Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Magn Reson Imaging. Author manuscript; available in PMC 2010 November 1.
Published in final edited form as:
PMCID: PMC2870717

Effect of Contrast Media on Single Shot EPI: Implications for Abdominal Diffusion Imaging

Vikas Gulani, M.D., Ph.D.,1,2,* Jonathan M. Willatt, M.D.,3 Martin Blaimer, Ph.D.,1,2,4 Hero K. Hussain, M.D.,3 Jeffrey L. Duerk, Ph.D.,1,2 and Mark A. Griswold, Ph.D.1,2



The goal of this study was to determine the effect of contrast media on the signal behavior of single shot echo planar imaging (ssEPI) used for abdominal diffusion imaging.

Materials and Methods

The signal of a ssEPI spin echo sequence in a water phantom with varying concentrations of gadolinium was modeled with Bloch equations and the predicted behavior validated on a phantom at 1.5 T. Six volunteers were given gadolinium contrast, and signal intensity (SI) time courses for regions of interest (ROIs) in the liver, pancreas, spleen, renal cortex and medulla were analyzed. The Student's t-test was used to compare pre-contrast SI to 0, 1, 4, 5, 10, and 13 minutes following contrast.


The results show that following contrast, ssEPI SI goes through a nadir, recovering differently for each organ. Maximal contrast related signal losses relative to pre-contrast signal are 20%, 20%, 53%, and 67%, for the liver, pancreas, renal cortex and medulla respectively. The SIs remain statistically below the pre-contrast values for 5, 4, and 1 minutes for the pancreas, liver, and spleen, and for all times measured for the renal cortex and medulla.


Abdominal diffusion imaging should be performed prior to contrast due to adverse effects on the signal in ssEPI.

Keywords: abdomen, diffusion, MRI, EPI, contrast, gadolinium, kidneys, pancreas, liver, spleen


Diffusion weighted imaging (DWI) is a long established technique for evaluation of the neurological system, but is relatively new in body imaging. It has, however, quickly become an important tool in the evaluation of abdominal pathology and function in the liver (1-4), pancreas (5-7), and kidneys (8-10). Applications include (but are not limited to) liver tumor detection and characterization of small lesions, evaluation and quantitation of hepatic fibrosis, diagnosis and characterization of solid and cystic pancreatic masses, evaluation of pancreatic exocrine function, evaluation of renal parenchymal disorders, and characterization of renal masses (1-10). An open clinical question is whether abdominal diffusion images should be acquired prior to or following contrast administration. The effect of contrast material could conceivably have positive, negative, or neutral effects on lesion conspicuity and detection, based on the various competing effects of contrast on signal intensity. Given the time pressure on the modern scanner schedule, it would be ideal to fit diffusion imaging into the present imaging protocols without expending additional table time. The typical abdominal imaging protocol consists of precontrast imaging, followed by multiple timed post-contrast series imaged at 20 – 180 seconds post contrast, and then delayed imaging at 4-6 minutes post contrast and in some cases 10-15 minutes post contrast (for example in the evaluation of cholangiocarcinoma). Thus there is a window of approximately 1 to 3 minutes between the 3 and 4-6 minute post contrast scans which could potentially be used for diffusion imaging. There are published studies in the literature stating that the signal intensity (SI) and measured apparent diffusion coefficient (ADC, a quantitative measure of the diffusivity) in the brain are not affected by administration of contrast (11-13) and that the same is true in the liver (14).

The anecdotal experience in the authors' experience, however, had been to the contrary: DWI of the abdomen obtained for clinical purposes appeared to be of higher quality if obtained before, rather than after, contrast. This hypothesis, however, must be proven with data.

Gadolinium contrast agents are utilized primarily for their T1 shortening effects, but it is well known that these agents also cause T2 shortening (15). The two properties have opposite consequences – T1 shortening causes an increase in signal intensity, particularly on T1 weighted images, while T2 or T2* shortening causes a decrease in signal intensity. The latter effect can dominate at high gadolinium concentrations, a phenomenon which is seen, for example in the bladder when gadolinium concentrates there after contrast excretion. The two effects are expected to also have opposing effects on lesion detectability – the improved signal to noise ratio (SNR) caused by T1 shortening will improve the available signal for additional diffusion weighting and also likely improve lesion detection, while the T2 shortening will degrade the signal, causing opposing effects on image quality, lesion detectability, and ADC quantification. Thus the purpose of this study was to quantitatively evaluate the effect of gadolinium administration on the quality of diffusion weighted images. It was hypothesized that T2 shortening provided by circulating contrast media significantly decreases the signal to noise ratio (SNR) in the heavily T2-weighted, single shot EPI images, leading to image degradation. This hypothesis was tested with theoretical calculation and experimental verification of predicted signal intensity in a simple spin echo EPI experiment on a gadolinium phantom of varying concentrations and time-course experiments on human volunteers.

