Diffusion-weighted imaging of the abdomen at 3.0 Tesla: Image quality and apparent diffusion coefficient reproducibility compared with 1.5 Tesla

Authors


Abstract

Purpose

To compare single-shot echo-planar imaging (SS EPI) diffusion-weighted MRI (DWI) of abdominal organs between 1.5 Tesla (T) and 3.0T in healthy volunteers in terms of image quality, apparent diffusion coefficient (ADC) values, and ADC reproducibility.

Materials and Methods

Eight healthy volunteers were prospectively imaged in this HIPAA-compliant IRB-approved study. Each subject underwent two consecutive scans at both 1.5 and 3.0T, which included breathhold and free-breathing DWI using a wide range of b-values (0 to 800 s/mm2). A blinded observer rated subjective image quality (maximum score= 8), and a separate observer placed regions of interest within the liver, renal cortices, pancreas, and spleen to measure ADC at each field strength. Paired Wilcoxon tests were used to compare abdominal DWI between 1.5T and 3.0T for specific combinations of organs, b-values, and acquisition techniques.

Results

Subjective image quality was significantly lower at 3.0T for all comparisons (P = 0.0078– 0.0156). ADC values were similar at 1.5T and 3.0T for all assessed organs, except for lower liver ADC at 3.0T using b0-500-600 and breathhold technique. ADC reproducibility was moderate at both 1.5T and 3.0T, with no significant difference in coefficient of variation of ADC between field strengths.

Conclusion

Compared with 1.5T, SS EPI at 3.0T provided generally similar ADC values, however, with worse image quality. Further optimization of abdominal DWI at 3.0T is needed. J. Magn. Reson. Imaging 2011;33:128–135. © 2010 Wiley-Liss, Inc.

DIFFUSION WEIGHTED MRI (DWI) with quantitative measurement of apparent diffusion coefficient (ADC) values is an increasingly applied technique in the abdomen. ADC measurements have shown potential in the evaluation of diffuse parenchymal disease such as liver fibrosis, chronic pancreatitis, and renal disease (1–7), as well as for the detection, assessment of grade and stage, and assessment of treatment response for abdominal and pelvic tumors (8–20). Despite an increasing spectrum of applications, DWI is not without limitations. For instance, most conventional DWI sequences use a single-shot (SS) echoplanar-imaging (EPI) technique that is prone to artifacts related to susceptibility and eddy currents, leading to image ghosting and distortion (21, 22). Also, DWI using an EPI technique is characterized by poor signal-to-noise ratio (SNR) (23). An additional limitation of DWI that impacts its quantitative application is that the ADC values obtained from DWI are impacted by patient-related, hardware-related, and sequence-related factors, thereby exhibiting an inherent degree of variability (24–30). Data evaluating the reproducibility of ADC values within the abdomen are limited: one recent study reported a mean coefficient of variation (CV) of 14% for abdominal ADC values at 3.0T (31), while another study reported a CV of ADC of 3.2% in the liver at 1.5T (4). To be able to apply ADC values clinically, it is critical to have an understanding of the reproducibility of these measurements.

The increasing availability of 3.0T systems introduces an important technical factor to consider for optimization of DWI, as it is possible that abdominal DWI may exhibit better image quality or ADC reproducibility at a particular field strength. Furthermore, if there were significant differences in ADC values between 1.5T and 3.0T, this would limit the ability to follow patients with intra-abdominal pathology at different field strengths as well as limit the applicability of the results of studies performed at 1.5T. As an example of the impact of field strength on DWI, brain diffusion studies have shown higher SNR but increased image distortion at 3.0T compared with 1.5T (21, 23), in addition to substantial differences in ADC values between these field strengths (32). While a recent study observed differences in ADC of the liver between 1.5T and 3.0T with no difference in ADC of the remaining abdominal organs, this study did not assess image quality or ADC reproducibility at a given field strength (33).

The objective of our study was to compare SS EPI DWI of abdominal organs in healthy volunteers between 1.5T and 3.0T in terms of image quality, ADC values, and ADC reproducibility.

