Materials and Methods

This was a prospective study of patients referred for prenatal MRI. It was conducted from early January to late March 2012 in a tertiary referral center. All patients were offered to undergo an US examination including MRI-US fusion imaging. In France, all women routinely undergo routine US examinations at first, second, and third trimester. In case of a previous history of congenital malformation or whenever fetal abnormality is suspected, additional US examinations are usually offered at appropriate times of the fetal development. MRI examinations are performed in addition to US in the late second or early third trimester of pregnancy.

All fetal MRI examinations performed in this study used a 1.5 Tesla (T) unit (General Electric Signa 1.5 T; GEMS, Milwaukee, WI) and body phased-array coil. The women were placed in the supine or lateral decubitus position. No maternal sedative or contrast agents were used. The routine prospective MR imaging protocol was applied including at least 3 planes (coronal, sagittal, and axial) acquired on the whole fetus with steady state-free precession (SSFP) (true fast imaging with steady state precession: repetition time/echo time [TR/TE]: 4.9/2.5; matrix: 224 × 224; FlipAngle [FA]: 65°; signal average: 1; slice thickness: 4 mm; distance factor: 30%, field of view [FOV]: 1) sequences. Images were acquired relative to the head and trunk of the fetus. No institutional review board approval was required because this study had no impact on routine prenatal care but to add 1 US examination immediately after MRI. Information was given and oral consent was sought before all examinations.

All US procedures were performed by trained operators (L.J.S., J.P.B., Y.V.) with no time constraint using the same 2-dimensional probes and ultrasound machine (3.5-5 MHz curvilinear abdominal or 6-9 MHz vaginal transducer–General Electric LOGIQ E9; GE Healthcare, GE Medical System Europe-78, Buc, France). Skills required from the operators were those of performing appropriate US examinations and of scrolling through MRI volumes. The ability to obtain MRI images (L.J.S., P.S., A.E.M.) was not necessary to perform fetal fusion imaging between MRI and US because it was driven by RT ultrasound US performed on the same day as MRI. MRI and US being performed in different locations within the hospital, this required mobilization of the women between the 2 examinations. The time interval between MRI and US was up to 3 hours, depending on the availability of machines and operators.

The fusion imaging system (GE Healthcare) was composed of a position-sensing unit mounted on an US unit, a magnetic field transmitter, and 2 sensors connected to a transducer bracket ( Figure 1 ). The Digital Imaging and Communications in Medicine (DICOM) volume dataset from preacquired MRI examination was loaded into the ultrasound system (LOGIQ E9; GE Healthcare). Image registration was performed first. It is the process of transforming different sets of data into one coordinated system. It is necessary to be able to compare or superimpose the data obtained from the different imaging modalities. Registration was either done by defining a common plane and 1 additional common point or by defining at least 3 common reference points. The operator performing the fusion examination identified these reference points/planes and registration was performed on each fetus. Practically, this required visualizing a characteristic landmark or plane on US images and marking that point or plane. After marking that landmark, the US image was fixed and the corresponding landmark on reference MRI images was sought for by comparison with the fixed US image. Once the landmark point on the reference MRI image was also marked, the registration was completed. US and MRI images were therefore matched. Once RT US examination was being performed, the MRI volume was also being scrolled automatically by the system, showing the same planes using the Volume navigation system (V Nav; GE Healthcare). Volume navigation was achieved by placing an electromagnetic transmitter near the area of scanning and attaching a pair of electromagnetic sensors to the probe. Both the transmitter and the sensors were connected to a position sensing unit embedded in the US machine. The magnetic tracking system determined the position of moveable sensors relative to a fixed transmitter within a defined operating volume. The 2 moveable sensors attached to the probe bracket precisely measured the magnetic field from a transmitter configured to generate a known set of field patterns. These transmitted patterns were arranged for the system to define a unique spatial position and orientation from the values measured by each sensor. Therefore, the accurate position and orientation of the US probe within the 3-dimensional (3D) MRI volume could be determined precisely. The integrated navigation software then fused the images in real time. The image displayed during the procedure allowed for either side-by-side display or for an image overlay where US overlapped MR images. The 2 modalities could be combined with an adjustable proportion and the addition of a Doppler signal was also possible. Once the US was fused, the RT US scanning navigated through the virtual cross-sectional DICOM dataset flawlessly in any plane with real-time reconstruction that matched the RT data from the US. Three senior operators were trained to use this US machine (L.J.S., J.P.B., Y.V.).

Fusion imaging system

Composed of a position-sensing unit mounted on an US unit ( star ), a magnetic field transmitter ( cross ), and 2 sensors connected to a transducer bracket ( arrows ). The main steps are highlighted.

MRI, magnetic resonance imaging; US , ultrasound.

Salomon. MRI-US fusion imaging. Am J Obstet Gynecol 2013 .

Registration was considered successful whenever it was possible to define a common plane and 1 additional common point or defining at least 3 common points on both the DICOM volume dataset and the live US image ( Figure 2 ). Volume navigation was considered successful when it was possible to virtually rescan the fetus with modified angulated sectional images of the original DICOM volume dataset for at least 30 consecutive seconds. This was performed again whenever the registration was not satisfactory because of fetal movements. This process was repeated as often as necessary to obtain a satisfactory overlap between the US and the MRI set.