Materials and Methods

Model/Phantom Experiment

The expected relaxation rates R1 and R2 in a saturation recovery experiment in the presence of various concentrations of contrast agent was modeled by the well accepted relationships (16):

[Equation 1]
[Equation 2]

where C is the concentration, T1,0 and T2,0 are the published intrinsic T1 and T2 values without contrast (6), and the longitudinal and transverse relaxivities α1 and α2 were obtained experimentally from measurements on gadolinium solutions. An expected signal versus concentration curve for water was generated from simple Bloch equations assuming TR/TE of 2400/71 ms. Imaging was performed on a 1.5 T Siemens Espree (Siemens Medical Solutions, Erlangen, Germany). All data analysis was performed offline in MATLAB (The Mathworks, Natick MA). The predicted behavior for water was tested on spin echo EPI on a phantom with gadolinium solutions of various concentrations ranging from 0 to 5 mM (imaging performed with a 12 channel body array coil, ssSE-EPI, TR/TE= 2400/71 ms, parallel imaging factor 2, 38 cm field of view (FOV), 150×150 matrix (MX), 5 mm slice thickness).

Volunteer Study

The study was performed under a protocol approved by the Institutional Review Board (IRB) and is HIPAA compliant. Informed written consent was obtained from all participants. Asymptomatic volunteers with no known renal dysfunction were administered gadolinium contrast (N=6, gadoversetamide - Optimark, Mallinkrodt Inc., St. Louis, MO, 0.1 mmol/mL, 0.1 mmol/kg up to maximum volume of 20 mL, 2 cc/sec followed by a 20 mL saline push), and serial ssSE-EPI images were obtained every minute (TR/TE 2400/71 ms, 36-38 cm FOV, 162×162 matrix, 5 mm slice thickness, b=0 & 500 s/mm2). SI was measured in 0.2 cm2 ROIs in the renal cortex and medulla, liver, pancreas, and spleen from the b=0 images, at two time points prior to contrast, and then at thirteen time points separated by 1 minute each, beginning immediately following contrast and ending at 13 minutes post injection. A total of 15 time points were thus measured. In the case of the kidneys, SI timecourses were measured for the renal cortex and medulla bilaterally. Since MRI SI is in arbitrary units and can vary widely across subjects, the measured SI at each time point was normalized for each subject and each organ by dividing the SI by the mean initial SI over the two measurements at 1 and 2 minutes prior to contrast administration. The mean signal intensity at each time point was also calculated and plotted for each organ over the individual subject time courses. The error on the mean SI was plotted as the standard deviation (sd) on the mean at each time point. An overall mean timecourse with error bars was plotted for each organ to compare the signal intensity behaviors of the organs. SNR measurements were not used, because with multicoil, parallel imaging data, noise is not uniform across the image and SNR measurements are not accurate. However, since noise is expected to be relatively stable over the timescale of these experiments, measuring a signal change in an ROI over time should be directly proportional to underlying changes in SNR.

For each organ, the normalized mean SI prior to contrast was compared to the intensity immediately post contrast, and 1, 4, 5, 10, and 13 minutes following contrast, using a two-tailed paired Student's t-test to analyze statistical differences in signal intensity.


The predicted SIs for a simple spin-echo experiment for water in the presence of various gadolinium concentrations and the measured signal intensity on a phantom are shown in Figure 1 in blue and green, with the measured intensities tracking with the theoretically predicted values.

Figure 1
Simulation (blue) and measured phantom (green) signal intensity for the predicted and actual behavior of a single shot spin-echo EPI sequence at different gadolinium concentrations.

Normalized signal time course data from volunteers for the liver, pancreas, spleen, renal cortex and renal medulla are shown in Figures 2--6.6. Each individual time course from a given subject is plotted with a different symbol and connected by corresponding lines, and the average time course (average +/- sd) is shown as a thick solid red line. For the renal cortex and renal medulla, the time courses for the right side in each subject are plotted as described above, while the time courses on the left side are connected by a dashed line. The average time course is plotted as for the other organs. The average time course for each organ/region is also plotted in Figure 7, to summarize these data and allow comparison between different organs. Finally, Figure 8 depicts images of the right kidney at 0 min, 1 min, 2 min, and 5 min post contrast administration, illustrating the effect of contrast on single shot EPI of the kidneys. Note the severe degradation of the image quality in the b=0/500 s/mm2 images in Figure 8 (g and h).