MATERIALS AND METHODS

Subjects

This prospective study was compliant with the Health Insurance Portability and Accountability Act and was approved by our local Institutional Review Board. Eight healthy volunteers (7 males, 1 female; mean age, 33 ± 12.8 years; range, 23– 62 years) were recruited for the study. Signed written informed consent was obtained from all volunteers. All scans were completed over a span of 5 months (August 2008 through December 2008).

MR Examination

Each volunteer underwent a total of four scans, consisting of two scans back-to-back on a 1.5T system (Magnetom Avanto, Siemens Healthcare, Erlangen, Germany, running software Syngo MR B15 at time of the study) and two scans back-to-back on a 3.0T system (Magnetom Trio, Siemens Healthcare, running software Syngo MR B13 at the time of the study). The 1.5T and 3.0T scans occurred on separate dates for all patients, separated by a mean delay of 12 days (range, 7– 21 days). All imaging was performed using a six-channel torso coil array and a six-channel spine coil array. Between the first and second scans at each of the two field strengths, the surface coil was removed and the patient taken out of the bore of the magnet, before putting the coil on again and proceeding with the second scan. No intravenous contrast was administered. Two scans were performed at both 1.5T and at 3.0T to allow for assessment of ADC reproducibility at each field strength as well as between field strengths. Each scan started with acquisition of a localizer sequence and an axial in- and out-of-phase T1-weighted sequence through the upper abdomen to assist in slice placement for the subsequent DWI sequences.

Each of the four scans for a given patient included six axial EPI diffusion-weighted sequences. The following parameters were constant for all six DWI sequences for all four scans for all subjects: TR 1900 ms, TE 76 ms (kept constant for all b-values), flip angle 90°, field of view (FOV) 262 × 350, matrix 144 × 192 (interpolated to 288 × 384), 10 slices, slice thickness 7 mm, 20% interslice gap, receiver bandwidth 1370 Hz/voxel, 1 signal average, parallel imaging using the generalized auto-calibrating partially parallel acquisition (GRAPPA) technique with parallel acquisition factor of 2. In addition, all scans were performed using fat suppression, achieved by means of the spectral-adiabatic inversion recovery (SPAIR) technique at 1.5T and by means of chemically selective fat suppression at 3.0T. Each of the first five diffusion acquisitions was obtained within a single breathhold of 19 s at end-expiration and using three b-values per acquisition. The combination of b-values was as follows: 0-50-100, 0-150-200, 0-300-400, 0-500-600, and 0-700-800 s/mm2. The sixth diffusion acquisition was performed using a free-breathing technique (without respiratory trigger) and with all 11 of the b-values that were used during the various breathhold acquisitions (0, 50, 100, 150, 200, 300, 400, 500, 600, 700, and 800 s/mm2), with a total acquisition time of 65 s. A b-value of 0 s/mm2 was included in each acquisition as no attempt was made as part of this study to separate perfusion and true diffusion effects (4, 27, 34). All of the diffusion sequences were performed using tri-directional diffusion gradients, with inline reconstruction of the trace images and of the ADC map, calculated using a mono-exponential fit for each sequence.

Data Analysis

Image Quality at 1.5T and 3.0T

Image quality was assessed for each scan using the trace images that were obtained with a representative baseline, intermediate, and high diffusion-weighting (specifically, the images obtained with b-values of 0, 400, and 800 s/mm2, hereafter referred to as the b-0, b-400, and b-800 images, respectively). Observer 1 (A.R., a board certified radiologist with 2 years of experience in body MRI at the time of image evaluation) subjectively rated image quality on the b-0, b-400, and b-800 images for each sequence for each scan and assigned a score from 1 to 4 for each of conspicuity of organ edges and ghosting/distortion artifacts (1 = poor image quality, considered nondiagnostic; 2 = fair image quality, somewhat impairing diagnostic quality; 3 = good image quality, not impairing diagnostic quality; 4 = excellent image quality). These two scores were summed to provide a maximal image quality score of 8 per subject for each sequence. The images were presented in random order to this observer, who was blinded to field strength, b-value, and breathhold or free-breathing technique.

