Initial experience of an investigational 3T MR scanner designed for use on neonatal wards

Magnetic resonance (MR) imaging has revolutionised the assessment of the brain in clinical practice and is the most accurate radiological method for detecting most brain pathologies in most age groups. The ability of MR imaging to produce high spatial and contrast resolution images without exposure to ionising radiation has ensured rapid clinical uptake. One group that has not benefited from MR imaging as much as others is neonates, in whom MR imaging of the brain is difficult. There is inherently poor contrast resolution in the neonatal brain due to its immature state of myelination and obtaining high-quality MR images in a non-sedated/non-anaesthetised baby is challenging because of movement.

A further problem is ensuring safe transfer of the baby from the neonatal units to the MR scanner. Many neonates who may benefit from MR imaging of their brain have unstable cardiovascular and respiratory function and transfer to the MR scanners (usually in a different part of the hospital or at a different hospital) introduces extra risk. The decision to perform a clinical procedure, including a diagnostic test, should be considered on the basis of a risk/benefit analysis and, for neonates, the decision is often not to perform MR imaging. In such cases, transfontanelle cranial ultrasound scanography (cUSS) is often relied upon as it can be done on the neonatal wards and has been used successfully over many years. These factors are well illustrated in a short educational video made at Sheffield Teaching Hospital by the Wellcome Trust [1].

It has been a long-term aim of our group to improve access to MR imaging of the brain in neonates and in this article we describe our initial experience of using a physically small, but high field MR scanner installed in the neonatal unit.

The investigational high-field (3-T) neonatal MR scanner MR employed in this study was designed and built by GE Healthcare, Milwaukee, WI, USA, and can be sited on non-reinforced office floors (weight 500 kg) in an area of approximately 6 m² including scan control and equipment rooms (Fig. 1). It is designed to have the full capability of a current clinical MR system but customised to scan neonatal brains. It was installed on the Neonatal Unit of our institution resulting in much shorter journeys for neonates to have a MR scan in comparison with the 300 m (including two lift journeys) to the existing MR facility in the radiology department. The 3-T static magnet is a closed loop cryogenic system with a low helium volume (30 l), which maintains the magnet at 4.2 K. The 5-Gauss fringe field is contained within the scanner room (2.0 m × 1.3 m) and has a field uniformity of 15 ppm within a 15-cm-diameter sphere at the iso-centre. The maximum gradient strength is 70 mT/m and the maximum slew rate is 300 T/m/s. dB/dt is kept below the FDA guidelines of 300 mT/m/ms by software and hardware control. The RF power of the transmitter is set at 4 kW to allow maximum SAR levels in the range of 2W/kg for babies weighing up to 6 kg. The magnet bore has 28.0 cm diameter and 50 cm length, the gradient bore is 21.8 cm and the internal diameter of the quadrature transmit-receive RF coil is 17.9 cm, which is the only coil available at present. The magnet, gradients and RF systems are linked to a clinical ‘front-end’ running software at release level DV25.0.

Fifty-four subjects were consented but in one case consent was withdrawn before the MR study was performed, and in another case, MR scanning was attempted but no diagnostic images were obtained because of equipment failure. Of the 52 babies with successful MR imaging, 60% were born at term and approximately 40% were born before 37 weeks GA. The median corrected gestational age at the time of the MR study was 39 weeks (interquartile range 36-40, full range 34-43 weeks GA), median weight was 2.8 kg (interquartile range 2.4-3.4 kg, full range 1.3-4.5 kg) and median head circumference 34.0 cm (interquartile range 32.6-35.8 cm, full range 28.3-39.0 cm). There was one reportable adverse event—a term baby with a pre-existing baseline bradycardia was recognised as having bradycardia during scanning. A 12-lead ECG study had confirmed normal sinus rhythm previously and the bradycardia was judged not to be attributable to the MR scan so the baby was discharged on the day of the scan and did not meet the local criteria for follow-up.

The MR imaging studies were considered to be evaluable in 52/52 (100%) subjects, of which 22/52 (42%) were fully evaluable and 30/52 (58%) were partially evaluable (Figs. 3 and 4). The image quality was judged to be at least comparable to the clinical 1.5-T scanner in 47/52 (90%) of cases and better in 34/52 (65%). Similar results were found for image contrast, signal-to-noise ratio, tissue contrast and homogeneity. The worst results were for artefacts, which were thought to be worse than expected at 1.5 T in 13/52 (25%) of cases (Table 2) and were predominantly due to subject movement.

