A dedicated neonatal brain imaging system

Neonatal Head Coil

To maximize SNR and support accelerated imaging, the coil should be as close fitting as possible and have as many channels as feasible, while retaining the load‐dominated condition 3. Given that infants across a range of postconceptual ages needed to be studied, a key parameter was the maximum head size to be accommodated. We therefore sought a robust estimator of the maximum diametric lengths for the oldest infants in the target age group and adopted the 95th percentile. Measurements were made of anterior–posterior (AP), right–left (RL) and inferior–superior (IS) diameters on previously acquired magnetization‐prepared rapid gradient‐echo (MPRAGE) brain images on 91 term‐born infants (age range at scan, 38.14–44.42 wk. The AP diameter was measured from the nasion to the occiput, on the mid‐sagittal slice (Fig. 1a, horizontal blue line). The RL diameter was determined by measuring biparietal diameter at the widest point on the mid coronal view (Fig. 1b, blue arrow) and the IS diameter was measured from the tip of the skull to the body of the third cervical vertebrae on the mid‐sagittal slice (Fig. 1a, vertical blue line). The resulting 95th percentile head dimension values taken for design purposes were: 145 mm (AP) × 120 mm (RL) × 130 mm (IS). Once other design factors relating to coil mechanics and infant handling were considered, the final coil elements were placed on a surface specified with each diameter increased by 10 mm. The final dimensions of the head coil were therefore 155 mm AP, 130 mm RL (Fig. 2b), and 140mm IS (Fig. 2c). The number of channels was set at 32, which was the maximum available on the standard interface of the MRI scanner being used. A 2‐cm² notch at the coil's open end was added to accommodate respiratory aids (Fig. 2c, green arrow).

Positioning Device

Our previous practice was to start with a sleeping infant, positioned within the adult head coil. Life‐monitoring devices (temperature, peripheral oximetry, electrocardiogram, and respiratory monitor) and auditory protection (earplugs molded from a silicone‐based putty (President Putty; Coltene/Whaldent, Mahwah, New Jersey, USA) placed in the external auditory meatus and neonatal earmuffs (MiniMuffs; Natus Medical Inc., San Carlos, California, USA) were then applied. No other sound attenuation device, such as an acoustic hood, was used. Swaddling and immobilization of the body was achieved using an air‐evacuated beanbag. Foam padding was then placed around the infant's head to prevent gross motion. These elements are essential, but the process was apt to disturb the infant, cause temperature loss during preparation and use up valuable sleep time. For the NBIS, we therefore sought to complete all preparation steps first, away from the MRI suite with the mother, a more favorable condition for settling the infant 1, 2. This introduced two challenges: 1) how to achieve the final required position within a small dedicated head coil in the MRI scanner bore and 2) how to ensure safe transport of the finally positioned and immobilized infant.

The solution was a rigid but lightweight protective “shell” (Fig. 2a(iii), 2d) that facilitates infant positioning, allows safe transfer of the infant that docks easily with the head coil (Fig. 2a(i)), and remains at the MRI device. Its purpose is to prepare and securely hold the infant, allowing the user to position the head deep into the head coil with virtually no disturbance to the sleeping infant.

The shell consists of a spheroidal headpiece (Fig. 2e) and v‐shaped base section to support the infants' body (Fig. 2d). The headpiece is shaped to precisely match the inner surface of the head coil so that correct positioning within the former guarantees correct positioning within the latter (Fig. 2f). By minimizing the thickness of materials used and using a rail system to achieve precise alignment (Fig. 2g), the distance from the inner surface of the headpiece to the coil elements was kept to <5 mm.

The v‐shaped base of the shell starts at 1 cm below the lower inner surface of the headpiece to allow space for a vacuum‐evacuated bag filled with polystyrene beads. When finally positioned, the inner surface of the vacuum bag molds to the infant's body, holding it in a fixed position, while the outer surface molds to the v‐shaped base, preventing lateral movement of the body independent of the head. To one side of the base there is a dedicated space to secure an oxygen saturation monitor and battery box (Fig. 2d, green arrow). This allows the user to manually lift the infant and shell with vital sign monitoring securely in place.

Once the infant is positioned and secured in the shell, safe handling is achieved by grasping the centrally located handlebar at the foot end of the shell and placing a hand under the headpiece, which has an anti‐slip matt surface. This allows the user to safely control the infant's head and body within the shell when lifting.

The coil and shell with the infant both sit on a heavy scanning base frame featuring a pair of rods that act as guide rails (Fig. 2g). Handles on either side of the frame allow it to be easily lifted on and off the scanner tabletop. The coil is free to slide along the rails between end stops and can be locked in place with a pair of cam fixings. The coil remains on the frame. The shell is placed onto the rails between its locating end stops. It has slots in its base so that it self‐centers when lowered into place (Fig. 2e, red arrow). When in position on the frame, the v‐shaped section has a 2 ° IS incline upward toward the head so that the infant does not lay completely flat, minimizing gastro‐esophageal reflux following feeding. All life monitoring can then be connected and checked for correct function. The head coil smoothly slides over the head section of the shell to a final preset fixed location, where it is locked into place. An acoustic hood is then placed over the assembly before the final movement to the magnet isocenter in the scanner bore (Fig. 2i).

