, MRI suggested the presence of a porencephalic cyst that was not found during prenatal and postnatal ultrasound; the neurological examination at 12 months of age was normal.
Table 4. Patients in whom neurosonography proved more accurate than magnetic resonance imaging (MRI)
|1||Suspected ACC||AV||31||AV, Suspected parenchymal abnormality||30||TOP, AV, no parenchymal abnormality|
|2||Discrepancy between HC and FL||Normal, small PVPC||37.5||Suspected parenchymal abnormality||36.5||Delivered, normal imaging and development|
|3||Asymmetric anterior horns||Asymmetric anterior horns||34||Suspected porencephalic cyst||34||Delivered, normal imaging and development|
|5||Brain and liver echogenicities||Findings consistent with CMV||24||MV||24||TOP, CMV|
|6||Echogenicity above corpus callosum||Pericallosal lipoma||35||Normal||35||Delivered, pericallosal lipoma, normal development|
|7||ACC||ACC, colpocephaly, IH cyst, migration disorder||26||ACC, colpocephaly, IH cyst||23||TOP, ACC, colpocephaly, IH cyst, migration disorder, partially fused thalamus|
Patient 4 (Table 4) was referred for MRI because of the presence of ‘atypical’ asymmetric ventricles at 22 weeks. These were considered atypical due to the fact that the asymmetry was most prominent at the level of the anterior horns. MRI performed at 24 weeks confirmed the presence of asymmetric ventricles. The neurosonographic examination at 25 weeks clearly demonstrated the presence of an abnormal parenchyma8 consistent with developing hemimegalencephaly. Another ultrasound examination performed at the referring hospital 4 weeks later showed only asymmetric ventricles, but at the same time a repeat MRI study demonstrated the hemispheric diffuse migration anomaly and callosal dysgenesis. TOP was performed at 31 weeks and confirmed the sonographic diagnosis of hemimegalencephaly. A retrospective analysis of the MRI images at 24 weeks depicted the presence of abnormal sulcation.
Patient 5 (Table 4) was referred for MRI because of the presence of ventriculomegaly in association with periventricular, intraparenchymal and thalamic hyperechogenic signals consistent with CMV infection as demonstrated by neurosonography (Figure 1a). MRI failed to visualize the foci of calcifications (Figure 1b). The diagnosis was confirmed by the presence of CMV in the amniotic fluid and autopsy.
Figure 1. Patient 5 (Table 4). (a) Sagittal transvaginal ultrasound scan showing parenchymal foci of increased echogenicity consistent with calcifications (arrows). (b) Normal appearance of the brain parenchyma by magnetic resonance imaging.
Download figure to PowerPoint
Patient 6 (Table 4) was referred for MRI because of the visualization of a hyperechogenic structure in close contact with the corpus callosum extending into the choroid plexus. MRI of the brain was initially considered normal (Figure 2a). The diagnosis of pericallosal lipoma was made by neurosonography (Figure 2b) and confirmed by CT (Figure 2c) and ultrasound postnatally. Retrospective analysis of the MRI examination showed the presence of the lipoma.
Figure 2. Patient 6 (Table 4). (a) Coronal magnetic resonance imaging scan showing a small, hypodense area adjacent to the corpus callosum (arrow). (b) Coronal transvaginal ultrasound scan showing an echogenic structure (arrow) in close proximity to the corpus callosum (thin arrow). Note the extension into the lateral ventricle (arrowhead). (c) Axial postnatal computed tomography image showing a midline lipoma with extension into the lateral ventricle.
Download figure to PowerPoint
Patient 7 (Table 4) was diagnosed as suffering from agenesis of the corpus callosum at 22 weeks and was referred for MRI. The MRI examination at 23 weeks confirmed the diagnosis. Neurosonography, performed 3 weeks later, depicted additional findings: delayed gyral and sulcal development and hypoplastic vermis. These findings were confirmed by autopsy. Both techniques failed to diagnose the presence of a partially fused thalamus.
In three (7.1%) cases, MRI diagnosis was more accurate (Table 5). In one patient with asymmetric ventriculomegaly demonstrated by ultrasound, MRI showed progressive unilateral ventriculomegaly and large subarachnoid spaces on the same side suggesting parenchymal damage. In one patient with asymmetric mild ventriculomegaly, MRI showed normal ventricles in a follow-up examination performed 5 weeks later. In one patient with bilateral mild ventriculomegaly and suspected dilatation of the third ventricle by neurosonography (Figure 3a, 3b), MRI depicted bilateral mild ventriculomegaly with a normal-appearing third ventricle and without associated parenchymal anomalies (Figure 3c).
