Contrast-enhanced 3D spoiled gradient-recalled echo (SPGR) sequences with parallel imaging are widely used for investigation of liver diseases [1, 2]. Using modern parallel imaging techniques, these sequences can acquire a high-resolution T1-weighted volumetric data set within 20–35 seconds, during which most patients can suspend respiration and eliminate respiratory motion artifact. However, patients with diminished (or no) breath-hold capacity present a challenge to this approach because either coverage or resolution must be sacrificed to shorten the acquisition time to the patient's breath-holding limit. Respiratory navigation of 3D SPGR sequences may provide a method to image these patients with maintained resolution and anatomic coverage. We present our initial clinical experience with a navigated version of 3D SPGR sequence with two-point Dixon water–fat separation for liver imaging.

The study was performed in accordance with our institutional review board procedures. All subjects signed informed consent before imaging. A 2D accelerated dual-echo 3D SPGR pulse sequence was modified to acquire periodic navigator data every 200 milliseconds using a low flip angle cylindrical excitation pulse.

Twenty-two consecutive patients (11 men, 11 women; mean age, 57 ± 15 years) were imaged for a variety of clinical indications (Table 1) on a single 1.5-T scanner (Signa HDx, GE Healthcare) between September 26, 2008, and February 4, 2009, using a 12-channel torso phased-array coil. The subjects were all patients who had been referred for upper abdominal imaging. None of the patients were known to have diminished breath-holding ability, and there were no exclusion criteria. After our standard abdominal imaging protocol (which included dynamic contrast-enhanced imaging) was completed, three sequences were performed: a standard breath-hold 3D SPGR, repeat of the same sequence with the patient breathing freely, and respiratory navigated 3D SPGR. These were performed approximately 5–8 minutes after injection of gadolinium contrast material. Both navigated and conventional sequences were performed with the following parameters: TR/TE, 6.9/2.4; flip angle, 12°; field of view, 34; 320 × 192 matrix; and pixel bandwidth, 391 Hz. Both sequences used 2D ARC (GE Healthcare)-accelerated parallel imaging (R = 2 × 2) [3] and two-point Dixon reconstruction with a phase-correction algorithm [4] to decompose water-only and fat-only images. A typical scan acquisition time for the navigated series was 1 minute. The navigator tracking pulse was placed on the highest point of the right diaphragm, with the trigger set to end-expiration and an acceptance window of ± 2 mm.

Subsequently, the images were graded independently by three radiologists according to criteria presented in Table 2. All three radiologists were certified by the American Board of Radiology, fellowship trained, and experienced (28, 11, and 2 years) independent interpreters of body MRI. The radiologists were blinded with regard to the image set being graded, although not to the clinical indication. After grading each series for overall image quality (criteria listed in Table 2), the radiologists then directly compared the navigated sequence with the breath-held and free breathing conventional series and ranked them on a 5-point scale (Table 3). The conventional breath-held and free breathing series were not compared with each other. The presence absence of saturation effects from the navigator pulse was recorded.

Agreement among readers was calculated unweighted kappa. Differences in quality ratings between scanning methods were tested by paired Wilcoxon's tests for each reader. Tests for pairwise superiority of breath navigation over each of the other two methods was done for each reader with a one-sided paired Wilcoxon's test using a null hypothesis of a median rating of zero (comparable quality). Fisher's exact test was performed for evaluation of differences between the sequences. All statistical analyses were done with Stata Release 9.2 (StataCorp).

When comparing the image quality ratings for each reader, there was little to no agreement among readers beyond that expected by chance (free breathing, κ = 0.02 and p = 0.424; navigation, κ = 0.17 and p = 0.076; breath-hold, κ = 0.02 and p = 0.441). In part, this may be because of the limited rating scale. However, ratings for the respiratory navigation (1.33 ± 0.51) method were intermediate between free breathing (0.47 ± 0.61) and breath-holding (1.83 ± 0.38) (Fig. 1). In no case did the navigator fail or did the navigated sequence fail to obtain images. Thirty-nine of 66 image quality ratings for the free breathing sequence were nondiagnostic, but only one of 66 navigated sequence ratings was nondiagnostic. None of the breath-hold rankings were nondiagnostic.