Immobilization

Due to the close fitting nature of the headpiece of the shell, foam padding such as that used previously to immobilize the head could not be used. Therefore, specifically designed thin immobilization devices were created to consistently position the head within the headpiece and reduce head motion (Pearltec, Zurich, Switzerland). These devices are made from thin sheets of microbial‐resistant thermoplastic polyurethane that satisfy clinical infection control requirements. Each device consists of a bead‐filled pad that conforms to the shape of the infants' head (Fig. 3a, orange arrow) backed by an inflatable air pocket (Fig. 3a, 3b, 3d, black arrows) that flexibly fills any remaining gap to the headpiece. These are positioned over the ear acoustic protection (Fig. 3a, 3c, green arrows) to reduce RL motion, and are simultaneously inflated using separate hand pumps (Fig. 3b, red arrow) allowing fine control to ensure both immobilization and comfort. The side cushions are 14 mm thick when deflated and can expand to 23 mm. An additional cushion is placed under the shoulders below the base of the neck (Fig. 3b). Inflating this cushion allows the user to change the AP tilt of the head without handling the infant.

Transport

A dedicated MRI compatible trolley was created to transfer the positioning shell with the infant in situ to the MRI scanner bed (Fig. 4a). The skeletal nature of its design ensured that any non‐MRI compatible devices placed on the trolley would be clearly visible and identified for removal before entering the magnet room. The trolley has a pair of rails identical to those on the MR scanner table base frame, so that the shell is securely held in place (Fig. 4b), but can be transferred on and off the trolley without disturbing the infant. The rails are held in an articulated structure with a locking mechanism (Fig. 4b(i), 4b(ii)). When locked, the shell is held in a fixed position, but it can be released to allow the shell to be gently rocked forward and backward to pacify the infant. The trolley has soft rubber wheels with brakes that are held locked unless manually released by depressing handles for steering the trolley (Fig. 4a(i)). This ensures that when not in motion, the trolley is in a secure stop position.

Gradient Noise

To avoid sudden sound changes that might disturb the infant's sleep, the MRI scanner software was modified to introduce a user defined start‐up period for all scans (Gyrotools, Zurich, Switzerland). During the start‐up phase, the sequence gradients are played out repeatedly with amplitude increasing linearly in time from zero to the required operating point. No radiofrequency pulses are applied during this period. Tests performed with pediatric populations have shown that during MRI, a slow ramp‐up in acoustic noise is effective at avoiding a startle response 9. In this study, a 5‐s ramp‐up period was used.

Preparation of the Shell and Infant Positioning

Details of the preparation of the shell are shown in Supporting Video S1. The shell is placed on top of the trolley with an oxygen saturation‐monitoring box in situ. A vacuum bag and a neonatal slide sheet are positioned on top of the v‐shaped base. A specially designed cotton wrap sheet that contains a pocket extension with a pull tag at the head end is placed on top of the slide sheet. The pull tag is placed through the bottom hole of the headpiece (Fig. 2e). Soft, thin (0.5 cm) foam padding and the neck/shoulder cushion are then placed into the pocket of the wrap sheet.

Before feeding, the infant has cardiac monitoring devices and ear defenses in situ and a fiber optic temperature probe (InVivo, Best, Netherlands) fixed into its axilla with micropore (3M, Maplewood, Minnesota, USA). Once asleep, the infant is then placed on the wrap sheet, with its head placed on the rectangular pocket near the headpiece of the shell. The oxygen saturation probe is attached to the infant's foot. The infant is then wrapped and the immobilization pieces are gently held to the infant's head and is allowed to settle. The infant's head is then glided into the headpiece (Supporting Video S2) by pulling on the tag of the wrap sheet. A second person gently guides the head to ensure it is not bumped. This allows safe positioning with virtually no disturbance to the infant and no stress applied to its neck. The additional holes in the headpiece (Fig. 2e, 2f) allow visual confirmation that the head is positioned as close as possible to the top of the shell headpiece ensuring maximum SNR.

Once the infant's head is in position, the inflatable cushions under the base of the infant's head and over the acoustic ear protection are inflated until the head is in a neutral position so that motion is reduced. The body of the infant is then immobilized by gently wrapping the vacuum bag around the infant and applying suction.

Using the trolley, the infant is then transferred to the scanner bed and the shell with the infant on board is lifted and placed on the frame. The head coil is positioned over the headpiece and secured into place. Once vital sign monitoring is satisfactory, an acoustic hood (Fig. 2i) is placed over the shell and the infant is positioned into the scanner bore.

