Optimization of single injection liver arterial phase gadolinium enhanced MRI using bolus track real-time imaging




To measure contrast agent enhancement kinetics in the liver and to further evaluate and develop an optimized gadolinium enhanced MRI using a single injection real-time bolus-tracking method for reproducible imaging of the transient arterial-phase.

Materials and Methods:

A total of 18 subjects with hypervascular liver lesions were imaged with four dimensional (4D) perfusion scans to measure time-to-peak (TTP) delays of arterial (aorta-celiac axis), liver parenchyma, liver lesion, portal, and hepatic veins. Time delays were calculated from the TTP-aorta signal, and then related to the gradient echo (GRE) k-space acquisition design, to determine optimized timing for real-time bolus-track triggering methodology. As another measure of significance, 200 clinical patients were imaged with 3D-GRE using either a fixed time-interval or by individualized arterial bolus real-time triggering. Bolus TTP-aorta was calculated and arterial-phase acquisitions were compared for accuracy and reproducibility using specific vascular enhancement indicators.


The mean bolus transit-time to peak-lesion contrast was 8.1 ± 2.7 seconds following arterial detection, compared to 32.1 ± 5.4 seconds from contrast injection, representing a 62.1% reduction in the time-variability among subjects (N = 18). The real-time bolus-triggered technique more consistently captured the targeted arterial phase (94%), compared to the fixed timing technique (73%), representing an expected improvement of timing accuracy in 28% of patients (P = 0.0001389).


Our results show detailed timing window analysis required for optimized arterial real-time bolus-triggering acquisition of transient arterial phase features of liver lesions, with optimized arterial triggering expected to improve reproducibility in a significant number of patients. J. Magn. Reson. Imaging 2011;33:110–118. © 2010 Wiley-Liss, Inc.

IT HAS BEEN widely accepted that optimal capture of arterial phase gadolinium-chelate enhanced (Gd) liver images are a critical component to diagnostically optimized MRI (1–8). Arterial-phase images may provide unique additional markers for differentiating characteristics of benign, malignant, hypervascular, and hypovascular primary and metastatic liver tumors (3, 4, 8, 9) and have been shown to provide analysis of active hepatitis in the setting of acute and chronic disease (5, 10).

Suggested methods for ensuring an optimal arterial phase liver MRI have varied between fixed and individually-tailored timing (3). The latter method involves either prescanning with a test bolus (3), or tracking bolus arrival in the descending aorta on-the-fly (11). These tailored timing strategies have become increasingly more accepted compared to fixed-timing in order to prevent the influence of variable arterial transit times on optimal tumor conspicuity. It is possible to further simplify the arterial-phase timing calculations and make this imaging step incorporated into the arterial phase acquisition by applying a real-time bolus-tracking methodology similar to the strategy employed for contrast enhanced MR angiography (MRA) (11–14). By identifying a common vascular reference point with minimized temporal proximity to optimal liver tumor arterial enhancement, one can expect to reduce the relative error of missed arterial-phase imaging that may result from variable vascular transit times. Real-time contrast bolus-triggering methodology for liver offers technical simplification and speed as compared to prescan test bolus techniques (3).

The technical requirement for capturing the critical arterial-phase is defined by the dynamic imaging point where there is maximal image contrast between arterial-enhancing liver lesions and liver parenchyma. The nature of dynamic contrast behavior in the upper abdomen is best analyzed through high temporal perfusion imaging, which has been previously studied for various applications (3–5, 8–10, 15, 16). It has been shown that there exists a finite “arterial window” following identification of peak aortic enhancement (3); however, the commencement and duration of this arterial window depends greatly on subject-specific vascular transit times, thereby compromising the use of fixed imaging delays after contrast administration. Though a previous investigation reported that peak tumor enhancement generally occurred at a finite time after peak aortic enhancement (15), we propose that the clinical application of optimal timing delays within the arterial window and how this information may be used to optimize the real-time bolus-tracking technique has not yet been fully measured.