Two babies had unexpected intracranial findings on their MR examinations that required input from the neonatology staff:

We have demonstrated that a physically small, 3.0-T MR scanner is a feasible option for imaging the neonatal brain on a neonatal unit without the need for additional major structural alterations. There are other methods for improving access to neonatal MR imaging such as MR-compatible transport incubators, which reduces the number of times the baby has to be handled during transfer and provides a controlled physical environment during scanning. The MR compatible transport incubator has a built-in MR coil allowing the baby to go into the scanner while in the incubator thereby reducing the number of handling events. Our group was involved in the early prototype trials of one such transport incubator system [3] and we have designed and built our own MR-compatible incubator [4], which is used clinically.

We maintain that an MR scanner on the neonatal unit is a fundamentally superior approach. It is generally not possible to install whole-body MR scanners on the neonatal unit in most hospitals because of the space required and floor loading. Hospitals with a whole-body MR scanner installed in the NICU have usually done so by designing a new department. Other groups have approached the problem by developing MR services optimised to meet the needs of imaging neonates on standard clinical MR scanners (1.5 T and 3 T) by using tailored equipment and/or care measures [5, 6]. There have also been other attempts to construct small footprint systems for neonatal imaging. Our group previously reported experience of using a 0.2-T permanent magnet system for neonatal imaging [7]. More recent developments have used a 0.023 T system [8] and the US FDA has recently cleared a small, 1.0-T neonatal MR system for neonatal use [9]. We came to the conclusion, based on our experience of neonatal imaging at 0.2 T, that the detailed requirements for neonatal brain imaging (e.g. spectroscopy, diffusion imaging, MR angiography) include at least 1.5 T. A similar view was reached by Tkach who successfully developed a small-footprint 1.5-T system sited on a neonatal unit that can image several anatomical areas [10]. They have subsequently reported their experience of imaging 492 premature neonates [11]. We believe that the theoretical improvements in the signal-to-noise ratio provided by imaging at 3 T offer the best potential imaging for the neonatal brain because of the requirements of high-quality multi-sequence imaging/spectroscopy in as short a time as possible. It should be noted that the US FDA guidelines confirm that MR devices with main static magnetic fields of ≤ 4.0 T should be classified as a ‘non-significant risk’ for neonates [12]. There are disadvantages of scanning at 3 T, however, such as increased SAR, increased acoustic noise and more artefacts arising from, for example, susceptibility effects and field homogeneity problems. The problem with artefacts is compounded by the small size of the investigational 3-T neonatal scanner and our clinical readers considered MR artefacts to be a concern in 25% of cases (although all cases were considered to be of diagnostic quality).

The safety of the babies was of paramount importance and required close collaboration between the MR and neonatal unit staff involved in transfer and scanning so that rapid assessment and transfer of babies back to the wards could be made if necessary. This was not required in any of the 52 babies scanned in this study although a consultant-level opinion was sought in three cases (one baby with a bradycardia and two babies with unexpected intra-cranial findings). The increased acoustic noise associated with using a 3-T MR scanner was successfully managed by the use of two types of ear protection. One of the major factors that influences image quality in neonatal imaging is movement of the baby and we have taken a feed and swaddle approach, which appears highly beneficial for settling the babies. Only 1 of 52 of the scans was stopped early because of subject movement and that examination was considered to be partially evaluable by the assessors. Close physiological monitoring of the babies did not reveal any adverse effects leading to problems with desaturation or thermoregulation. Our results showed no problems with temperature control and support the findings of Cawley et al. who measured the core temperatures of neonates during scanning at 3 T and found no significant effects [13]. We must stress, however, that the babies included in this study were “physiologically stable”, which may explain the absence of serious adverse events during the scan procedure. Similar studies are required in the future when acutely unwell and particularly low-birth-weight premature babies will be scanned.

At the planning stage of this study we were prepared to take image quality that was ‘no worse’ than that obtained from our routine clinical scanning on a 1.5-T system as an acceptable outcome. This ‘equivalence’ approach was justified by the extra advantages in terms of safety by having the investigational 3-T neonatal scanner on the neonatal unit. Our results show that the image quality was at least equivalent to 1.5 T in the majority of cases (90%) and were judged to be better than the 1.5-T images in 65% of cases. The investigational 3-T neonatal scanner has the full range of MR sequences that we would expect to use in the clinical environment, all of which produce high-quality images. The next stage of the programme is to see if this can be reproduced in babies that have acute brain pathology. Once CE marking is obtained we will start to perform standard-of-care neuroimaging on that scanner. We will judge whether the optimised sequences developed through the study demonstrate the pathology or if further scan parameter optimisations are necessary.

In conclusion, we describe a new class of high-field MR scanner designed specifically for imaging neonates that is small enough to be sited on neonatal units. Our initial assessments of safety and technical performance are favourable and the next stage is to move on to evaluate diagnostic performance and diagnostic/clinical impact.