During the design process of the NBIS, there was also a review of what actions to take in the event that an infant wakes up. Previous practice was to try to resettle the infant by whatever means were appropriate at the time, which could include a gloved fingertip to suck on, or with parental permission, use of a pacifier and sucrose. In some cases the infant was removed from its position and resettled. With NBIS, a much more standard approach has been adopted. If an infant wakes up, the soothing devices described above are given but without any handling of the infant (i.e., the infant remains in situ).

SNR Evaluation: Phantom

An experiment was performed to assess the possible SNR gain of the 32‐channel neonatal coil compared with a 32‐channel adult receive array (Philips, Best, Netherlands). Both coils were used to scan a dedicated phantom designed to fit the neonatal coil. The phantom was 3D‐printed from nylon using selective laser sintering and was formed from two halves, each of which included a threaded section, so that they could be screwed together and glued to achieve a water‐tight join. A filling spout at the top could be plugged with a rubber bung containing an injection membrane for final top‐up using a syringe. Its outer shape matched the headpiece of the shell. It had 10‐mm‐thick walls, and the interior surface was varnished to ensure water‐tightness (Fig. 5(iii)). The phantom was filled with physiological saline (0.9%), doped with Gadovist to achieve a T1 of approximately 300 ms. For SNR evaluation of the adult coil, the phantom was placed in its geometrical center, supported by foam padding.

A 3D spoiled gradient echo sequence was acquired (FOV = 140 × 140 × 140 mm, resolution = 1 × 1 × 1 mm, TE = 6 ms, TR = 30 ms, flip angle = 30 °, bandwidth = 200 Hz/pixel). The raw k‐space data were exported and reconstructed offline. Individual complex images for each coil element were reconstructed by simple Fourier transform and combined by taking the sum‐of‐squares. SNR maps were created by dividing the sum‐of‐squares image by the average of the noise signal in a background region of interest. This process was repeated for both coils. A correction factor may be applied to the mean of the absolute noise signal to convert this to the standard deviation of the unrectified noise, but this was not applied, because it was common to both measurements being compared.

SNR Evaluation: Infant Brain

In vivo evaluation of SNR was obtained by performing a dMRI acquisition (TR = 7856 ms, TE = 49 ms, voxel size = 2 × 2 × 2 mm, FOV = 224 × 224 × 98 mm, number of directions = 32, maximum b factor = 750 s/mm²) on two infants. The first infant was imaged at 44 + 3 wk PCA, age at birth 40 + 1 wk using the 32‐channel adult head coil, and the second infant was imaged at 43 + 2 wk PCA, age at birth 41 + 2 wk using the 32‐channel neonatal coil. The SNR in both data sets were ascertained using Image J software v1.49 (National Institutes of Health, Bethesda, Maryland, USA). A mask of the B0 image was created and the mean and standard deviation of the signal across all volumes in both dMRI data sets was estimated.

Acoustic Noise Attenuation

Infants scanned using the adult coil and the NBIS both had the same ear protection described previously. However, for NBIS we employed an acoustic hood. We were not able to reliably compare measurements of the noise attenuation for the NBIS and the adult coil because precise results depend heavily on consistency of the microphone position. We therefore measured the sound attenuation with and without the hood as a function of frequency. The test was performed using the DHCP multiband (MB) functional MRI (fMRI) sequence, because it has more consistent nonvarying noise levels. A Castle GA221 MR‐compatible microphone (Castle Group Ltd, Scarborough, UK) was placed pointing in an RL orientation at the isocenter of the magnet and at the center of the hood. The level of noise (dBA) generated at each octave band was measured for the open bore and with the hood in situ.

Adult Coil

From December 18, 2013, until November 17, 2014, 243 infants underwent an MRI of the brain using a Philips 32‐channel adult head coil. Only term‐born infants who underwent imaging without sedation were included in the analysis (n = 68). Infants were positioned in the adult head coil and were monitored and immobilized as described previously. Once in position, and once vital sign monitoring was stable, the MRI scan protocol was started (Table 2). The number of infants with a completed MRI protocol and images with no overt motion artifacts was ascertained.

NBIS with Dedicated Neonatal Coil

Between February 6, 2015, and November 17, 2015, 171 infants underwent brain MRI using the NBIS. Only term‐born infants imaged without sedation were included in the analysis (n = 132). The infants were prepared as described in the “Preparation of the Shell and Infant Positioning” section in this article. The number of infants with a completed full MRI protocol and images with no overt motion artifacts was ascertained.

Acoustic Noise Attenuation Evaluation

The acoustic hood produced an overall reduction in noise level of 10.6 dBA at a midrange frequency of 1 kHz, with a maximum attenuation of 17.5 dBA at the higher frequency of 4 kHz. A graph showing the noise level at different octaves with and without the hood and the total attenuation provided by the hood can be seen in Supporting Figures S1a and S1b, respectively.