ADC Values of Abdominal Organs at 1.5T and 3.0T

A separate observer (M.O., observer 2, a fourth year medical student who had undergone a training session by B.T., a board certified radiologist with 6 years of experience in body MRI) placed regions of interest (ROIs) on the b-0 image that was included in each separately acquired sequence for each scan using a dedicated MRI processing workstation (Leonardo, Siemens Healthcare). ROIs were placed on the right lobe of the liver (the left hepatic lobe was not assessed due to the risk of cardiac motion artifacts), spleen, pancreatic body, and both renal cortices (Fig. 1). Care was taken to avoid vessels and artifacts in ROI placement. Three circular 100-pixel ROIs were placed within each anatomic location (aside from the right and left renal cortices, within which two ROIs were placed in each). The four scans (two at 1.5T and two at 3.0T) for each subject were viewed simultaneously, with ROIs placed in a location as similar as possible between scans. The ROIs were transferred from the b-0 images to the corresponding ADC maps (Fig. 1). The mean ADC obtained from the 3 ROIs placed on each organ for each sequence (aside from the four ROIs placed on the kidneys) was calculated to yield a single ADC value per organ for each sequence. Therefore, a total of four ADC values were obtained for each organ for each sequence (two at 1.5T and two at 3.0T) in each subject.2

Figure 1.

a–l: Axial fat suppressed breathhold single shot echo-planar diffusion-weighted images of the abdomen in a 39-year-old healthy male volunteer (TR 1900/TE 76/matrix 144 × 192, slice thickness 7 mm, b-values of 0, 700, and 800 s/mm2, b0 images shown only). Images were obtained at both 1.5T (two left columns) and 3.0T (two right columns). Three circular regions of interest (ROIs, circles) of approximately 100 pixels were placed within the right hepatic lobe, pancreatic body, and spleen, and two circular ROIs within the left and right kidneys, on the b0 image (a,c,e,g,i,k) and copied to the corresponding ADC maps (b,d,f,h,j,l). Note increased ghosting on the 3.0T images (arrows on c,g,k).

Figure 2.

Axial fat suppressed breathhold single shot echo-planar diffusion-weighted images of the abdomen in a 23-year-old healthy male volunteer (TR 1900/TE 76/matrix 144 × 192, slice thickness 7 mm, b-values of 0, 400, and 800 s/mm2). The b-0, b-400, and b-800 diffusion images obtained at 1.5T (top row) and 3.0T (bottom row) are shown. Note increased ghosting (arrows) on 3.0T images.

Statistical Analysis

MedCalc version 10.4 (Frank Schoonjans, Mariakerke, Belgium) was used for all computations. A series of paired Wilcoxon tests was used to compare subjective image quality, ADC, and coefficient of variation (CV= SD ADC/mean ADC) between 1.5T and 3.0T measurements. A single abdominal organ, b-value combination, and acquisition technique (breathhold versus free-breathing) was assessed for each comparison, such that each subject only contributed one set of values at 1.5T and 3.0T for each comparison. For the comparisons of subjective image quality and ADC between 1.5T and 3.0T, the values obtained for each of the two scans at a given field strength were averaged to obtain a single value for each magnet. All reported P values are two-sided and considered statistically significant when less than 0.05.

RESULTS

Image Quality

There was a trend toward lower subjective image quality scores at increasing b-values for both breathhold and free-breathing DWI at both 1.5T and 3.0T (Table 1; Figs. 1, 2). At each of the selected b-values (0, 400, and 800 s/mm2) and for both breathhold and free-breathing techniques, subjective image quality within the abdomen was statistically better at 1.5T than at 3.0T (P ranging from 0.0156 to 0.0078 for all comparisons) (Table 1).

Table 1. Mean and Standard Deviation (SD) of Image Quality Scores for Each Field Strength at b-0, b-400, and b-800 for Breathhold and Free-Breathing DWI Techniques Obtained in 8 Volunteers at 1.5 and 3.0T*
b-Value1.5T3.0TP**
 Breathhold
  • **

    Maximum image quality score = 8. Scores were significantly lower for 3.0T for all b-values. P values for significant differences are bolded.

  • b

    Paired Wilcoxon test.