While most MR systems are equipped with bolus-tracking sequences, its use has not been widespread in dynamic liver imaging. Particular to liver imaging, the dynamic bolus-tracking methodology is currently limited by the general uncertainty of the step-wise procedures for the acquisition. Namely, there is insufficient documentation in the existing MRI literature for: 1) defining vascular reference points; 2) optimizing timing delays between the arrival of contrast bolus, providing the breath hold commands, and initiation of sequence acquisition; and 3) adapting imaging sequence programming to bolus-timing specifications.

This study is designed to further develop and evaluate Gd-enhanced MRI of the liver using a dynamic real-time bolus-tracking method to capture the time-sensitive arterial phase. The primary aims are to: 1) quantify liver perfusion kinetics during the arterial-dominant phase; 2) determine an optimized timing delay for use with real-time bolus tracking; and 3) provide a measure of the clinical impact of optimized real-time bolus tracking on accuracy and reproducibility of capturing the transient arterial-phase.


Study Patients

This study was Health Insurance Portability and Accountability Act (HIPAA) compliant and was approved by our institutional review board including use of approved informed consent practice and documentation. Initially, 18 subjects (eight male, average age 49.7 years; range 22 to 67 years) exhibiting hypervascular liver lesions during routine MR abdominal examinations were included in the study of perfusion kinetics. The nature of the liver lesions was determined from known patient history reports and confirmed by each case showing well-accepted characteristic MRI criteria (6). The findings showed 10 subjects with focal nodular hyperplasia, five with hepatocellular carcinoma, two with carcinoid, and one with benign adenoma. For subsequent evaluation of real-time bolus tracking, 100 consecutive patients (ages 21 to 80 years, average 52.2 years; 41 males and 59 females) were imaged, while another 100 control patients (50 before and 50 after, ages 17 to 82 years, average 54.4 years; 38 males and 62 females) were imaged using the fixed timing technique (fixed 20 seconds timing delay following injection, taken from previously published technique for routine arterial phase liver imaging [4,8,17]). All patients were referred for abdominal Gd-enhanced MRI examinations as per routine clinical indications and not specifically referred for this study. The experimental/control group was comprised of patients with a history of liver disease in 48/42 of patients (tumor N = 15/16, chronic hepatitis N = 33/26), or disorders involving kidney (N = 15/16), pancreas (N = 7/13), pelvic neoplasms (N = 9/10), bowel (N = 4/9), adrenal (N = 2/1), vascular (N = 4/6), other (N = 11/3).

MR Perfusion Acquisition

A total of 18 subjects underwent perfusion MRI performed on an 18-channel 1.5T Siemens Avanto system (Erlangen, Germany) equipped with gradients of 40 mT/minute and 200 T/m/second slew rate using anterior and posterior phased array surface coils for signal reception. Following a routine clinical liver examination,, which revealed an arterial-enhancing tumor, a subsequent accelerated 3D gradient echo (3D GRE) perfusion scan was performed at least 10 minutes later in a transverse plane encompassing the tumor, liver, descending aorta and portal vein using 35 slices and 3.5-mm slice thickness (18). If multiple lesions were present, the 3D volume was prescribed to include the larger neoplasms while ensuring coverage of the other tissues of interest. The sequence parameters for the perfusion scan were TR/TE/flip = 2.3 msec/0.7/15, 192 matrix (60% phase resolution) with interpolation to 384, 350 mm field of view (FOV) (75% rectangular FOV), bandwidth (BW) = 1300 Hz, and integrated parallel acquisition techniques (IPAT) factor = 3. This resulted in an average temporal resolution of 2.0 seconds per 3D volumetric image set. Additionally, slab-selective excitation and spatial presaturation were implemented to prevent fold-over in the slice direction. The sequence was performed sequentially with 60 dynamic scans, which resulted in a total perfusion imaging time of approximately 120 seconds. The perfusion sequence was initiated simultaneously with an injection of 0.05 mmol/kg gadobenate dimeglumine (Gd-BOPTA) (MutliHance, Bracco, Italy) at 2 cc/second with the subject instructed to breath-hold to tolerance and then breathe slowly.

