The current standard of care for evaluation of vesicoureteral reflux (VUR) is x-ray voiding cystourethrography (VCUG). The pediatric VUR guidelines panel for the American Urological Association urged the development of “less traumatic methods to determine whether reflux is present...as well as techniques that result in less radiation exposure” [1]. Alternatives to x-ray VCUG include nuclear scintigraphy [2], which offers lower radiation dose; enhanced sonography, which is free of ionizing radiation [3]; and Doppler sonography, which avoids radiation and catheterization [4]. Although x-ray fluoroscopy provides images with excellent spatial resolution, MRI affords better soft-tissue contrast and multiplanar information. Previous attempts at detecting reflux of gadolinium with MRI were performed without real-time continuous interactive imaging for the duration of the examination and thus yielded lower sensitivity for VUR with MRI than x-ray VCUG [5, 6]. Therefore, we aimed to develop an MRI voiding cystography protocol with continuous real-time true MR fluoroscopy and to validate this protocol in a pilot study against the reference standard x-ray voiding cystourethrography.

Imaging was performed on an open-bore 0.5-T magnet (Signa SP, GE Healthcare) that allows continuous access to the child and real-time MR fluoroscopy (simulates x-ray fluoroscopy). In addition, the system incorporated a hybrid x-ray system developed and constructed at our institution that is compatible with MRI (Fig. 1A). This system has a fixed anode tube with a maximum kVp of 110, automatic exposure control, a continuous fluoroscopy mode with 30 frames per second, a pulsed fluoroscopy mode, and a single-shot imaging mode [7, 8]. An amorphous silicon flat-panel x-ray detector was placed under the child (Fig. 1A). Permanent magnets were mounted in the x-ray detector to reduce field homogeneity from magnetic components in the detector. Additionally, x-ray transparent MR receiver coils constructed at our institution were used [9], which were made from aluminum (Fig. 1B).

Eight patients (five boys and three girls) with known VUR were enrolled in the study after approval of our institutional review board and informed parental consent. Ages ranged from newborn to 11 years, with a mean of 2.9 years. Patients were recruited from a cohort of patients returning for annual follow-up evaluation in the urology clinic at our institution who required an x-ray VCUG as part of their routine care. Any patients with a documented minimum grade 2 by the International Classification System (ICS) for VUR were eligible [10]. Exclusion criteria were contraindications to MRI (e.g., pacemaker, ferromagnetic clips, etc.) or a documented gadolinium allergy. No sedation or anesthesia was used.

For each patient, the study protocol consisted of placement of a urinary catheter by the patient's urologist, who positioned the patient in the scanner and remained by the side of the patient for the duration of the examination, monitored for adverse events, and received the results of both the MRI and the x-ray studies. A diaper was worn by each patient to protect the magnet from contamination. Two radiologists (with 4 and 30 years of experience) controlled the MRI scanner and x-ray fluoroscopy unit and interpreted each examination. As part of the research study, the MRI was performed by first obtaining single-shot fast spin-echo coronal images of the torso. Then, dilute gadolinium (1:100 in normal saline) was infused by gravity into the urinary bladder under MR fluoroscopic observation, with a spoiled gradient-echo sequence permitting real-time control of flip angle, field of view, scan plane orientation, slice thickness, saturation pulses, and matrix, as directed by the radiologists. Gadolinium was infused to give bright signal on the gradient-echo sequence because saline would have a signal similar to urine. The field-of-view was chosen to include the kidneys, ureters, and bladder. The image frame rate was approximately one image per second using a TR of 15 milliseconds with spatial resolution approximating 1 × 2 mm. Slice thickness varied from 1 to 2 cm. MR fluoroscopic examination continued until voiding nearly emptied the bladder.

An ICS grade for VUR was then assigned to each renal collecting system by consensus between two radiologists. A fractional grade was assigned if a clear ICS grade could not be determined (e.g., 3.5 if between a 3 and 4). Immediately after MR cystography, a radiographic voiding cystography was performed with 60 kVp and approximately 2 mAs. Through the same urinary bladder catheter used for the MRI, the bladder was gravity filled with iodinated contrast material under intermittent fluoroscopic observation until the patient voided again. As before, an ICS grade for VUR was assigned by consensus. The results of both the MRI and the x-ray fluoroscopic study were relayed to the referring pediatric urologist. Agreement of VUR grading between x-ray and MRI for each collecting system (eight patients, 16 collecting systems) was then determined as strict (grading difference of < 1) or loose (grading difference of < 2). The frequency of strict and loose agreement was then determined, along with a 95% CI for the frequencies of agreement.