08.0 ± 0.06.9 ± 0.60.0156
4007.8 ± 0.55.6 ± 0.80.0078
8006.9 ± 0.04.5 ± 1.10.0078
 Free-breathing
08.0 ± 0.06.6 ± 0.70.0078
4007.5 ± 0.55.6 ± 1.20.0078
8006.6 ± 1.04.0 ± 1.20.0078

Comparison of ADC Values Between 1.5T and 3.0T

In one volunteer, the prescribed field of view did not cover the pancreas; therefore, no pancreatic measurements were obtained for this subject, and only seven subjects were included in the assessment of pancreatic ADC (Table 2).

Table 2. Mean and Standard Deviation (SD) of ADC (in mm2/s) for Each Abdominal Organ With Breathhold and Free-Breathing DWI at 1.5 and 3.0T*
Organb-Values1.5T3.0TP**
 Breathhold
  • **

    There were no significant differences between 1.5 and 3.0T, except for lower liver ADC for breathhold b0, 500, 600 at 3.0T. P values for significant differences are bolded.

  • b

    Paired Wilcoxon test.

Liver0, 50, 1002.85 ± 0.422.97 ± 0.700.4609
Liver0,150, 2002.34 ± 0.422.13 ± 0.630.5469
Liver0, 300, 4001.63 ± 0.231.49 ± 0.470.3828
Liver0, 500, 6001.41 ± 0.291.21 ± 0.310.0156
Liver0, 700, 8001.20 ± 0.251.12 ± 0.360.4609
Kidney0, 50, 1003.01 ± 0.292.87 ± 0.410.5469
Kidney0,150, 2002.68 ± 0.242.62 ± 0.240.6875
Kidney0, 300, 4002.39 ± 0.162.34 ± 0.160.1484
Kidney0, 500, 6002.19 ± 0.152.14 ± 0.140.3828
Kidney0, 700, 8002.09 ± 0.12.05 ± 0.160.3828
Pancreas0, 50, 1002.94 ± 0.382.92 ± 0.350.9375
Pancreas0,150, 2002.76 ± 0.462.55 ± 0.510.4687
Pancreas0, 300, 4001.89 ± 0.131.80 ± 0.290.2969
Pancreas0, 500, 6001.59 ± 0.151.58 ± 0.241.00
Pancreas0, 700, 8001.47 ± 0.111.46 ± 0.171.00
Spleen0, 50, 1001.26 ± 0.261.44 ± 0.360.2500
Spleen0,150, 2001.07 ± 0.131.21 ± 0.180.1094
Spleen0, 300, 4000.98 ± 0.121.05 ± 0.080.2187
Spleen0, 500, 6000.97 ± 0.260.99 ± 0.150.7422
Spleen0, 700, 8000.91 ± 0.090.95 ± 0.230.7422
 Free-breathing
Liver0 to 8001.03 ± 0.330.89 ± 0.310.0781
Kidney0 to 8002.08 ± 0.131.99 ± 0.160.1484
Pancreas0 to 8001.26 ± 0.161.39 ± 0.250.0937
Spleen0 to 8000.96 ± 0.020.98 ± 0.210.9453

At both 1.5T and 3.0T, all organs demonstrated a decrease in ADC as the b-value increased, as shown previously (25,26,33). There was no significant difference between ADCs at 1.5T and 3.0T for any organ using any b-value or breathing technique, except for lower liver ADC for breathhold b0, 500, 600 at 3.0T. Although seven of the eight volunteers had lower liver ADC at 3.T with free-breathing technique, this difference did not reach statistical significance. (Table 2; Fig. 3)

Figure 3.

Bar graphs show the ADC values measured for liver, spleen, kidneys (average between right and left kidney), and pancreas in each of eight volunteers (seven volunteers for the pancreas) at 1.5T and 3.0T using free-breathing technique and b-values ranging from 0 through 800 s/mm2. Dark gray bar = 1.5T, light gray bar = 3.0T. For this sequence, there was a trend toward lower liver ADC values at 3.0T in most volunteers, without reaching significance.

One of the volunteers had a consistently low liver ADC on each of the assessed sequences. Review of this patient's in- and out-of-phase T1-weighted images demonstrated substantial loss of signal in the liver on the in-phase images, consistent with iron deposition, which may account for the low ADC.