MR Dynamic Liver Acquisition

For arterial, venous, and delayed imaging in examinations undergoing either fixed-time or real-time bolus-track timing, Gd contrast (Gd-BOPTA, MutliHance, Bracco, Italy) was administered at 0.05 mmol/kg and 2 cc/second, followed by a 20 cc normal saline flush at 2 cc/second. Imaging was performed on either a 1.5T Siemens Avanto, as previously described, or a Philips 1.5T Intera System (Best, The Netherlands), equipped with a 4-element torso array body coil. The procedure for using bolus tracking for arterial-phase imaging consisted of four sequential elements: 1) real-time bolus monitoring concurrently with contrast administration; 2) stopping bolus-tracking when contrast is visualized in the abdominal aorta at the level of the celiac axis (diaphragm); 3) instituting a timing delay for breath hold commands; and 4) executing of a breath hold 3D GRE acquisition. A timing chart of the events structure is given in Fig. 1.

Figure 1.

A timing diagram outlining the series of events to execute arterial-phase liver imaging with real-time bolus-tracking preparation. The initiation of real-time imaging is followed by contrast administration. The bolus is followed to the abdominal aorta (at the level of the celiac axis), upon which real-time imaging is ceased and breath hold instructions are given within a prescribed time window.

Real-time bolus-track imaging was implemented using a 2D GRE sequence, with acquisitions every 0.5 to 1.0 second and on-the-fly image reconstruction (FOV = 450 mm, matrix = 256, TR/TE/flip = 4.1/1.23 ms/10°, slice thickness = 50 mm, 100 dynamics, scan time = 100 seconds). The sequence was oriented in the coronal plane, along the abdominal aorta, and initiated at the same time as contrast administration. The real-time images were displayed using an inline viewer on the console. A manual pause was implemented after cessation of the bolus-tracking sequence during which breath-holding instructions were given to the patient before initiating 3D GRE. The duration of this pause was 6 to 7 seconds, and was optimized using the results of the perfusion data acquired in this study and the approximate center of k-space within the 3D GRE technique.

Arterial-phase imaging was obtained using a 3D GRE imaging sequence with a segmented k-space acquisition. The estimated time to the center of k-space was <5 seconds, ensuring the initial part of the scan effectively coincides with the collection of the desired image contrast. The 3D GRE imaging parameters were the same for both the fixed and bolus-tracking method. 3D GRE was acquired in 17 seconds during patient breath hold with the following parameters: 360 mm FOV (75-90% phase FOV), 256 matrix (70% phase resolution), TR/TE/flip = 4.1/1.7 msec/10°, BW = 400 Hz/pixel, 94 slices, slice thickness = 3 mm, acceleration factor = 2 (sensitivity encoding [SENSE] or generalized autocalibrating partially parallel acquisitions [GRAPPA]) and spectral adiabatic inversion recovery (SPAIR) fat saturation (19). Following the arterial phase imaging image set, venous and delayed interstitial phase image sets were obtained 60 and 180 seconds after contrast administration, to complete the comprehensive multi-phase clinical examination.

Perfusion Image Evaluation

Perfusion data was exported off-line for analysis (Osiris 4.19, Geneva, Switzerland) where individual dynamic slices containing tissues of interest were isolated. Signal measurements were obtained from aorta (at the level of the celiac axis), enhancing neoplasm, remote liver, portal vein, and hepatic vein using a region-of-interest (ROI) tool and propagated to all dynamic images. The size of the specific tissue in question determined the extent of the ROI, while care was taken to avoid inclusion of surrounding tissues. For small regions, such as hepatic veins, the ROI was limited to a small number of pixels; however, at least three other ROIs were drawn if possible, and the intensity time-courses were averaged together. Liver ROIs were drawn in a region neighboring the arterial-enhancing tumor (if possible) to avoid the possible effects of B1 and B0 inhomogeneity. Compensation for significant motion was achieved by manually moving individual ROIs as needed. In addition to the mean signal in the ROI, the standard deviation (SD) was recorded, which was used as an indicator of motion contamination or inclusion of surrounding tissue signals. ROIs were redrawn or repositioned if SD fluctuated more than 100% of what was expected from a nominally drawn region.

