2In this study, we present the design, construction, and characterization of five size-matched 32-channel receive-only array coils for highly accelerated pediatric brain imaging at 3 T. The characterization of coil performance included the evaluation of noise correlation, SNR, and G-factor maps using size-matched phantoms. Furthermore, the coil performance was evaluated with in vivo scanning on adults with small enough heads to fit the 4-year-old and 7-year-old pediatric coils and in a child evaluated in a separate clinical study. Our evaluation shows that a tailored array approach for size-matched pediatric brain MRI provides significant SNR gains for both accelerated and unaccelerated imaging. Furthermore, it underscores the inefficiency of using adult or single channel volume coils for pediatric brain examination. For the optimal use of array detection, regarding SNR and encoding performances, the surface coils should closely surround the imaging volume of interest. Furthermore, handling and patient comfort are important considerations when targeting pediatric populations. Although the approach required building multiple arrays to size-match each age-group, it is important to understand the potential gains of this approach to motivate investment in a flexible, or adjustable size approach.
A number of technical issues arise in the implementation of a large channel-count array with relatively small element size. In particular, the interelement decoupling, QU
ratio, SNR performance, and location of preamplifiers become more challenging, whereas commercial 32-channel head coils for adults with a loop diameter of ~90 mm are typically constructed out of flexible circuit material, array coils with smaller elements size show eddy current losses in the conductors of the neighboring elements can be significant and lead to a lower QU
ratio and SNR using this circuit approach (6
). Spatially sparse conductors, such as wire, as well as relocating the preamplifier and its motherboard 2–3 cm from the loop elements, reduces the losses in the copper for the dense arrays. We also found that the ability to mechanically optimize the overlap between two loops by bending the wire facilitated the element decoupling procedure. However, we still could measure a reduced unloaded Q
, when the loop under test was placed in an array configuration suggesting that losses within the conductors of neighboring elements were still present. We found that the negative impact of these losses increases as the loop diameter is reduced. While loop diameters of 90 mm show only an increased resistance of 3.2% due to the neighboring elements, the increase was 8.2% for the 60 mm diameter elements used for the neonate array coil. However, the QU
-ratio for all constructed array coils show sample noise dominance, but this metric was also less favorable for smaller loop sizes (e.g., 6-month-old or neonate arrays.) Thus, maintaining body noise dominance in the small loop elements is more challenging for two reasons; the intrinsically smaller QU
-ratio in the isolated loop, and the increased effect of losses within the surrounding elements.
When loop elements become smaller, and therefore, the array is more densely packed, the positioning of the preamplifier becomes more challenging. For both, SNR gain with optimum noise figure and preamplifier decoupling, it is important to align the preamplifier to z
). This reduced degree of freedom for preamplifier positioning is conflicting likely with optimum cable routing. Mainly, the routing of the output coaxial cable should not pass near (<~2cm) of another preamplifier’s input to avoid positive feedback loops. This was especially challenging for the neonate and the 6-month-old array coils.
Measurements of the transmitted RF field with and without the pediatric 32-channel coils inside the scanner ranges from 1.0 to 3.2% in transmit power difference. This indicates sufficient coil element detuning and suppression of common mode currents on the cables. The modest increase in power requirement, when arrays are present, might be due to power dissipation in the copper-wire used in the element construction, cables, and element-adjacent preamplifier, including circuitry and motherboard. Similar observation can be found in literature about large-count array coils for adults (6
The noise correlation between the coil elements over all 32-channel arrays revealed a mean coupling value of 12%. However, some pairs in each coils showed correlations over 35%. All of these high coupling pairs were identified to be adjacent to each other but not overlapped. These were the coil pairs, where one loop was located in the posterior coil segment and the other in the frontal paddle. Examination of different paddle positions on the phantom showed significant changes in the level of correlations of those pairs. However, the noise-covariance weighted root sum-of-squares reconstruction method is able to reduce the impact of such element coupling within the array by utilizing coil sensitivity and noise correlation information. This image reconstruction method offers the array coil designer increased freedom in array layout, beyond overlapping elements. But the presence of this “gap” between frontal paddle elements and the posterior elements led to the gap seen in the SNR maps, as seen in the sagittal maps.
All constructed arrays showed increased SNR compared with the adult 32-channel coil. Because of the spatial SNR variation associated with array coils, the SNR gain of the constructed pediatric array coils was higher at the periphery of the brain but also showed moderate improvements at the center. Furthermore, the SNR comparison in the “whole phantom” ROI showed less favorable SNR improvements compared to the “whole brain” ROI. The whole phantom ROI included areas such as the face, for which the pediatric brain array coils are not intended to image, whereas the comparison coils (such as the adult 32-channel and the pediatric birdcage) are more effective at receiving signal in this area.
Comparisons between G-factor maps obtained from commercially available 32-channel adult head coil and the five constructed arrays, showed that the multisize approach by using incrementally sized 32-channel coils for pediatric imaging produce overall significantly more favorable G-factors compared to a “one-size fits all” approach, required when adult coils are used. The spatial variations of the G-factors still show the highest noise amplification in the central brain regions when the pediatric coil arrays are used. However, in relative comparison with the 32-channel adult coil, the central brain regions show the highest improvement in G-factor. The G-factor improvement is likely attributed to the close fitting and smaller elements, which offers a stronger spatial modulation of signal intensity and thus improved ability to unalias folded images (SENSE method) or synthesize spatial harmonics (SMASH or GRAPPA methods). Thus, the accelerated images obtained from the pediatric arrays provide the ability to accelerate at approximately one unit higher at a given noise amplification compared with the adult array.
