Journal Logo

Research Article: Observational Study

Reproducibility of Dynamic Contrast-Enhanced MRI in Renal Cell Carcinoma

A Prospective Analysis on Intra- and Interobserver and Scan–Rescan Performance of Pharmacokinetic Parameters

Wang, Haiyi MD; Su, Zihua PhD; Ye, Huiyi MD; Xu, Xiao PhD; Sun, Zhipeng MD; Li, Lu MD; Duan, Feixue MD; Song, Yuanyuan MD; Lambrou, Tryphon PhD; Ma, Lin MD

Section Editor(s): Roviello., Giandomenico

Author Information
doi: 10.1097/MD.0000000000001529

Abstract

INTRODUCTION

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has evolved from a tool for qualitative analysis, subjectively judging the enhancement of a target area on a visual basis and semiquantitative analysis/characterization of tumors using curvology,1,2 to a fully quantitative tool for analyzing/evaluating parameters generated using pharmacokinetic models.3,4 Through kinetic modeling of signal intensity changes resulting from the passage of contrast agent through the tumor vascular bed, DCE-MRI can produce physiologically related parameters, including the volume transfer constant from plasma to the extracellular extravascular space (EES) (Ktrans), the efflux rate constant from EES back to plasma (Kep), the ratio of the EES volume to tissue volume (Ve), and the ratio of blood plasma volume to tissue volume (Vp), which together may reflect tumor perfusion, vascular volume, and angiogenesis.

To date, pharmacokinetic parameters derived from DCE-MRI have been used extensively to evaluate tumor vessels, mainly reflecting 2 aspects: first, the characterization of benign versus malignant tumors5 and the grading of malignant tumors,6 and second, the evaluation of treatment effects, such as the therapeutic effects of molecularly targeted drugs7 and chemotherapy.8

In renal tumors, DCE-MRI has played a key role in the delineation and characterization of masses, such as differentiating among RCC subtypes9 and evaluating the therapeutic effects of molecularly targeted therapies.10,11 Especially for the latter, pharmacokinetic parameters derived from quantitative DCE-MRI seem promising and may become surrogate biomarkers, which may be more convenient and accurate for monitoring dynamic changes in RCCs after receiving molecularly targeted drugs, potentially avoiding the need for repeated biopsies.

Because DCE-MRI showed potential in these areas, its reproducibility in quantitative analyses has drawn a lot of attention. To date, the methods and focus of DCE-MRI reproducibility studies have varied from assessing intra- and interobserver reproducibility to understanding scan–rescan reproducibility. However, for RCCs, the reproducibility of DCE-MRI in terms of pharmacokinetic analyses has not yet been thoroughly reported, in terms of observer or scan–rescan reproducibility.

In addition, conventional pharmacokinetic DCE-MRI has been performed using continuous scanning with high temporal resolution for pharmacokinetic analyses of target lesions. However, this sacrifices spatial resolution, and subsequently lowers the capacity of DCE-MRI sequence to cover a large scanning volume of interest. Furthermore, this method, in clinical practice, cannot meet the demands of high-quality DCE-MR images of abdominal organs, such as the liver and kidney, because of respiratory motion. Orton et al12 reported the feasibility of obtaining functional liver perfusion estimates using a sequential breath-hold protocol. Thus, in the present study, in an attempt to balance the requirements of both clinical practice and pharmacokinetic analyses, we performed renal DCE-MRI as subjects held their breath, resulting in a noncontinuous scan mode. The aim of this study is to assess the intra- and interobserver as well as scan–rescan reproducibility of quantitative parameters of RCCs with noncontinuous DCE-MRI imaging.

MATERIALS AND METHODS

This prospective and observational study was approved by the local Institutional Review Board (Ethics Committee of Chinese PLA General Hospital). Written informed consents were obtained from all patients.

Patients

Patients with suspicious RCC during the computed tomography and ultrasound examinations were recruited at diagnosis from the urological clinic at our hospital between September 2012 and November 2012. To be included in this study, patients had to meet the following inclusion criteria: age ≥18 years old, glomerular filtration rate >60 mL/min, and size of lesions >1.0 cm in diameter to avoid partial volume artifact concerns, and clear cell RCCs—most common pathologic subtype. Exclusion criteria included common exclusion criteria for MRI scans and the use of Ga-related contrast, lesions with complete necrosis or cystic degeneration confirmed in MR examination and patients with poor DCE-MRI quality. Poor imaging quality should mainly meet the criteria—severe motion artifacts appeared in enhanced MRI and the images cannot be used for further evaluation.