Mean signal intensities for each region were expressed as a function of time, with t = 0 signifying the bolus injection point. Since all intensity measurements were obtained from the same perfusion scan (and often the same slice), rescaling or normalization of the data against background noise was not necessary, given constant receiver gain settings throughout the acquisition. Therefore, the mean signal difference between two tissue regions was considered an appropriate estimation of tissue contrast. This estimation was used to calculate tumor-to-liver tissue contrast at each time point for each subject, thereby providing an additional perfusion curve detailing changes in tumor conspicuity following gadolinium injection.

The specific time points of interest extracted from the perfusion curves included time-to-peak (TTP) signal for each tissue, along with TTP tumor-to-liver contrast (TPC). TTP for each tissue was recorded as the time point from contrast injection to maximum mean signal intensity of the specific tissue in question. TPC was similarly calculated as the time to maximum tumor-to-liver signal difference. Additionally, both of these measurements (TTP and TPC) were expressed relative to peak aorta signal (TTPaorta). This extra measurement was made since TTPaorta is considered the time point real-time bolus tracking would be halted in clinical liver examinations to initiate subsequent arterial-phase 3D imaging. This adjustment places emphasis on the TTP tumor enhancement following bolus arrival in the aorta at the level of the celiac axis, which is central to the bolus-tracking timing method, and needed for technique optimization. These measurements were indicated with the “plus” symbol (+) (i.e., TPC+) to signify temporal measurements relative to peak aorta signal. Since TPC represents a singular point of maximum contrast within a dynamically changing process, a temporal range was also defined for each subject to represent an “imaging window” encompassing heightened lesion contrast during the arterial-phase. The extent of this temporal window was based on lesion-contrast measurements at 60% of maximum, which depended on the signal characteristics of each subject. This value is not an absolute value, but used for the purposes of referencing for this study.

A cumulative average of temporal measurements was conducted for all 18 subjects, along with measurements of SD, and significance between the two largest sub-groups (focal nodular hyperplasia [FNH] and hepatocellular carcinoma [HCC]) were determined via an unpaired Student's t-test, with the significance level P = 0.05. TTPtumor+ and TPC+ were used to define the manual delay between cessation of real-time bolus tracking and arterial-phase 3D acquisition for clinical imaging.

Clinical Image Evaluation

Evaluation of gadolinium liver enhancement was determined in relation to the hepatic vessels containing contrast following previously published guidelines (3–6, 8): Optimal timing, which was our targeted phase of vascular enhancement, was defined as presence of contrast in the hepatic artery and portal vein, but not in the hepatic veins; Late timing was defined as contrast appearing in the hepatic veins; Early timing was defined up to the point when contrast appears in the hepatic artery but not yet in the portal vein. All 200 patient examination images were loaded onto a computer with image review software (eFilm, MERGE eMed, Milwaukee, WI, USA) and reviewed separately by two experienced reviewers having 12 and 5 years of MRI experience. Discrepancies would be documented subsequently for further review. Annotation was suppressed from the images and the patient list was ordered by the hospital record number to create a randomized mixing of control and test patients. All qualitative scores were tabulated categorically, whereupon each category was represented as a percent of the total patient population. This representation produced a measure of timing error, consistency, and accuracy. By determining the ratio of “optimal” arterial-phase images between bolus-tracked and fixed-timed techniques, we evaluated the degree of improvement of one method over the other for reproducible acquisition of targeted vascular enhancement. A two-sample test for equality of proportions was used to test the null hypothesis that both methods have the same diagnostic accuracy.

To estimate bolus transit time variability among individuals, all real-time bolus-tracking images were saved and analyzed to record the transit time from injection site to abdominal aorta using image time-stamps. The durations were averaged among the 100 cases (mean ± SD), and the maximum and minimum times were noted to determine the range of values. To compare the data against the fixed-timed technique, 10 seconds were added to the bolus transit durations to indicate the beginning of 3D GRE. Statistical significance of the difference between bolus and fixed times (against zero) were determined using a Student's t-test with P < 0.05.


MR Perfusion Evaluation

Figure 2 shows a sample perfusion plot from one subject, detailing the temporal relationships between the tissues of interest, and the ideal timing alignment of 3D GRE for reference. A summary of the mean delay times for the 18 subjects are given in Table 1, with subgroup results for FNH (N = 10), and HCC (N = 5). No significant differences were found between groups during the arterial-phase.