Similarly, in the side-by-side comparison of diagnostic image quality, there was again little agreement among readers beyond that expected by chance (navigation vs free breathing, κ = 0.15 and p = 0.040; navigation vs breath-holding, κ = 0.19 and p = 0.010). However, on the basis of pairwise comparison, navigation was significantly better than free breathing and significantly worse than breath-holding (p < 0.0001 in each case) (Fig. 2). Navigation had a greater proportion of “much better” ratings over free breathing than breath-hold did over navigated (47% vs 18%, p = 0.001 by Fisher's exact test).

In no case were saturation effects noted in the liver, diaphragm, or remainder of the image. The scan acquisition time ranged between 65 and 85 seconds, with a mean duration of 74 ± 6 seconds.

Although respiratory navigation has previously been successfully used in cardiovascular delayed enhancement imaging [5–8], to our knowledge, this is the first report of the integration of this technique with a 3D SPGR sequence typically used for abdominal imaging applications.

Although performance of the navigated sequence was inferior to breath-hold examinations in our study, there was improvement when compared with free breathing series (Figs. 3A, 3B, 3C, 4A, 4B, 4C, 5A, 5B, and 5C). Although the image quality comparisons indicated that navigated images did not reach the quality of the breath-hold images, the average navigated rating (1.33) is much closer to the average breath-hold rating (1.83) than the average free breathing rating (0.47). The fact that 47% of the navigated images were graded “much better” than free breathing images, whereas only 18% of the breath-hold images were graded “much better” than the navigated images also indicates that navigated images were closer in image quality to the breath-hold than the free breathing conventional images. No saturation effects from the navigator pulse were noted on the diaphragm or liver; this is likely because of the low flip angle used.

The residual motion artifact on the respiratory navigated sequences probably occurs because the navigated acquisition allows a small range of free breathing motion within its data acceptance window. It is also possible that there could be some inaccuracies in sensing the location of the diaphragm. However, because the only current clinical alternative for patients with limited or no breath-holding capacity is to decrease the inherent resolution of the images, even a modest improvement is useful if the residual ghosting artifacts can be read through and the inherent resolution of the images is preserved. One approach under investigation to mitigate these residual motion artifacts is the addition of slab tracking within the acceptance window, in which the excited volume is adjusted in real time according to the amount of detected motion. Alternatively, the acceptance window could be narrowed to reduce motion artifacts, with a corresponding increase in acquisition time.

A significant drawback to the technique is the additional scanning time required for respiratory navigation, which currently prohibits acquisition of discrete hepatic arterial and portal venous phase images. Although our study was performed beginning approximately 5 minutes after injection of IV contrast material, early dynamic perfusion imaging is standard for many clinical situations (especially liver imaging) and has been widely adopted in surveillance for hepatocellular carcinoma in cirrhotic patients [1, 9]. The longer scanning time with this technique (just over 1 minute) does not allow the necessary temporal resolution to resolve these phases of intravascular and parenchymal enhancement.

Techniques are under investigation to improve acquisition efficiency including modified view ordering schemes and increased acceleration factors. However, a clinical approach to such patients with current technology might involve initial sequential acquisition of low-spatial-resolution and higher-temporal resolution images within the patient's breath-hold time, followed by higher-spatial-resolution imaging with respiratory navigation. Further work is needed to optimize this technique, including investigation of patients with truly diminished breath-holding ability.

A. C. Brau and Y. Iwadate are employed by GE Healthcare. R. J. Herfkens holds a research grant from GE Healthcare.

Address correspondence to P. M. Young ([email protected]).