Sample size estimation: sample size estimation for intraclass correlation coefficient (ICC) was performed using Power Analysis & Sample Size Software, PASS 11.0 (NCSS, LLC. Kaysville, UT). The preset condition included observers (n = 3), R1 (expected ICC = 0.9), R0 (acceptable lowest ICC = 0.75), α = 0.05, and β = 0.20. At last the smallest sample size (k) was 19.

Procedure

MRI Technique

Patients underwent DCE-MRI twice: the first scan was within 48 hours of the initial diagnosis and the second scan was 48 to 72 hours after the first scan. MRI examinations were performed on a 3.0 T scanner (GE Discovery MR 750, GE Healthcare, Milwaukee, WI; maximum gradient strength 50 mT, maximum slew rate 200 mT/s) with an 8-channel surface phased-array coil. Patients practiced breathing techniques before each scan, which included breathing quickly during a nonscanning break and then breath-holding in the same position for as long as possible. Care was taken to ensure that, for each patient, rescanning was performed in the same lying position and the same anatomical location as the first scan. Routine clinical axial and coronal T2-weighted imaging was performed in all patients before dynamic studies to help localize and delineate tumors. The imaging protocol for DCE-MRI consisted of a precontrast T1 mapping sequence and a DCE sequence. The former included 5 consecutive axial 3-dimensional (3D) spoiled-gradient recalled-echo sequences for liver acquisition with volume acceleration (LAVA) with an array of flip angles (3°, 6°, 9°, 12°, and 15°) in breath-hold mode. Then, an axial DCE sequence (flip angle, 12°)—repeated scanning during 12 seconds of breath-holding for 2 phases and subsequently 6 seconds of breathing—was performed for 4.4 minutes to monitor contrast passage (Figure 1). Scanning parameters were as follows: repetition time 2.8 milliseconds, echo time 1.3 milliseconds, matrix 288 × 180, FOV 38 × 38 cm, slice thickness 6 mm, number of excitations 1, bandwidth 125 kHz, and parallel imaging acceleration factor 3. The contrast agent—Gadodiamide (0.1 mmol/kg, Omniscan, GE Healthcare)—was given intravenously when the second scan was started at a rate of 2 mL/s using a power injector (Spectris; MedRad, Warrendale, PA). The contrast bolus was flushed with 20 mL normal saline, administered at the same rate, to improve bolus coherence.

FIGURE 1
FIGURE 1:
Dynamic contrast-enhanced MRI protocol. MRI scan was performed in a noncontinuous way—repeated scanning during 12 seconds breath-holding for 2 phases and subsequent 6 seconds breathing. The intravenous injection of MR contrast agent was started simultaneously when the second scanning session was initiated. MRI = magnetic resonance imaging.

Image Postprocessing and Analysis

All DCE-MRI analyses were conducted using open-source software packages, including the R package (http://dcemri.sourceforge.net/) and a medical image nonrigid registration package (http://cmictig.cs.ucl.ac.uk/wiki/index.php/NiftyReg).

All images were transferred to an Omni-Kinetics workstation (GE Healthcare, LifeScience, China) for analysis. Each patient's breath-holding position differed slightly and a soft organ as kidney changed its shape nonrigidly. Many articles have proposed using image registration methods13 to handle body motion within time domain.11,14 Here, the workstation provided an automatic nonlinear registration framework15 to help remove any error of misalignment between consecutive MRI scans, thus making our results more accurate (for more visual results, please refer to Supplemental Digital Content, http://links.lww.com/MD/A431). The registration framework used a free-form deformation algorithm16 as the main registration engine and mutual information as the correspondence metric.17