Figure 2.

a (graph): Time-signal intensity graph and perfusion images from one subject corresponding to the relative enhancement of a tumor with arterial-dominant enhancement, in comparison to liver, hepatic artery, portal veins, and hepatic veins. The timing relationships for the bolus-tracking technique using centric 3D GRE, as compared to non-centric 3D GRE, is shown along the time-axis below the graph. b (images): The image numbers correspond to the numbered arrows denoted on the graph. Following peak abdominal aortic signal (image 1; dashed arrow), arterial lesion begins to enhance (image 2; solid arrow) until signal intensity (image 3) approximately 11 seconds after bolus detection. The lesion then becomes inconspicuous, as liver signal increases, lesion washout occurs, and hepatic veins fill (image 4). The hatched area between the tumor and liver parenchyma curves denotes the period during which the tumor develops higher signal intensity in contrast to liver. The central region of k-space is depicted for two 3D GRE variants (“k = 0 (effective)”). Note that noncentric 3D GRE may be initiated earlier after bolus detection to ensure capture of the high tumor contrast region.

Table 1. Average Time-to-Peak (TTP) Signal Enhancement (and Range) for Selected Tissues (N = 18)
TTP-TissueSubjectFrom injection (second)From abdominal aorta (second)
  • *

    P < 0.05 HCC vs. FNH.

AortaTotal (N = 18)24.0 ± 4.7 (13.3–30.6)0
 FNH (N = 10)24.1 ± 4.40
 HCC (N = 5)24.0 ± 3.60
LesionTotal (N = 18)33.2 ± 5.9 (20.8–44.8)9.4 ± 2.9 (5.4–14.5)
 FNH (N = 10)33.6 ± 4.99.6 ± 3.2
 HCC (N = 5)32.9 ± 4.99.1 ± 2.0
Lesion-liver contrastTotal (N = 18)32.1 ± 5.4 (22.6–44.8)8.1 ± 2.7 (4.0–14.4)
 FNH (N = 10)32.1 ± 5.68.0 ± 3.4
 HCC (N = 5)32.1 ± 2.98.2 ± 1.4
Portal veinTotal (N = 18)37.1 ± 6.5 (29.7–51.0)13.0 ± 4.9 (5.4–22.1)
 FNH (N = 10)34.3 ± 5.211.0 ± 4.2
 HCC (N = 5)40.3 ± 8.316.8 ± 5.5
Liver parenchymaTotal (N = 18)54.2 ± 10.8 (37.5–75.0)30.3 ± 11.5 (15.0–59.2)
 FNH (N = 10)49.1 ± 9.125.0 ± 7.1
 HCC (N = 5)67.8 ± 8.2*44.3 ± 7.1*
Hepatic veinTotal (N = 18)65.5 ± 11.6 (45.0–90.1)41.1 ± 12.4 (22.5–61.2)
 FNH (N = 10)61.8 ± 12.637.2 ± 12.3
 HCC (N = 5)71.2 ± 11.353.1 ± 11.5

For the perfusion subjects studied, the average TTP abdominal aorta signal from injection was 24.0 ± 4.7s (range, 13.3–30.6 seconds; median, 24.3 seconds). The TTPtumor intensity from injection was 32.4 ± 5.9s (range, 20.8–44.8 seconds; median, 31.9 seconds), which was significantly longer than TTPaorta (P < 0.05). TTPtumor relative to peak bolus arrival in the aorta at the level of the celiac axis, which is the scenario if arterial-phase imaging is triggered using real-time bolus tracking, was 9.4 ± 2.9 seconds (range, 5.4–14.5; median, 8.1 seconds). Comparing TTPtumor and TTPtumor+, the latter strategy results in a 62.1% reduction (24.0–9.1 seconds) in the time-range of peak tumor enhancement when the injection point is used as t = 0. The TPC+ was 8.1 ± 2.7 (range, 4.0–14.4 seconds; median, 8.0 seconds), which was not significantly different than TPtumor+ (P > 0.05). Moreover, absolute lesion contrast (with remote liver) differed by <3% between TPC and TTPtumor measured values.