Comparison of ADC Reproducibility Between 1.5T and 3.0T

CV of ADC was generally similar between field strengths, with no statistically significant difference in reproducibility of ADC between 1.5T and 3.0T for any of the performed comparisons (Table 3). The CVs ranged for liver from 9.7% to 16.2% at 1.5T and from 9.9% to 16.1% at 3.0T, for kidney from 3.7% to 7.1% at 1.5T and from 3.2% to 11.5% at 3.0T, for pancreas from 8.0% to 17.7% at 1.5T and from 6.7% to 16.0% at 3.0T, and for spleen from 8.9% to 12.2% at 1.5T and from 8.6% to 28.5% at 3.0T. Based upon these ranges, ADC reproducibility was considered excellent in the kidney (in general, excellent reproducibility is defined by CV < 10%) and moderate in the liver, spleen, and pancreas.

Table 3. Mean Coefficient of Variation (%) of ADC for Each Abdominal Organ for Both Breathhold and Free-Breathing DWI at 1.5 and 3.0T
Organb-Values1.5T3.0TP*
 Breathhold
  • *

    Paired Wilcoxon test. There was no significant difference between 1.5T and 3.0T for any comparison.

Liver0, 50, 10013.511.60.9453
Liver0,150, 20016.216.10.9453
Liver0, 300, 40011.812.41.00
Liver0, 500, 60014.29.90.2500
Liver0, 700, 8009.713.50.8438
Kidney0, 50, 1007.111.50.6406
Kidney0,150, 2004.05.40.6406
Kidney0, 300, 4004.63.20.6406
Kidney0, 500, 6004.14.40.9453
Kidney0, 700, 8004.56.00.5469
Pancreas0, 50, 10011.116.00.8125
Pancreas0,150, 20017.713.60.4687
Pancreas0, 300, 40012.56.70.8125
Pancreas0, 500, 60012.110.20.3750
Pancreas0, 700, 8008.08.80.5781
Spleen0, 50, 10010.828.50.2301
Spleen0,150, 20011.98.60.9453
Spleen0, 300, 40012.211.00.8438
Spleen0, 500, 6009.110.40.5469
Spleen0, 700, 8008.912.50.8438
 Free-breathing
Liver0 to 80011.515.30.4609
Kidney0 to 8003.76.40.3828
Pancreas0 to 80011.08.20.9375
Spleen0 to 80011.615.80.2301

DISCUSSION

There is increasing evidence supporting the clinical utility of ADC quantification in the abdomen as well as of the serial assessment of such values. For instance, ADC values have shown potential for the characterization of parenchymal diseases in the liver, kidney, and pancreas (1–7), as well as for the prediction and follow-up of treatment response of malignancies such as hepatocellular carcinoma and colorectal liver metastases (8–20). With the increasing availability of 3.0T systems, it is important to recognize any possible impact of field strength upon ADC values of abdominal organs; such an impact could limit the ability to perform serial assessment of ADC values using magnets of different field strengths. If ADC values do vary between 1.5T and 3.0T, then it would be useful to know if it is preferable to perform studies incorporating ADC quantification at a particular field strength in terms of achieving optimal image quality and ADC reproducibility across serial examinations.

We did not observe significant differences in ADC between 1.5T and 3.0T for any abdominal organ. In comparison, Dale et al observed a significant increase in liver ADC at 3.0T, with no significant difference in ADC between 1.5T and 3.0T for the pancreas and spleen. These authors attribute a lower liver ADC at 1.5T to “noise floor” issues from lower SNR at this field strength. However, the liver has the lowest T2 relaxation time of the abdominal organs assessed in this study (35–37), and in an earlier intra-individual assessment of relaxation times at 1.5T and 3.0T, the liver showed the largest relative decrease in T2 relaxation times at 3.0T among the upper abdominal organs (35). It is possible that these T2 properties of the liver offset the potential increase in SNR at 3.0T, thereby mitigating the impact of a possible “noise floor” at 1.5T upon differences in liver ADC between 1.5T and 3.0T. Recent data available in abstract format from a sample of 30 volunteers showed significantly lower hepatic ADC at 3.0T compared with 1.5T (using b0-600), which also is different from the results of Dale et al (33).