Data Collection

Calculation of Pharmacokinetic Parameters

The widely used 2-compartment extended-Tofts model18 (Equation 1) with population-based arterial input function (AIF)19 (Equation 2) was used to calculate parameters:

where in Equation 1, Ktrans is the transfer constant from plasma to the EES; Ve is the ratio of the EES volume to tissue volume; Vp is the ratio of blood plasma volume to tissue volume; Kep = Ktrans/Ve is the efflux rate constant from EES to plasma; and Cp(t) and Ce(t) are the contrast agent concentrations in the plasma and EES, respectively, and where in Equation 2, D = 1.0 mmol/kg, a1 = 2.4 kg/L, a2 = 0.62 kg/L, m1 = 3.0 and m2 = 0.016.20

Region of Interest Selection

Images were transferred to a Sun workstation (Sparc 10, Sun Microsystems, Mountain View, CA), where pharmacokinetic parameters were measured using the ImageJ software (National Institutes of Health, Bethesda, MD). Using reference information from anatomic axial and coronal T2-weighted images and postcontrast T1 images, 3 radiologists (ZS, FD, YS, all board-certified radiologists engaged in abdominal imaging for 8, 10, and 9 years, respectively) were instructed to place region of interests (ROIs) on the slice with the largest diameter of tumors on the dynamic images of DCE-MRI, covering the whole tumor where possible but excluding pulsatile artifacts from blood vessels and susceptibility artifacts from the adjacent bowels. Then, the same ROI was copied to all 4 parametric maps (Ktrans, Kep, Ve, and Vp) (Figure 2).

FIGURE 2
FIGURE 2:
Graph showed placement of ROI on parametric maps of K trans (A), and V e (B) for renal cell carcinoma. K trans = transfer constant from plasma to the extracellular extravascular space (EES), ROI = region of interest, V e = the ratio of the EES volume to tissue volume.

Because values of Ktrans >1.2 min−1 are commonly considered to represent errors because of measurement of pseudo-permeability in large blood vessels,21,22 any pixels with Ktrans >1.2 min−1 or with Ve beyond the range of 0% to 100% were excluded from the parametric maps. Thus, we used the histogram function and set minimum and maximum values of Ktrans (0, 1.2 min−1), Kep (0, 15 min−1), Ve (0, 1), and Vp (0, 1); the mean values of Ktrans, Kep, Ve, and Vp were automatically calculated within the preset range.

Three observers (No. 1, ZS; No. 2, FD; and No. 3, YS) measured parameters of 2 DCE-MRI scans once, respectively, to examine interobserver and scan–rescan reproducibility. Then, the first observer (ZS) measured first scan again to evaluate intraobserver reproducibility.

Statistical Analyses

Intra-, Interobserver, and Scan–Rescan Differences in Pharmacokinetic Parameters

Intraobserver differences was assessed using paired t tests. Interobserver and scan–rescan differences were evaluated using two-way analysis of variance.

Variability Analyses

Intra-, interobserver, and scan–rescan variability of pharmacokinetic parameters were evaluated using the test–retest root-mean-square coefficient of variation method23 to estimate the within-subject coefficient of variation (wCoV) of the whole study group for a pairwise comparison of all possible observer combinations. The wCoV (expressed as a percentage) is used to estimate variability by quantifying the variation in Ktrans, Kep, Ve, and Vp for a single individual from the study group for any pairwise comparison. Variability was considered acceptable when wCoV was within the goal of current quantitative imaging initiatives (Quantitative Imaging Biomarkers Alliance; coefficient of variation <20%).24

Agreement Analyses

Intra-, interobserver, and scan–rescan agreements of pharmacokinetic parameters were evaluated using the ICC with a two-way mixed effect model. The agreement was defined as good (ICC > 0.75), moderate (ICC = 0.5–0.75), or poor (ICC < 0.5).

All statistical analyses were performed with the SAS software (ver. 9.3; SAS Institute Inc, Cary, NC) and GraphPad Prism (ver. 6.0; GraphPad Software, Inc, La Jolla, CA). P values <0.05 were considered to indicate a statistically significant difference.

RESULTS

Patients and Lesions Characteristics

A total of 28 patients with renal lesions underwent DCE-MRI scanning according to the given MR protocol. After the analysis of imaging quality and histopathology, 7 cases were excluded because of poor imaging quality (n = 2), papillary RCC (n = 1), chromophobic RCC (n = 3), and a benign renal angiomyolipoma lesion (n = 1). This resulted in 21 effective paired data sets of clear cell RCC cases (17 men, 4 woman; age 37–69 years, mean age 54.6 years, mean size, 5.0 ± 2.2 cm).