Figure 3 depicts the phases of bolus arrival in the abdominal aorta and lesion enhancement characteristics for the 18 subjects studied. Despite the identification of absolute peak tumor enhancement, a finite phase of heightened lesion contrast (defined as 60% of maximum) was found following peak bolus signal (TTPaorta), which spanned 12.8 ± 6.9 seconds (N = 18). This “imaging window” commenced 2.9 ± 2.5 seconds and ceased 15.5 ± 6.9 seconds after TTPaorta. These data suggest that arterial-phase imaging should be initiated early after bolus detection near the celiac axis (<20 sec), despite the duration of the “arterial window” (41.1 ± 12.4 seconds; range, 22.5–61.2 seconds; median, 41.9 seconds), which can be defined from peak aorta to peak hepatic vein enhancement.

Figure 3.

The temporal arterial imaging phases are shown for the 18 patients studied. White bars represent the delay from the start of the contrast injection to bolus peak in the abdominal aorta at the celiac axis. TTP and TPC markers indicate the respective additional delay to the optimal imaging time point for arterial phase enhancing tumors. Solid light gray bars represent the perfusion time window defining a tumor-to-liver contrast above a given threshold fraction of peak contrast (60% of maximum). There is a 62% reduction in the range of bolus transit times to peak lesion contrast when abdominal aortic triggering is used in place of the start of the injection. The dark bars represent the additional delay to peak hepatic vein enhancement.

TTP portal vein occurred 13.0 ± 4.9 seconds (range, 5.4–22.1 seconds; median, 12.0 seconds) relative to the bolus reference point (TTPaorta). From the data, tumor enhancement coincided closely with portal venous enhancement, judging by TTP times and perfusion curves (P = 0.08). Since this trend was valid in all subjects, this finding suggests that the portal vein may be used as a visual indicator for optimal tumor enhancement, or to determine whether arterial-phase timing was early or late. This finding quantitatively confirms existing criteria for determining optimal arterial phase imaging (17).

MR Bolus-Tracking Liver Evaluation

Figure 4 summarizes the assessment of the arterial phase liver images acquired using the real-time bolus-tracking technique in comparison to the routine empirically-timed technique. The bolus-triggered technique was optimally timed in a greater number of patients (94%), compared to the fixed timing technique (73%). Examples of optimal, early and late timing are also shown for reference. The improvement achieved using bolus tracking, in terms of the additional number of cases showing accurate capture of the targeted phase of enhancement, was seen in 21 patients (28.8% improvement). The bolus-triggered technique had significantly higher accuracy when compared to fixed timing (P = 0.0001389). A 95% confidence interval for the difference in accuracy between the bolus-triggered and fixed timing method is 0.1013 to 0.3187. The image review was considered straightforward in all cases and there was 100% agreement between the reviewers' evaluation of arterial phase images. Reviewers subjectively noted no consistent detectable difference in image quality between the two techniques. In comparison to the fixed-timed method, real-time bolus monitoring required added visual awareness of contrast agent arrival in the abdominal aorta, and attention to initiating the breath hold command within the defined timing delay window. However, these steps mimicked those used in routine contrast-enhanced MRA and no cases in this study experienced notable errors in execution among technologists.

Figure 4.

Number of patient cases showing optimal (central portal vein, arrow), early (hepatic artery, arrow), or late (hepatic vein, arrow) liver enhancement using the bolus-triggered and fixed-timing technique.

Upon analysis of the image series collected from the real-time bolus track sequence in 100 cases, a mean transit time of 18.2 ± 4.1 seconds was determined, with a range of 12 to 31 seconds (Fig. 5). This represents the duration from injection site to the abdominal aorta at the celiac axis (the point when bolus-tracking was stopped). Given an approximately seven seconds delay for providing breath hold instructions, the 3D GRE was commenced an average of 25.2 ± 4.1 seconds from the beginning of contrast administration, which was significantly greater than the fixed-timing procedure conducted at 20 seconds (P < 0.0001). Aligning the empirical timing method with the perfusion results (Fig. 3), fixed timing delays of 20 and 25 seconds (with the center of k-space occurring approximately five seconds later) miss thepredefined enhancement imaging window in 50% (9/18) and 27.8% (5/18) of perfusion analysis patients, respectively.