In addition to hardware optimization in pediatric MRI, image reconstruction algorithms should be adapted to the specific requirements of the pediatric imaging and the small loop diameters of the pediatric arrays. For example, comparisons with the 32-channel adult head coil images show that the pediatric array images are darker at the center of brain. In contrast, the SNR maps show that these regions are not reduced in SNR. Instead, the signal intensity normalization algorithm performs suboptimally on the small-loop pediatric array coils although our postprocessing software can handle these intensity variations. Another example image reconstruction improvements in pediatric MRI is shown in the recent study from Vasanawala and coworkers (18
), demonstrating a combination of parallel imaging and compressed sensing (17
) in pediatric MRI with promising results. This study translated successfully the L1
-SPIR-iT method (18
) into a feasible clinical pediatric MRI protocol and obtained high image quality with a reduced acquisition time. In principle, the compressed sensing and parallel imaging approach is synergistic with the SNR and acceleration improvements of the dedicated pediatric array coils.
The use of MRI in pediatric imaging remains a challenging undertaking due to practical, methodological, and analytical issues that arise when imaging young populations. Beyond clinical MRI applications, there is a progressive use of functional MRI (fMRI) in pediatric research studies (19
). Functional MRI allows monitoring of functional developmental processes during brain maturation and may provide the basis for early detection of pathophysiologic processes; a prerequisite for functionally guided therapeutic interventions. Achieving this goal requires high quality structural and functional MRI data, often in studies without anesthesia. Head motion has been identified as the most common reason for scan failure in pediatric fMRI (24
). The ability to highly accelerate structural scans is therefore desirable for pediatric brains studies since it reduces scan time (and the probability of motion) significantly. For single shot fMRI acquisitions, the acceleration does not translate directly to improved motion mitigation (but to reduced susceptibility distortions), but due to the tightly fitting array coils the degree of motion is mechanically constrained. The close-fitting array coils limit the ability of the subject to rotate their freely head. Furthermore the narrow fit around the neck prohibits the subject from sliding out of the coil. These design choices raise the concern about the psychologic reaction to having one’s head “captured.” However, our collaborative pediatric imaging studies (as well as our experience with adult subjects in the 4-year-old and 7-year-old arrays) showed that the helmets do not induce additional anxiety. Note that the frontal paddle is held in place with a simple articulating hose tubing, so a typical 4-year-old child would have no problem exiting the coil on its own.
Although the helmets constrain motion, they still allow side-to-side motion. Although centimeters of motion are not possible, the head can move against the compressed padding. In this respect, the situation is not much different from using the vacuum bead immobilization approach (commonly used at children’s hospitals) or wedging foam pads within the adult coil to immobilize the subject. Ultimately, the helmets were designed to be fairly comfortable to lie on with minimal foam padding. This stems from the bowl shaped support for the occipital pole, the curving neck support and the lack of any structures over the eyes, nose, and mouth region. The latter is particularly important to avoid rebreathing exhaled CO2 rich air (CO2 inhalation causes anxiety or panic).
Using tight fitting pediatric array coils clearly reduces the range of choices of noise protection. In fact, common used MRI ear-muffs for noise mitigation are not compatible with the arrays. Often, however, regular ear-muff are not used on younger children (< about 3 years old) anyway, when vacuum bead head-holders are used to immobilize the anesthetized head inside the adult coil. In our approach, we used insertable pediatric ear-plugs together with external “mini-muffs.” These single use devices are made from thin acoustic barrier foam (similar to the insert plugs) and adhere with via an adhesive surface to the pinna. When auditory stimulus is needed, it must be applied either through an ear-bud based MRI compatible system or externally, through a loudspeaker powerful enough to be heard thru the ear protection.
An additional potential application for the increased parallel imaging performance of the pediatric arrays is to use the reduced gradient encoding to lower acoustic noise rather than reducing susceptibility distortion (25
). In this scenario, the echo-planar imaging echo train length is not shortened, but the reduced number of phase encoding steps afforded by accelerated acquisition is used to lower the BW (and gradient amplitude) of each readout line, resulting in an acoustically quieter scan. If the spatial resolution and acquisition duration are unaltered in the single-shot echo-planar imaging sequences the ramp time of the readout gradient, a sampling BW, and the gradient amplitude can both reduced by a factor approximately equal to the acceleration rate lowering the Lorentz forces within the gradient coil. This has potential for a more comfortable environment during a pediatric MRI scan, or it can be used when auditory stimulations are applied for fMRI.
Finally, there is a potential to reduced scan-slot durations by using accelerated scans and reducing the need for anesthesia and sedation. The latter slows down the pediatric MRI patient flow and raises the costs significantly (7
), thereby reducing the population of patients that can benefit from the diagnostic abilities of MRI. Dedicated sized-matched pediatric array coils can potentially relieve these limitations and establish pediatric MR imaging to a wider range of children with no or lighter sedation/anesthesia. Ultimately, this can reduce costs in pediatric radiology divisions and offset the high cost of age-specific array coils.