MR image quality was assessed retrospectively by determining contrast-to-noise ratios (CNRs) for the urethra and urinary bladder for each patient. CNRs were also measured in structures identified by reflux of gadolinium. To determine CNRs, the image that best delineated the structure of interest was selected. Then three regions of interest (ROIs) were placed. The first was the largest possible circular region over the structure of interest. The next was the largest possible circular region over a homogeneous region adjacent to the structure of interest that visually had signal intensity closest to that of the structure of interest. Finally, the third ROI was the largest circular background region in the image. The CNR was taken as the ratio of the difference of mean signal intensities between the first two ROIs to the SD of the third ROI.

Additionally, image quality was assessed retrospectively by grading the quality of the delineation of structures of the urinary tract. For each structure that was outlined by gadolinium on MRI and iodinated contrast medium on x-ray fluoroscopy, the quality of delineation of the structure was graded on a 3-point scale by consensus between two pediatric radiologists: 1, detectable but margins not clear; 2, margins mostly delineated; and 3, all margins sharply delineated. Finally, for each examination, overall image quality was judged retrospectively on a 3-point scale: 1, poor, nondiagnostic; 2, good, diagnostic but with some uncertainty; and 3, excellent, no doubt about diagnosis.

Both MRI and x-ray VCUG studies were performed in each patient, with no adverse event. Grading of reflux for each renal collecting system is delineated in Table 1, which shows the average of the grading as determined by two radiologists. The grade of reflux was different by less than one grade in all but two renal collecting systems (Figs. 2A, 2B, and 2C), resulting in a sensitivity for MRI of 88% (95% CI, 72–100%). If the definition of agreement between MRI and x-ray is loosened to a difference of less than two ICS grades, the sensitivity of MRI is 94% (95% CI, 82–100%). In one collecting system, reflux was detected at MRI but not at x-ray VCUG (Figs. 3A, 3B, and 3C). In another collecting system, reflux was seen in the renal pelvis at x-ray VCUG but only in the ureter at MRI.

CNRs for various structures are shown in Table 2. The mean CNR was highest for the urinary bladder. Mean CNR exceeded 25 for the urethra, ureters, and renal collecting system. A comparison of the quality of the delineation of structures identified on both MRI and x-ray fluoroscopy is shown in Table 3.

By this measure of image quality, x-ray fluoroscopy was superior (Wilcoxon's signed rank sum test, p < 0.05) to MRI in delineation of the ureters and renal collecting system. The mean score for evaluation of the delineation of the urinary bladder was 2.25 by MRI versus 3 by x-ray fluoroscopy (p < 0.05). The urethra was seen at MRI in six of our eight cases (Figs. 4A, 4B, and 4C). On retrospective review of captures from x-ray fluoroscopy, an image of the urethra was present in four cases; of these four cases, the mean score for x-ray was 2.75 versus 1.75 for MRI (statistically insignificant, with p = 0.25). Overall image quality was graded as 3 for all x-ray fluoroscopy studies, as 3 on four of the MRI studies, and as 2 on the remaining four MRI studies, for a mean score of 3 on x-ray versus 2.5 for MRI (p = 0.06).

At present, urinary tract imaging consisting of x-ray VCUG and renal sonography is recommended for infants and young children up to approximately 6 years with first-time urinary tract infection (UTI) [11]. The objective of the VCUG is to show VUR, which predisposes to permanent renal damage in the setting of UTI [12]. Failure to accurately diagnose VUR may occasionally result in end-stage renal disease, requiring dialysis or renal transplantation [13]. VUR also has been shown recently to be more common in asymptomatic siblings of children who have reflux and in children with parents who have had reflux [14, 15]. In addition, 2–4% of obstetric sonograms detect fetal urinary tract dilation that may be secondary to VUR [16]. These patient groups frequently undergo VCUG.

X-ray VCUG is performed primarily in children, who are at the greatest risk for the harmful effects of ionizing radiation. Although accurate for characterizing and grading reflux, the examination exposes the child to radiation, particularly in the region of the kidneys and reproductive organs. A standard examination involves a scout radiograph and subsequent fluoroscopy [17]. Intermittent fluoroscopic monitoring during the examination results in additional radiation exposure, varying with patient size, patient ability to cooperate, examination difficulty, and fluoroscopist skill [18]. Patients with documented VUR will often undergo yearly x-ray VCUG or radionuclide cystography as they await spontaneous resolution or a decision for surgical correction.