Pharmacokinetic Parameters of RCC

After setting the upper and lower limits for each kinetic value, the mean values of each pharmacokinetic parameter of each ROI were automatically calculated. They are listed in Tables 1 and 2.

TABLE 1
TABLE 1:
Kinetic Parameters and Comparison Between 2 Measurements for First MR Scan
TABLE 2
TABLE 2:
Kinetic Parameters Measured by 3 Observers for First and Second MR Scan

Analysis of Differences in Kinetic Parameters

There were no statistically significant intra- or interobserver differences in any kinetic parameters examined (Table 1), or between MRI scans (Table 3; all P > 0.05). The scan–rescan values measured by observer 1 for all of the parameters are presented in Figure 3.

TABLE 3
TABLE 3:
Interobserver and Scan–Rescan Comparison of Kinetic Parameters in Light of Two-Way ANOVA
FIGURE 3
FIGURE 3:
Graphs showed comparison of DCE-MRI kinetic parameters of 2 MRI scans of RCC for K trans (A), K ep (B), V e (C), and V p (D). DCE-MRI = dynamic contrast-enhanced magnetic resonance imaging, K trans = transfer constant from plasma to the extracellular extravascular space (EES), K ep = efflux rate constant from EES to plasma, MRI = magnetic resonance imaging, RCC = renal cell carcinomas, V e = the ratio of the EES volume to tissue volume, V p = the ratio of blood plasma volume to tissue volume.

Variability Analysis

For intraobserver variability, all parameters except Vp demonstrated low variation (Figure 4), where Ktrans and Ve showed similar wCoVs (1.2% and 0.9%), lower than Kep (3.7%). Variation in Vp was extreme (58%). For interobserver variability (Figure 5), Ktrans, Kep, and Ve showed similarly low wCoVs (2.5%, 3.1%, and 2.9%, respectively); Vp also demonstrated large variation (38%). In addition, each combination of comparison showed similar results to overall variability. Furthermore, scan–rescan variability (Figure 6) was greater than intra- and interobserver variability. Ktrans and Kep showed similar wCoVs (15.6%, 15.4%), and variation in Kep was lowest (10.1%) and variation in Vp was highest (48%). In terms of both intraobserver and scan–rescan variability, Ve showed the lowest variation.

FIGURE 4
FIGURE 4:
Graph showed intraobserver variability. Whereas K trans, K ep, and V e showed similarly low variability, V p showed highest variability (range of 41%–98 %). K trans = transfer constant from plasma to the extracellular extravascular space (EES), K ep = efflux rate constant from EES to plasma, V e = the ratio of the EES volume to tissue volume, V p = the ratio of blood plasma volume to tissue volume, wCoV = within-subject coefficient of variation.
FIGURE 5
FIGURE 5:
Graph showed interobserver variability for all possible observer combinations. K trans = transfer constant from plasma to the extracellular extravascular space (EES), K ep = efflux rate constant from EES to plasma, V e = the ratio of the EES volume to tissue volume, V p = the ratio of blood plasma volume to tissue volume, wCoV = within-subject coefficient of variation.
FIGURE 6
FIGURE 6:
Graph showed scan–rescan variability. V e showed lowest variation, whereas V p showed highest variation. K trans = transfer constant from plasma to the extracellular extravascular space (EES), K ep = efflux rate constant from EES to plasma, V e = the ratio of the EES volume to tissue volume, V p = the ratio of blood plasma volume to tissue volume, wCoV = within-subject coefficient of variation.

Agreement Analysis

The ICCs of all kinetic parameters in terms of both intra- and interobserver tests were all >0.75 and ICCs of Ktrans, Kep, and Ve were also >0.90, which are excellent agreements (range, 0.991–0.999; P < 0.001), with Ktrans showing the highest intra- and interobserver ICC (0.999 and 0.993, respectively) (Table 4). However, in scan–rescan agreement analyses, the ICC of Kep showed the highest value (0.906, 95% CI: 0.710–0.972, P < 0.001) followed by that of Ve (0.764 95% CI: 0.378–0.925, P = 0.001). Ktrans and Vp demonstrated a moderate agreement with an ICC of 0.686 (95% CI: 0.212–0.898) and 0.657 (95% CI: 0.164–0.888), respectively (Table 4).