Figure 5.

A histogram representation of timing durations from injection (t = 0) to abdominal aorta at the level of the celiac axis using the real-time bolus-tracking timing information (N = 100). The mean transit time was 18.2 ± 4.1 seconds.


Our results show that fixed-timed arterial-phase imaging was not accurate in over a quarter of patients. Although fixed-timed examinations have been used routinely in gadolinium-enhanced liver studies, the approach may fail in cases where the vascular transit time is altered, such as in patients with poor cardiac output (3). An earlier report using separate preimaging test bolus injection for timing has been shown to improve reproducible acquisition of arterial phase enhanced liver images (3). Our study objectives included extending these observations to bolus tracking methodology, with the intent to simplify and reduce the number of separate steps required to execute a patient-specific timing method. Bolus tracking and an automated triggering methodology for liver imaging has also been described previously (11, 12). However, there has been limited investigation into the delay time needed from the reference trigger point (i.e., abdominal aorta or celiac axis) to peak tumor enhancement, especially considering there are several variations of the 3D arterial-phase imaging sequence, where the center of k-space (predominant source of image contrast) can exist at the beginning or middle of data acquisition. In our study, we measured the liver perfusion-signal kinetics in individuals with known hypervascular benign and malignant tumors, and applied the results to optimize the real-time bolus-tracking strategy for reproducible arterial-phase liver imaging. Our results showed that improvement in arterial phase image capture should be expected in over one quarter (28.8%) of clinical cases using the proposed bolus-triggered methodology.

The sequence of events for arterial-phase imaging with real-time bolus tracking are : 1) start bolus-tracking scan with real-time reconstruction; 2) administer contrast agent; 3) visualize appearance of contrast at reference point (i.e., celiac axis); 4) stop bolus-tracking scan; 5) initiate breath hold commands (over a prescribed timing delay, i.e., seven seconds); and 6) execute the 3D GRE sequence. The steps provide versatility to accommodate specific variables, such as the use of noncentric 3D GRE sequences, or alternate reference points and timing delays. These adaptations, in light of these issues, are discussed below. Another key feature is that this technology is universal; since the tools presented in this study parallel contrast-enhanced MRA, optimization can be made generically, with the idea that all users can easily implement new strategies.

This study shows evidence that accurate capture of liver arterial-phase can be attributed mostly to the ability of real-time bolus tracking to resolve the strong variance in contrast arrival time into the aorta. We found a range of peak TPC of 24.0 seconds (20.8–44.8 seconds), which is evidence that vascular transit times play a large potential role in hypervascular tumor MRI, and that contrast-enhanced liver examinations should be customized to the individual patient. The TTPtumor and TPC following bolus detection in the abdominal aorta (TTPtumor+ and TPC+) was 9.6 ± 3.2 seconds and 8.1 ± 2.7 seconds, respectively, with a range among patients of 9.9 seconds. With this smaller time range, the potential for missed arterial-phase diminishes, thereby improving reproducibility. Previous work by Materne et al (11) concluded TTP contrast-to-noise ratio (CNR) of tumor was 7.5 seconds, which is similar to the TPC values measured here. A difference, at least in part, may be attributed to the acquisition technique; their study was restricted to aortic ROI placement in the plane of the tumor, which may vary to some degree along the length of the abdominal aorta, depending on where the tumor is located in the liver. Our study utilized a 3D acquisition with aortic ROIs taken in a consistent plane to reflect the position of bolus track detection.