Figure 2
Individual time course and average (thick red line) normalized signal intensity (SI/SIave) in the liver. Error bars for the average signal intensity are calculated as the standard deviation on the mean for each time point.
Figure 6
Individual time course and average (thick red line) normalized signal intensity (SI/SIave) in the renal medulla. Error bars for the average signal intensity are calculated as the standard deviation on the mean for each time point. For each subject, data ...
Figure 7
Normalized signal intensity (SI/SIave) time courses in the liver, pancreas, spleen, renal cortex, and renal medulla.
Figure 8
Single shot EPI images cropped from a single time course to show the kidneys with b=0 (a-d) and b=500 s/mm2 (e-h), at times 0 (a,e), 1 min (b,f), 2 min (c,g), and 5 min (d,h) post contrast administration.

Table 1 summarizes normalized signal intensities prior to contrast (-1 minutes), immediately post contrast (0 minutes), and at 1, 4, 5, 10, and 13 minutes post contrast. The two tailed paired Student's t-test comparisons between the precontrast (t=-1 minute) and the post contrast time points above are given in parenthesis within the same table. Comparisons reaching statistical significance (p<0.05) are denoted in bold.

Table 1
Change in Normalized Signal Intensities Over Time


As can be seen from Figure 1, at very low gadolinium concentrations, water signal intensity is expected to increase, but then decrease as gadolinium concentration increases. The measured phantom data (green) follow the predicted trend, though at higher contrast agent concentrations measured signal intensities are slightly lower than the model. This may relate to variations in contrast agent concentration, and also to the fact that the model we employ is simplistic, using only pure T1 and T2 effects in the Bloch equations, and not taking into account other mechanisms of signal loss such as off-resonance distortions or relaxation during RF pulses. The important point, however, is that at higher concentrations of contrast agent, the signal loss caused by T2 decay outweighs the enhancement effect of the agent. Since adding diffusion weighting causes a significant net loss of signal, the goal is generally to preserve as much signal as possible so that there is sufficient signal to allow acquisition of higher b-value DWI, and improve lesion detection and characterization. At higher concentrations of gadolinium, this experiment clearly shows that the baseline signal intensity for EPI would be non ideal (i.e., non-maximal) for DWI.

The volunteer time course data show reproducible signal behavior in EPI images of the liver, pancreas, spleen, and renal cortex and medulla (Figures 2--7,7, numerical comparisons in Table 1). These organs/regions were chosen because these are the major areas of interest in the upper abdomen, and DWI of the regions must be performed with these regions in mind. There is a statistically significant drop in signal in all organs after injection of contrast. With the exception of the spleen, in which there is a recovery of signal within two minutes, it takes several minutes for the signal in the remaining organs to return to statistical baseline (presumably the T1 shortening effects of the contrast agent dominate in the spleen). An additional observation about signal behavior in the liver is that there is a second sharp decrease in signal intensity between 2 and 3 minutes after contrast. This drop (approximately 8%) is also found to be significant (p=0.005) and may relate to arrival and concentration of contrast from the portal circulation, though this hypothesis would require further experimental testing.

There is actually a reversal of cortex/medulla contrast in the kidney roughly 2 minutes after injection (Figures 5 - -8).8). Importantly, for the liver, pancreas, renal cortex and medulla, there are contrast related signal losses of 20%, 20%, 53%, and 67%, respectively, all of which are statistically significant. The signal is measured to be statistically below baseline for the first four minutes for liver, first eight minutes and at 13 minutes for pancreas (signal measurements and t-test calculations for minutes 6-8 are not shown in Tables 1 and 2), and beyond 13 minutes for either renal cortex or medulla. These results confirm the preliminary observations on a single subject published in abstract form previously (17). The results for these organs indicate that if images are obtained after contrast injection there would be loss of valuable signal, a key problem for an already SNR starved technique such as DWI which relies on further signal loss to achieve the desired image contrast and quantification. Moreover, since dynamic contrast enhanced images are typically obtained at 20 to 180 seconds after contrast, if the DWI were to be performed post contrast, images would be most likely obtained 4-6 minutes after injection, the signal nadir for the kidneys and pancreas, and a time at which liver signal is also low. This means that the DWI would be performed at the most sub-optimal time if the organ of interest is any of these three. Even if diffusion imaging is performed more than 10 minutes post injection of contrast, the situation is suboptimal for evaluating the kidneys, pushing forward a strong argument that diffusion weighting should not be performed in the post-contrast set of sequences in the abdomen and doing so could negatively affect the diagnostic performance of the sequence. The drop in signal could result in low SI lesions to drop to noise level, rendering them indetectable, and could adversely affect the conspicuity of other lesions. Quantitation of ADC in lesions and surrounding parenchyma will be adversely affected because less signal will be available to obtain higher diffusion weighting, and also quantitation of the ADC is less accurate with noisier data.