Figure 3. Patient 3 (Table 5). (a) Coronal transvaginal ultrasound scan showing mild bilateral ventriculomegaly with concurrent dilatation of the third ventricle. (b) Sagittal transvaginal ultrasound scan showing a large third ventricle (3V). (c) Coronal magnetic resonance image showing mild bilateral ventriculomegaly, however the third ventricle is not enlarged.
Download figure to PowerPoint
Table 5. Patients in whom magnetic resonance imaging proved more accurate than neurosonography
|1||AV||AV||25||Ventriculomegaly, large SAS||28||TOP, ventriculomegaly|
|2||AV||AV||22||Normal ventricles||27||Delivered, normal imaging and development|
|3||AV||MV, third ventricular dilatation||23||MV||25||Delivered, normal imaging and development|
In three (7.1%) patients the identified pathologies were differently interpreted; each examination provided another aspect of the anomaly or a definitive diagnosis was not possible (Table 6). In one patient a periventricular cyst was considered to be a periventricular pseudocyst at the level of the caudate by neurosonography and a cyst at the level of the centrum semiovale by MRI. The child is neurologically normal at the age of 12 months. Unfortunately, the parents refused the performance of any kind of diagnostic imaging. In a second patient, ultrasound diagnosed the presence of severe ventriculomegaly with large periventricular cysts and MRI depicted severely damaged white matter surrounding the ventricles. Amniocentesis was positive for CMV and the imaging findings were later confirmed in the autopsy. In a third patient who was referred because of suspected lobar holoprosencephaly, ultrasound raised the possibility of septo-optic dysplasia due to the lack of the septum pellucidum while MRI demonstrated bilateral ventriculomegaly with parenchymal damage. The pregnancy was interrupted but an autopsy was not performed. A retrospective evaluation of this case showed that the MRI diagnosis was probably more accurate.
Table 6. Patients in whom neurosonography and magnetic resonance imaging provided different aspects of the diagnosis or in whom the diagnosis was not confirmed
|Patient||Referral ultrasound||Neurosonography||MRI||Postnatal follow-up|
|1||Caudate cysts||PVPC||32||Centrum semiovale cyst||34||Normal NE, no imaging|
|2||Ventriculomegaly||Ventriculomegaly, PVPC||37||Ventriculomegaly, white matter anomaly||37||TOP, CMV|
|3||Lobar holoprosencephaly||Ventriculomegaly, suspected SOD||23||Ventriculomegaly, parenchymal damage||23||TOP, no PM|
- Top of page
The introduction of MRI for the study of the fetal brain is appealing and has been received with enthusiasm. Five large studies regarding fetal brain MRI have recently been published. Sonigo et al.1 reported their results on 400 prenatal MRI examinations, Levine et al.2 on 90, Wagenvoort et al.3 on 19, Simon et al.4 on 66 and Kubik-Huch et al.6 on 30. The conclusion of four of these studies1–4 was that MRI added valuable information not obtained by ultrasound. Kubik-Huch et al.6 published the only study in which a statistical analysis of the results was performed and they did not find a statistically significant difference between sonography and MRI. The claim that MRI is superior to ultrasound, particularly in the diagnosis of migration disorders, callosal anomalies and pathologies of the posterior fossa2, 4, 5, has not been scientifically proven. We have recently discussed the possible biases in the studies comparing ultrasound and MRI7: an overenthusiastic approach towards the new technique; a comparison between a referral routine transabdominal ultrasound examination and an MRI performed at a neuroradiological unit at a third-level facility; a detailed description of the MRI technique with minimal or no description of the ultrasound technique and the equipment employed; MRI images of excellent quality while the ultrasound pictures are either not supplied or are suboptimal; and the period between the index ultrasonographic examination and the MRI study, although crucial, is not specified.
Sonigo et al.1 published their experience with approximately 400 prenatal MRI examinations. Although a systematic comparison between MRI and ultrasound was not performed, the authors reached the conclusion that prenatal MRI is a valuable complementary tool when ultrasound is incomplete, doubtful or limited. Indications for fetal MRI were: cerebral anomalies detected by ultrasound in which the prognosis depends on the presence of associated anomalies not observed during the ultrasound examination and fetuses at risk of CNS diseases. The authors were not able to diagnose by ultrasound anomalies associated with agenesis of the corpus callosum such as neuronal migration anomalies, posterior fossa anomalies, cortical atrophy and microcalcifications.