Despite the potential advantages of imaging at a higher field strength (22), our results do not show a clear benefit of performing abdominal DWI at 3.0T. Subjective image quality, which in this study was determined on the basis of image ghosting and sharpness of the liver edge, was significantly worse at 3.0T for all evaluated DWI sequences. Hunsche et al and Kuhl et al (23, 39) also observed worse image distortion at 3.0T in intra-individual assessments of DWI in the brain at 1.5T and 3.0T. The increased distortion of SS EPI at higher field strength relates to the vulnerability of the underlying EPI technique to phase-related errors. Increased field inhomogeneity and susceptibility effects at 3.0T can lead to even more severe phase-related errors and subsequent worse geometric distortion of DWI. While the use of a radial multi-shot turbo spin-echo technique has been explored as a means to reduce image distortion for abdominal DWI (40), this approach is generally associated with a substantial increase in imaging time.

The reproducibility of ADC measurements at constant field strength, a critical factor to consider when performing serial ADC assessments, was equivalent between the two field strengths. The range of CV of ADC that we observed at 1.5T and 3.0T for the liver, pancreas, and spleen was generally similar to the results of Braithwaite et al (31), who observed a mean CV of ADC of 14% in the short- and midterm for these same three organs when performing abdominal DWI in healthy volunteers at 3.0T, although larger than the mean CV of ADC of 3.2% in the liver at 1.5T reported by Patel et al (4). We believe that CV of ADC in this range is adequate for longitudinal abdominal ADC measurements and supports the role of the ADC value as a reliable biomarker of disease when DWI is performed at constant field strength.

We note several differences in methodology between our study and that of Dale et al (33) who also recently assessed for differences in ADC values of the abdominal organs between 1.5T and 3.0T. First, we compared the relative performance of DWI at each individual field strength in terms of subjective image quality and reproducibility of ADC; these features were not assessed in the study by Dale et al. Also, our study included assessments of both breathhold and free-breathing DWI, whereas Dale et al evaluated only free-breathing DWI. Finally, Dale et al did not include the kidneys in their study, which we did evaluate.

Several limitations of this study warrant mention. First, our sample size was small, reflecting our initial experience. Second, we only assessed healthy volunteers, such that differences between 1.5T and 3.0T in the utility of ADC measurements for abdominal pathology were not assessed. Third, only free-breathing and breathhold DWI were assessed. While it would have been useful to also compare respiratory-triggered DWI between field strengths, a respiratory-triggered DWI sequence was not available on our institution's 3.0T system at the time of this study. Given the potential impact of cardiac motion on ADC values in the upper abdomen (41), cardiac triggering has also been recently explored as a means of improving ADC reproducibility (42), which was not examined in this study. Fourth, as image quality was rated by a single observer, inter-observer variability for image quality was not assessed. Fifth, different techniques were used for achieving fat-saturation at 1.5T and 3.0T, which may have influenced image quality. Sixth, our results may not apply to other 1.5T and 3.0T MRI systems. Finally, while the benefit of parallel imaging for improving image quality of DWI has been clearly demonstrated (43–46), parallel imaging greatly complicates the assessment of image SNR. For instance, Dietrich et al demonstrated that calculations of SNR based upon a single acquisition using an ROI placed in the background air to measure image noise do not agree with true SNR in the presence of parallel imaging (47). Therefore, SNR of SS-EPI DWI at 1.5T and 3.0T was not assessed using this approach.

In conclusion, the reproducibility of ADC measurements obtained solely at either 1.5T or 3.0T was sufficient to allow for serial measurements at constant field strength. We observed no significant difference in ADC in the liver, kidney, pancreas, or spleen between field strengths (except for lower liver ADC for breathhold b0, 500, 600 at 3.0T), with similar reproducibility of ADC at each of the two field strengths. Image quality was consistently worse at 3.0T than at 1.5T. Continued optimization of abdominal DWI at 3.0T is, therefore, warranted.

Ancillary