Figure 5
Individual time course and average (thick red line) normalized signal intensity (SI/SIave) in the renal cortex. Error bars for the average signal intensity are calculated as the standard deviation on the mean for each time point. For each subject, data ...

The observed behavior is different from that reported in the literature on the brain where it has been shown that the presence of gadolinium contrast has little effect on diffusion weighted imaging (11-13). This difference likely relates to a higher concentration of contrast agent in abdominal organs (particularly the kidney), and the absence of a blood brain barrier in the abdomen. Also, the dynamic nature of enhancement in the abdomen has a clear effect on the signal behavior in the abdomen after contrast administration. The most likely explanation for the difference between these results and those reported for the liver previously, which suggested that there is no difference in the SNR of liver DWI obtained before and after contrast, is that the single time point used for post contrast imaging in the previous study is long enough post contrast that the liver signal had recovered to baseline or near baseline (14). Also, the previous study was performed at 3.0 T and the present study is at 1.5 T. Relaxation properties due to contrast agent are different at high fields, and thus time course SIs could differ as well.

As long as the DWI is performed at a time where the signal behavior of the sequence is not dynamically changing, one would not expect the ADC to change significantly due to the presence of contrast. These experiments show (Figures 2--77 and Tables 1 and 2), however, that dynamically changing signal behavior would be nearly impossible to avoid after contrast, particularly in the kidneys. The adverse effect of data noise on the accuracy of calculated diffusion coefficients in both isotropic and anisotropic diffusion was not evaluated here, but has been previously well documented (18-20).

As mentioned previously, lesion characterization and ADC quantification, especially tumor detection and characterization in the liver, pancreas and kidneys are critical reasons to perform diffusion imaging in the abdomen. Therefore an important additional consideration is the effect of contrast on tumors. The dynamics of contrast behavior in tumors and the effects on EPI are not predictable, and indeed likely vary from tumor to tumor. In the liver, for example, the contrast related alteration in ssEPI signal behavior may be drastically different for an arterially perfused tumor such as hepatocellular carcinoma than for a relatively poorly perfused mass such as cholangiocarcinoma. It is not possible to say a priori whether a given tumor will behave like the renal cortex/medulla in which the SI is adversely affected 13 minutes or more beyond the administration of contrast, or more like the spleen, where there may even be an increase in baseline signal due to contrast. If for a given mass at a given time point there is a contrast related signal loss sufficient to drop the higher b value SI to noise levels, then the measured ADC would also be affected (the calculated ADC can be predicted to be lower than the actual value due to the SI hitting a noise floor and not changing due to increased diffusion weighting). This unpredictable behavior is another reason to avoid diffusion imaging in the abdomen after contrast.

In conclusion, this work shows that in planning whether to place the diffusion sequences for the abdomen before or after contrast, the effect of the contrast on the EPI signal intensity should be taken into account, and assuming normal physiology, the DWI should be performed before contrast. If the decision is made to place the diffusion imaging after contrast, a minimum of 6 minutes post contrast should be allowed when normal circulation is expected, in which case the effect of changing signal behavior in the SS-EPI sequences generally employed will likely preclude accurate characterization of the kidneys.

Figure 3
Individual time course and average (thick red line) normalized signal intensity (SI/SIave) in the pancreas. Error bars for the average signal intensity are calculated as the standard deviation on the mean for each time point.
Figure 4
Individual time course and average (thick red line) normalized signal intensity (SI/SIave) in the spleen. Error bars for the average signal intensity are calculated as the standard deviation on the mean for each time point.