Examples of fused images displayed side-by-side

From case 21A and case 13B. A, Axial view of the fetal head, close to the biparietal plane. Note the appropriate registration of the 2 images, with the sylvian fissure ( arrow ), the lateral ( star ), and third ( cross ) ventricles. B, Profile of a fetus affected by Robin sequence. Note the micrognatia ( arrow ).

Salomon. MRI-US fusion imaging. Am J Obstet Gynecol 2013 .

Results

Over the study period, 59 fetal MRI examinations were performed for fetal or placental indications. Fifteen women declined the fusion examination for personal reason. Twenty cases were not included because no operator was available within a few hours from MRI. In total, 24 pregnancies underwent fusion imaging following 25 MRI examinations (cases 16 and 17 from the same pregnancy) at a median gestational age of 31 (range, 24–35) weeks. The primary indication for MRI was the suspicion of an anomaly of the fetal brain, thorax, gastro-intestinal tract, urinary tract or of the placenta in 12, 5, 4, 3, and 1 case, respectively ( Table 1 ).

Data registration and matching were feasible within 2 minutes in all cases. This did not change with gestational age. Registration was performed for each fetus. Registration landmarks varied depending on the anatomic region to evaluate and on the fetal position at both MRI and US examinations. In all cases, landmarks had to be clearly identified at both MRI and US. This was obtained by using standardized planes of the fetal head or trunk, such as biparietal diameter, 4-chamber view, abdominal circumference. Table 2 provides with examples of planes that could easily be defined on both imaging modalities and that were used for data registration in our cases. The additional common point(s) used for matching should also be clearly identified on both modalities. Examples of usable landmarks are given in Table 2 .

After appropriate registration, volume navigation was achieved in all cases. However, significant fetal movement required repeat-registration in 15 (60%) of cases. Repeat registrations are only time consuming but can be performed as often as required by fetal movements. Fusion imaging allowed superimposing MRI and US images therefore providing with real time imaging capabilities and high tissue contrast. Figures 2-5 show examples of fused images. It also allowed adding real time Doppler signal on MRI images ( Figures 4 and 5 ). The average duration of the overall additional scan with fusion imaging was 10 ± 5 minutes.

Example of fused images displayed side by side

Fusion imaging allowed to correctly identify the gallbladder ( arrow ) on magnetic resonance imaging as it appeared in T1 hypersignal (from case 3). The other longer hypersignal is related to meconium in the colon.

Salomon. MRI-US fusion imaging. Am J Obstet Gynecol 2013 .

Examples of overlaid images with Doppler signal (case 15)

A, Axial plane of the fetal head, showing the Doppler signal of the polygon of Willis ( arrow ) and the middle cerebral artery ( arrow ) superimposed on the magnetic resonance imaging image. B, Sagittal view of the fetal head, showing the Doppler signal of the pericallosal artery superimposed on the magnetic resonance image.

Salomon. MRI-US fusion imaging. Am J Obstet Gynecol 2013 .

Combined analysis of MRI and US plus Doppler in a case of pulmonary sequestration

Case 11, allowing definitive identification of the feeding vessel of the sequestration ( arrowhead ) branching from the aorta ( arrow ).

MR, magnetic resonance; US , ultrasound.

Salomon. MRI-US fusion imaging. Am J Obstet Gynecol 2013 .

In cases of brain indications (cases 1, 5, 6, 8, 10, 12, 15, 18, 20, 21, 24, and 25), US was a simple tool to analyze the normal anatomy as well as most abnormalities of the fetal CNS. However, the ossification of the cranial bones hampered appropriate evaluation of the hemisphere closest to the transducer especially in late gestation. In contrast, if the fetus was in a vertex presentation, the high-frequency transvaginal probe added valuable information. In cases with a suitable fetal position, 3D sonography and Doppler examination also helped in the examination of fetal brain. MRI overcame such technical difficulties. MRI allowed for a better evaluation of the posterior fossa, brain stem, corpus callosum, and cortical development than US alone. It was also possible to visualize small and specific structures such as optic nerves, olfactory bulbs or semi circular canal and to obtain cerebral biometry and evaluation of brain parenchyma. In thoracic fetal diseases, MRI permitted to evaluate the normal and abnormal lung and to estimate the lung volume as compared with the normal GA-adjusted volume. This was particularly relevant in cases of congenital diaphragmatic hernia (cases 16, 17, and 23) where MRI also offered better images to assess the size of the defect and to determine the organs involved. MRI also provided with a delineation of lung masses (cases 7 and 11). In contrast, US examination that allowed the superimposition of Doppler signal provided with useful information for the diagnosis of lung masses and helped locating the feeding vessels. In gastrointestinal tract anomalies (cases 3, 9, 13, and 14), US allowed to assess bowel peristaltic movements and helped locating the level of obstruction, especially in esophageal atresia. MRI was informative on bowel maturation by assessing the presence of intraluminal meconium. MRI was also useful in determining the level of the atresia, affecting either the esophagus or the small bowel (cases 3 and 9). In genitourinary tract anomalies, MRI was helpful to evaluate normality of the spine and to look for associated abnormalities of the pelvis or genitalia (cases 2, 4, and 19). However, US allowed for a quantitative evaluation of amniotic fluid and repeated assessment of bladder volume. Last, MRI was used once (case 22) for further delineation of a placenta accreta, as it helped obtaining a larger field of view and a better visualization of the posterior part of the uterus than when using US.