Interactive neonatal gastrointestinal magnetic resonance imaging using fruit juice as an oral contrast media

12 different contrast media were prepared, as follows: Tap water, standard infant formula milk (SMA Gold, SMA nutrition, Berkshire, UK), Iopamidol (Gastromiro, Bracco, High Wycombe, UK), and the following 9 commercially available (standard supermarket stock) fruit juices: orange juice, two examples of pineapple juice (one from juice; PJ, and one reconstituted from concentrate, PJ2), apple & beetroot juice, prune juice, blackcurrant juice, blueberry juice, raspberry juice, and blackberry juice. Three of these products (PJ2, prune, and blueberry juice) required further dilution to achieve a carbohydrate concentration of <10%, all other media were imaged undiluted (see Table 1). Estimated nutritional and paramagnetic agent contents of each juice are given (Table 2).Each proposed contrast agent was placed within a multi-compartment phantom positioned inside an 8 channel brain array coil, and images acquired on a standard 1.5 T whole-body MR system (Signa HDx; GE Healthcare, Waukesha, WI; Figure 1).

The T₁ values were measured using a 2D inversion recovery (IR) prepared fast radiofrequency (RF) spoiled gradient echo (FSPGR) sequence with variable inversion times (TI) at nine different TIs (40, 80, 160, 320, 640 ms, and 1.28, 2.56, 5.12, 10.24 sec). Other imaging parameters were: field of view (FOV) 25×25 cm, slice thickness 5 mm, matrix 256 × 256, TR 7.2, TE 3.3 ms.

The T₂ values were measured using an 8-echo spin echo sequence. Eight different echo times (TE) were used, 30, 60, 90, 120, 150, 180, 210, 240 ms. Other imaging parameters were: FOV 25×25 cm, slice thickness 5 mm, matrix 256 × 256, TR 2000 ms.

We then measured the signal intensity of each contrast medium using the IR-SSFSE sequences at variable TI times (0–4000 ms), to simulate the in-vivo environment, and subtracted the signal intensity of milk to predict the possible optimal TI times in vivo.

Interactive MRI in-vivo was performed using an IR-SSFSE pulse sequence. This modified pulse sequence has been custom developed to operate within a proprietary interactive imaging interface (i/Drive Pro Plus; GE Healthcare) which allows interactive control over many imaging parameters such as image plane location, phase encode ordering, and the use of a 180° inversion pulse that can be interactively toggled on and off prior to the SSFSE readout. The inversion time (TI) between the 180° inversion pulse and the SSFSE readout could also be interactively controlled during imaging, by inputting different IR times in ms. Initial imaging parameters for the SSFSE sequence were: FOV 25×25 cm, matrix 256×256, TR 2500 ms, TE 50.1 ms, single slice thickness of 10 mm. No motion suppression or breathing suppression technique was used.In volunteer imaging, initial localiser images and non-interactive axial and coronal FIESTA and SSFSE sequences were obtained in a supine position to outline upper gastrointestinal anatomy and localise pre-existing small bowel contents. Coronal interactive IR-SSFSE was then used during free breathing to vary the TI at 100 ms intervals (between 0 and 2000 ms), and signal intensities were measured in 3 regions of interest (ROIs): stomach, proximal small bowel (duodenum) and mid-small bowel (jejunum; Figure 2), in order to characterise the signal properties of pre-existing bowel contents.

We then repeated the study with increasing TIs within 10 minutes of the oral administration of 200 ml pineapple juice to three adult volunteers, and after 20 ml pineapple juice administration to a single neonatal infant (2 week old 3 kg male infant). We did not use any form of respiratory or cardiac gating.

Figure 1 demonstrates the phantom used for in-vitro testing. The T₁ and T₂ relaxation times for the contrast media tested are reported in Table 1: pineapple juice from juice (PJ) was found to have both the shortest T₁ (243 ms) and T₂ (48 ms) relaxation times.Assuming that milk was likely to be the main constituent of pre-existing bowel contents in pre-weaned infants, a representative graph of the relative difference in signal intensity between PJ and formula milk is shown in Figure 3, demonstrating that maximal differences are seen in-vitro with TIs between 500 – 900 ms.

We subsequently tested pineapple juice as an in vivo contrast agent using the IR-SSFSE sequence: we measured signal intensity in three different ROIs (Figures 2), with increasing T1 at 100 ms intervals (0 – 2000 ms; Figure 4) before and after oral PJ administration. By subtracting the signal intensity of distal (jejunal) small bowel contents, we plotted differences in signal intensity (Figure 5) and showed that maximum differences between stomach and jejunum contents were observed at TIs of 1000 ms, 1100 ms, and 900 ms in the three adult volunteers respectively (Figure 6). We were unable to obtain good images of every TI in the neonate following 20 ml PJ administration due to patient motion, but maximal signal differences were seen at 1000 ms (Figure 7).

One limitation of this feasibility study could be the variable content of commercially available fruit juice. We did not determine the exact constituents of the different fruit juices used in this study, but relied on the manufacturers details provided on the packaging and estimated provided by the USDA National Nutrient Database (http://​ndb.​nal.​usda.​gov). Since the concentrations of fruit pulp and therefore paramagnetic properties of diluted commercial fruit juices may vary, the ability to interactively optimise the IR timing becomes even more important as a way to compensate for these variations.

We acknowledge that we did not include any patients in this volunteer feasibility study, and have tested only a limited number of volunteers. The full diagnostic potential of this technique now needs to be evaluated in a larger patient population.