TABLE 4
TABLE 4:
Intra-, Interobserver, and Scan–Rescan Agreement Analysis

DISCUSSION

The reproducibility of pharmacokinetic parameters derived from DCE-MRI has been a controversial topic. Most studies have relied on scan comparisons obtained at 2 separate time points in the same patients without any intervening therapy.22,23,25–29 Other studies simply assessed the within-subject variation in an identical DCE data set, making measurements at multiple times within and between observers.30–33 A particular strength of our study is that it covers both aspects, and thus is a more comprehensive analysis of the reproducibility of pharmacokinetic parameters produced by DCE-MRI.

Previous studies have used different parameters for reproducibility analyses, depending on each study's aims. The most commonly used parameters have been Ktrans, permeability surface (PS) product, and Ve, followed by Kep; Vp is seldom used. Kep is always larger than Ktrans in value. In a range of studies on tumors, Kep has been 2 to 5 times higher than Ktrans, and Ve has ranged from 20% to 50%. In the present study, Kep was 0.760 to 0.839 min−1, nearly double the Ktrans value of 0.450 to 0.466 min−1, whereas Ve was 0.551 to 0.560, consistent with previous studies.31,34 Our Ktrans values were similar to those of a previous study35 on pharmacokinetic parameters of RCC (0.51 ± 0.34 min–1). However, our Vp value (1.900–2.498 ×10–6) was much smaller than in previous studies, which might have been influenced by the noncontinuous scanning mode that we used.

In the present study, Ktrans, Kep, and Ve showed low variation in both intraobserver (wCoV: 2.1%, 3.7%, and 0.9%, respectively) and interobserver (wCoV: 2.5%, 3.1%, and 2.9%, respectively) variability analyses. This is consistent with the results of some previous studies. Davenport et al31 found that Ktrans, Kep, and Ve had similarly low variation (wCoV: 5.7%, 5.0%, and 3.9%, respectively) in assessments of uterine fibroids, and Beresford et al30 reported 4.5%, 2.3%, and 1% wCoV in expert intraobserver analyses and 6.0%, 5.2%, and 2.7% wCoV in expert interobserver analyses, respectively, in assessments of breast cancer.

In the present study, intra- and interobserver agreements were excellent (all values >0.75). Similar results were reported by Davenport et al31 (interobserver agreement: 0.88, 0.98, and 0.87 ICCs for Ktrans, Kep, and Ve, respectively) and Braunagel et al's32 research on RCCs (ICC ranging from 0.79 to 0.97 for PS, Ve, and Vp in both intra- and interobserver agreement). However, unlike our results, Braunagel et al reported that PS was less reproducible than Vp, whereas Ve showed the lowest reproducibility (ICC = 0.6 and 0.64 in intra- and interobserver agreements), probably because of its insufficient scanning time (≤240 seconds). A previous study36 showed that permeability parameters (Ktrans or PS and Ve) are more dependent on total acquisition time than perfusion parameters (Kep, Vp) and are sensitive to small differences in data with increasing variability.

Furthermore, regarding scan–rescan reproducibility, we found that Ktrans, Kep, and Ve had similar test–retest wCoV (15.6%, 15.4%, and 10.1%, respectively). Similar wCoV values for Ktrans and Ve have been reported for abdominal tumors (19% and 14%) 23, and carotid atherosclerotic plaques (18% and 12%),29 but higher than that of gliomas (7.7% and 6.2%),22 which was probably because of inevitable motion of targeted abdominal organ, though imaging registration was used. In addition, the Ve values in the previous studies have been the most reproducible in test–retest analyses. Even in studies in which not all parameters have been acceptable, Ve has shown the lowest variation.26,28

However, in our study, the Vp value showed the largest variation among all parameters and across all analyses, probably because of the slow sampling method; nonetheless, this result is similar to or greater than the reported values regarding test–retest evaluations of carotid atherosclerotic plaque (40%)29 and brain and abdominal tumors (39% and 30%, respectively).23 For interscan agreement analysis, Kep showed highest ICC (0.906). And Ve has higher ICC (0.764) than Ktrans (0.686), which was in accordance with the previous studies in gliomas 22 and uterine fibroids.24 Combined abovementioned variability with agreement analysis, Ktrans, Kep, and Ve were reproducible parameters in this noncontinuous DCE-MRI reproducibility studies.