Levine et al.2 reported on 91 patients who underwent both ultrasound and MRI examinations. They concluded that the MRI findings led to a change in the ultrasound diagnosis in 40% of the cases and enabled better prenatal counseling. A statistical analysis to prove this conclusion was not performed. The major problem with this study is that the authors were not able to diagnose by ultrasound agenesis of the corpus callosum, porencephaly, arachnoid cysts and cortical defects. This is a major weakness since these pathologies should be demonstrated, at least in some patients, by an ultrasound scan performed at a tertiary-level referral center.
Wagenvoort et al.3 reported on 41 prenatal MRI studies performed for diverse indications in patients with an equivocal ultrasound examination. Nineteen fetuses were evaluated because of suspected CNS pathology. All the studies were performed within 1 week of the referral ultrasound examination. MRI added information in 10 (52.6%) patients, a result similar to that obtained in our study when comparing the referral ultrasound scan and the information added by neurosonography or MRI. In three fetuses MRI helped elucidate the presence of the corpus callosum, in three to differentiate between an arachnoid cyst and a porencephalic cyst, in two to study the infratentorial structures, in one to differentiate between holoprosencephaly and hydrocephaly and in another one to rule out the presence of an occipital bone defect. MRI confirmed the ultrasound diagnosis in six patients, and produced false-negative and false-positive results in one patient.
In the present study we attempted to overcome the described biases: both examinations were performed at third-level facilities; the time elapsed between the examinations was reduced as much as possible and the cases in which more than 2 weeks had elapsed are discussed separately; the cases in which the MRI was performed as a follow-up examination are indicated, state-of-the-art examinations were performed with both techniques; the techniques and the employed equipment are described; and ultrasound and MRI images are presented for comparison.
In our study MRI did not prove superior to neurosonography. The possible explanations for this result, which differs from previously published studies, are as follows: (1) the neurosonographic examination was performed by a dedicated neurosonographer with knowledge in brain development and anatomy, and the interpretation of the findings was made by a multidisciplinary team that also included a pediatric neurologist and a geneticist, both experienced in brain malformations and neurogenetic syndromes; (2) at the beginning of the study there was a difference between the expertise of the neuroradiologist and the neurosonographer; and (3) the MRI and ultrasound were done in two different and independent facilities that did not rely on each other for refinement of the diagnosis.
We found that in 29/42 (69.1%) patients the interpretation of the dedicated neurosonographic examination and the MRI was similar. In 14 of these patients both the neurosonographic and MRI diagnosis were different from the referral one. Thus, if the patients would have only been referred for an MRI by the referring ultrasound unit it could have been falsely concluded that MRI provided additional information or a different diagnosis in 48.3% of these patients (Table 3, Patients 1–10, 16, 19, 20, 29).
A retrospective review of the MRI examinations in which the interpretation was discordant revealed that in 3/7 cases the initial MRI diagnosis was not accurate and it indeed demonstrated the same findings that were diagnosed by neurosonography (Table 4, Patients 4–6).
In all three cases where MRI was more accurate, MRI was performed more than 2 weeks following the ultrasound scan so it can be assumed that a follow-up ultrasound scan at the same time would have depicted the same findings since ventricular size is easily measured by ultrasound. In 2/7 patients where neurosonography was more accurate the examination was also performed more than 2 weeks after the MRI.
Our study demonstrates that MRI performed before 25 weeks' gestation may be misleading. In countries where TOP may be performed late in gestation it seems wise to postpone the MRI to 30–32 weeks in order to obtain maximal information.
Based on both previously published studies and the present study it is still difficult to define clear-cut indications for fetal brain MRI. The indications are influenced by the family's attitudes toward TOP. MRI is clearly indicated when TOP beyond 24 weeks is an option and the diagnosis by a dedicated neurosonographer is not possible or unclear. A possible indication is psychological reassurance of the parents confirming either pathology or normality by both techniques. Further research on specific pathologies such as mild ventriculomegaly, asymmetric ventriculomegaly and posterior fossa anomalies is expected to provide answers regarding clinical indications for fetal brain MRI. The advantage of performing a prenatal MRI in order to avoid general anesthesia after delivery has not yet been proven and it still seems that the results of postnatal imaging are superior.
We conclude that until large and well-controlled studies are performed it seems premature to state that fetal brain MRI is more accurate than dedicated neurosonography. It is clear that ideally both the neuroradiologist and neurosonographer have to be equally proficient in the fields of neuroanatomy and fetal brain development and in the technical aspects of their imaging modality. They should discuss the cases with each other in order to maximally refine the diagnosis and they must work with a team that includes a pediatric neurologist, a geneticist and a neuropathologist in order to provide the most accurate counseling to the parents.