Given the need for multiple examinations with ionizing radiation, several investigators have explored MRI as an alternative. One of these efforts relied on the assumption that dilation of a ureter or renal pelvis occurs on micturition [19], finding high sensitivity and specificity. Another approach detected reflux on the basis of temporal signal intensity curves after the administration of IV gadolinium, with a spike in the signal intensity in the collecting system an indication of reflux [20]. Each of these two techniques used sedation. In an approach similar to our study, Lee et al. [5] instilled gadolinium in the urinary bladder and sought evidence of VUR by obtaining coronal T1-weighted images immediately after voiding. The study of Lee et al. was performed under sedation and did not permit real-time monitoring for reflux or interactive control of scanning parameters. Along the same lines, Teh et al. [6] reported a case of reflux in an adult by filling the bladder with gadolinium and monitoring for reflux with MR fluoroscopy; when reflux was detected, a 3D T1-weighted acquisition was triggered.

In a pilot study on an open, interventional magnet, we have shown the feasibility of MR voiding cystography with a fluoroscopic real-time interactive spoiled gradient-echo sequence for the evaluation of VUR and determined an estimate of the sensitivity of the method. Although image quality was better with x-ray fluoroscopy than with MRI, MRI did show a case of reflux not seen by x-ray fluoroscopy. The technique has the advantage of lacking ionizing radiation. Additionally, when combined with T2-weighted and IV contrast-enhanced imaging, which were not performed in this study, MRI may provide a comprehensive evaluation, assessing reflux and renal scarring.

Furthermore, volumetric T2-weighted and contrast-enhanced T1-weighted acquisitions would provide improved resolution, which would make assessment of smaller features, such as calyceal blunting, easier to assess. As a cross-sectional imaging technique, the ureterovesicular junction is well delineated, although the inferior spatial resolution may compromise evaluation of the urethra. An additional issue with cross-sectional imaging is that brief reflux into a collecting system that is not in the plane of imaging during the reflux may not be detected. This, along with intermittent reflux, is a possible explanation of the case of grade 2 reflux at x-ray fluoroscopy that corresponded to only grade 1 reflux at MRI. In our experience, the duration of the study is simply determined by the time required to fill a urinary bladder by gravity until the patient voids. However, this study was performed on an open MRI system, permitting the physician to position the child over the receiver coils without the use of a restraint. We are currently gaining preliminary experience performing examinations on the more commonly available closed-bore systems.

MRI offers the advantage of continuous surveillance for VUR during the entire filling, waiting, and voiding phases of the study, theoretically offering higher sensitivity. Given ionizing radiation concerns and the as-low-as-reasonably-achievable principle, this is not feasible with x-ray VCUG. This, or intermittent reflux, may explain the single patient in our study in whom reflux was identified using MR cystography but not x-ray cystography. Cystosonography has been reported in Europe for diagnosis of VUR as an alternative without radiation [3, 4]. Cystosonography has not been widely accepted in the United States, possibly because the method has a prolonged learning curve and is challenging for the radiologist to interpret or that the small-field-of-view images are difficult for the urologist to evaluate.

Our study has several limitations. The small sample size yielded a relatively large 95% CI for the sensitivity of MRI for reflux. Additionally, examinations were performed on a 0.5-T open magnet. Translating this technique to the ubiquitous closed-bore systems offers the advantage of improved image quality with higher temporal and spatial resolution. However, controlling the position of the patient may then require a papoose for younger children. Another study limitation is the lack of blinding of radiologists to results of the MRI portion of the study when performing and interpreting the x-ray portion. Blinding was not feasible because the two examinations were performed in rapid succession. Thus, intermittent VUR, both in presence and degree, limits true comparison of x-ray and MRI techniques. Furthermore, the x-ray portion of our study was performed with a system designed and constructed in house as opposed to commercially available equipment, which may have better performance. Finally, comparison of MR voiding cystography is made to x-ray VCUG, which does not have perfect sensitivity and specificity.

In this pilot study, we have shown the feasibility of MR voiding cystography with a real-time interactive MR fluoroscopic technique on an open MRI magnet for the evaluation of VUR in children. Further work is required to translate this technique to the more ubiquitous closed-bore magnets and better define the accuracy of the technique.

Supported by the Research and Education Foundation of the Society for Pediatric Radiology, the Packard Foundation, and Michelle Ferrari.

Address correspondence to S. S. Vasanawala ([email protected]).

This is a Web exclusive article.