Grant support: NIH 1KL2RR024990 (VG), NIH 1R21DK073649-01A2 (HKH), and Siemens Medical Solutions (JLD and MAG)


1. Koinuma M, Ohashi I, Hanafusa K, Shibuya H. Apparent diffusion coefficient measurements with diffusion-weighted magnetic resonance imaging for evaluation of hepatic fibrosis. J Magn Reson Imaging. 2005;22:80–85. [PubMed]
2. Naganawa S, Sato C, Nakamura T, et al. Diffusion-weighted images of the liver: comparison of tumor detection before and after contrast enhancement with superparamagnetic iron oxide. J Magn Reson Imaging. 2005;21:836–840. [PubMed]
3. Quan XY, Sun XJ, Yu ZJ, Tang M. Evaluation of diffusion weighted imaging of magnetic resonance imaging in small focal hepatic lesions: a quantitative study in 56 cases. Hepatobiliary Pancreat Dis Int. 2005;4:406–409. [PubMed]
4. Sun XJ, Quan XY, Huang FH, Xu YK. Quantitative evaluation of diffusion-weighted magnetic resonance imaging of focal hepatic lesions. World J Gastroenterol. 2005;11:6535–6537. [PubMed]
5. Inan N, Arslan A, Akansel G, et al. Diffusion-Weighted Imaging in the Differential Diagnosis of Cystic Lesions of the Pancreas. American Journal of Roentgenology. 2008;191:1115–1121. [PubMed]
6. Lee SS, Byun JH, Park BJ, et al. Quantitative analysis of diffusion-weighted magnetic resonance imaging of the pancreas: Usefulness in characterizing solid pancreatic masses. Journal of Magnetic Resonance Imaging. 2008;28:928–936. [PubMed]
7. Erturk SM, Ichikawa T, Motosugi U, et al. Diffusion-Weighted MR Imaging in the Evaluation of Pancreatic Exocrine Function Before and After Secretin Stimulation. The American Journal of Gastroenterology. 2006;101:133–136. [PubMed]
8. Cova M, Squillaci E, Stacul F, et al. Diffusion-weighted MRI in the evaluation of renal lesions: preliminary results. Br J Radiol. 2004;77:851–857. [PubMed]
9. Squillaci E, Manenti G, Cova M, et al. Correlation of diffusion-weighted MR imaging with cellularity of renal tumours. Anticancer Res. 2004;24:4175–4179. [PubMed]
10. Thoeny HC, De Keyzer F, Oyen RH, Peeters RR. Diffusion-weighted MR imaging of kidneys in healthy volunteers and patients with parenchymal diseases: initial experience. Radiology. 2005;235:911–917. [PubMed]
11. Yamada K, Kubota H, Kizu O, et al. Effect of intravenous gadolinium-DTPA on diffusion-weighted images: evaluation of normal brain and infarcts. Stroke. 2002;33:1799–1802. [PubMed]
12. Fitzek C, Mentzel HJ, Fitzek S, et al. Echoplanar diffusion-weighted MRI with intravenous gadolinium-DTPA. Neuroradiology. 2003;45:592–597. [PubMed]
13. Chen G, Jespersen SN, Pedersen M, et al. Intravenous administration of Gd-DTPA prior to DWI does not affect the apparent diffusion constant. Magn Reson Imaging. 2005;23:685–689. [PubMed]
14. Chiu FY, Jao JC, Chen CY, et al. Effect of intravenous gadolinium-DTPA on diffusion-weighted magnetic resonance images for evaluation of focal hepatic lesions. J Comput Assist Tomogr. 2005;29:176–180. [PubMed]
15. Caravan P, Ellison JJ, McMurry TJ, Lauffer RB. Gadolinium(III) Chelates as MRI Contrast Agents: Structure, Dynamics, and Applications. Chem Rev. 1999;99:2293–2352. [PubMed]
16. Haacke EM, Brown RW, Thompson MR, Venkatesan R. Magnetic Resonance Imaging: Physical Principles and Sequence Design. New York: John Wiley & Sons; 1999.
17. Gulani V, Blaimer M, Nour SG, et al. International Society of Magnetic Resonance in Medicine. Berlin: 2007. Effect of Contrast media on the Signal Intensity of Single-Shot EPI for Diffusion Imaging of the Body; p. 3833.
18. Pierpaoli C, Basser PJ. Toward a quantitative assessment of diffusion anisotropy. Magnetic Resonance in Medicine. 1996;36:893–906. [PubMed]
19. Basser PJ, Pajevic S. Statistical artifacts in diffusion tensor MRI (DT-MRI) caused by background noise. Magnetic Resonance in Medicine. 2000;44:41–50. [PubMed]
20. Chen B, Hsu EW. Noise removal in magnetic resonance diffusion tensor imaging. Magnetic Resonance in Medicine. 2005;54:393–401. [PubMed]