In this study, it was found from the bolus-tracking and perfusion data that the range of vascular transit times was 18.2 seconds (12–31 seconds) and 17.3 seconds (13.3–30.6 seconds), respectively, using a standard infusion rate of 2.0 cc/second. Alternate infusion rates would also cause variance in fixed-timing schemes. While the time to peak arterial enhancement, and the time duration of increased arterial-to-liver enhancement should change in relation to the length of contrast infusion time, these effects will be amplified using fixed-timing schemes, and minimized using an aortic/hepatic arterial bolus trigger point. It is expected that optimal arterial contrast enhancement (lesion-to-liver ratio) should be achieved by infusing contrast over the shortest possible time duration. This may be a relative advantage of using the highest possible infusion rate, or shortest possible infusion duration that is possible for a given contrast. That was part of our rationale for using a contrast agent with a higher relaxivity, which then allowed for a reduction of dose to 0.05 mmol/kg, while preserving signal per dose. When given at the recommended rate of 2 cc/second, this achieves a 50% reduction in the infusion time. Even though transit data was not recorded from the fixed-timed subjects in this study, this large variance observed with a 2.0 cc/second infusion rate is likely the primary cause of erroneous arterial-phase timing with the fixed-timed method (27% error). Even though a fixed timing delay of 20 seconds was used as the standard of comparison in this study, a higher fixed timing delay of 25 seconds using the same 3D GRE sequence (center of k-space approximately 30 seconds postcontrast) would still miss the predefined imaging windows in 27.8% (5/18) of patients who underwent perfusion analysis (Fig. 3). Real-time bolus triggering effectively normalizes the contrast transit-delay time to a specific reference point (celiac axis). With a smaller transit delay following cessation of real-time bolus tracking at the celiac axis, arterial-phase imaging can be more reliably captured in a majority of altered vascular states. It is noteworthy that human interaction with operator-dependent bolus detection and activation of sequence acquisition may represent another source of timing fluctuation between patients. This increases interest in implementation of an automated bolus-track liver examination based upon machine-based bolus detection and image acquisition triggering.

Even though this study used a particular 3D GRE sequence with the center of k-space near the beginning of the acquisition, it is feasible to use other formulations of 3D GRE with bolus triggering, provided that the optimal arterial-phase image contrast is achieved when peak tumor enhancement data is acquired over the central k-space lines of the imaging sequence. Contrary to the centric 3D GRE used in this study, conventional 3D GRE sequences typically acquire k-space data linearly or radially over many segments and the center of k-space is passed multiple times over the scan duration. In effect, this averages the dynamic image contrast, such that the middle of the acquisition is representative of the peak image contrast time point. For optimized triggering using linear 3D GRE, the timing delay of eight to nine seconds should be reduced such that the duration to k = 0 falls within the time-region of high lesion contrast, which must also account for the total acquisition time. Our perfusion data shows a relatively high lesion-to-liver contrast (60% of maximum) around peak lesion contrast, which spans 12.8 ± 6.9 seconds and extends on average 15.5 seconds after bolus detection. This observation allows for some imprecision of 3D GRE timing after bolus trigger, while still enabling improved timing accuracy with high lesion arterial phase contrast.

Limitations of this study include the absence of contrast-to-noise comparisons between fixed and bolus-timed strategies. These were not possible since the absolute time-point could not be adequately controlled in the two experiments. For this, the same individual must be imaged with both techniques. Similar argument may be made regarding comparisons of alternate 3D GRE sequences or alternate contrast infusion strategies, where a higher infusion rate with shorter bolus should lead to higher lesion CNR. These represent evaluations outside the scope of this study. Another limitation is that the reason for cases of early or late enhancement using bolus tracking in our results could not be determined for every case. As discussed, one explanation is altered hepatic artery-portal vein transit times, but the absolute physiologic reason was not possible to measure in our study design. The correlation between cardiac output and vascular transit times was not included in this study, but could be the subject of future investigations. Technical factors including human operator error is another potential consideration. Automation of the triggering steps may yield even higher reproducibility. Finally, it should be noted that, while we see no marked differences between the timing characteristics of distinct types of arterial phase enhancing liver tumors, further investigation may more fully outline the differences between various hyperascular lesions.

In conclusion, we present a more detailed temporal analysis of enhancement events that follow bolus contrast administration, essential for understanding and further developing methodology for accurate and reproducible real-time bolus-triggered acquisition of transient arterial phase 3D GRE MRI of the liver. Automated bolus-tracked liver examination is feasible and may further reduce the potential for operator error while further improving reproducibility, simplicity and speed for acquiring individually compensated dynamically enhanced liver MRI.