It should be noted that in our study scan–rescan reproducibility was lower than intra- and interobserver performance. Given proper training and guidance, interpersonal difference can be trivial. However, although image registration method was used, there is still accountable error introduced into dynamic images. Image registration can improve the accuracy of results but cannot totally fix the problem. Patients’ hemodynamic information changes such as heart rate could also affect scan–rescan reproducibility.

In order to ensure the precision of DCE-MRI regarding the course of MRI scans, we made a great deal of efforts. Before scanning, we gave our technologists MRI scan training to ensure consistency in regards to parameters, the location plane between scans, and rules for measurements of pharmacokinetic parameters. We also gave patients respiratory training and advice on how to avoid discomfort. During scans, we used a series of 3D LAVA sequences, a type of spoiled-gradient echo-pulse sequence, which can markedly reduce the scanning time, especially for a large volume of interest (eg, whole kidneys), compared with 2D sequences, and still maintain a high signal-to-noise ratio. In addition, scanning as patients hold their breath can help alleviate the effects of respiratory motion. We set the breathing interval to 6 seconds, as short as possible. The noncontinuous scanning mode can lead to a low sampling rate; however, down sampling only slightly affects the fit quality at lower sampling resolution (6 and 24 seconds).37

After MRI scanning, data processing was initiated. Before extracting pharmacokinetic parameters, we tested a published 3D nonrigid image registration method to correct misalignment caused by respiratory motion, which worked well and probably contributed to the near-ideal reproducibility values. During parameter extraction, the most sensitive method to a dynamic scan's temporal resolution is AIF. In noncontinuous scanning mode, it is almost impossible to have an identical AIF when performing scans twice in the same patient. This led us to use a population-based AIF method, rather than a personal AIF. This not only helped address temporal resolution difficulties, but also reduced AIF ROI location and sizing errors that have been reported previously.38 The population-based AIF works equally well as the individual AIF for estimating pharmacokinetic parameters, as confirmed by several investigators.39–41 Other reasonable alternatives for measuring DCE-MRI kinetics include a reference region model,42 a double-reference model,43 and a multiple-references model.44 The advantage of these reference methods is that there is no requirement for direct AIF measurement. However, these models are not as widely used as the Tofts model, and are beyond the scope of this article.

Rules for ROI analysis using parametric maps have to be consistent and clearly defined. We chose ROIs covering the whole tumor at the slice with the maximum diameter instead of one or multiple smaller ROIs, based on a previous study32 that concluded that the best interobserver and intraobserver correlations are obtained when a whole-tumor ROI is defined. In addition, in a previous study, the interobserver variability of Ktrans was significantly lower in whole-lesion ROI analysis (10.6%) than in 3 out of 4 ROI analyses (20.1%) of uterine fibroids.33

This study has some limitations. First, the relatively low temporal resolution and short acquisition time might have led to inaccurate Vp values, in turn leading to large variation in reproducibility analyses. Second, we simply investigated 1 slice of tumors, which may not represent whole tumors. A 3D whole-tumor volume analysis with more pixels may reduce the variation in mean values of quantitative parameters and improve interobserver reproducibility compared with using slices, although a previous study45 on primary RCC reported no significant differences in functional parameters in this respect. Thus, single-slice analysis may save time and workload. Third, because of necessity of image registration and establishment of kinetic parametric maps, the whole analysis process was relatively time-consuming, this is not ideal in real clinical practice, further more easily handled software or accelerating method should be investigated. Fourth, only clear cell RCC was enrolled in this study because clear cell RCC is the most common subtype and other subtypes were relatively less; it should be prudent that the result of this study is applied into other RCC subtypes because of biological difference of different tumors.

CONCLUSION

Our results indicate that DCE-MRI is a promising and reliable tool for pharmacokinetic analysis of RCC with good intra- and interobserver reproducibility and moderate to good scan–rescan performance using Ktrans, Kep, and Ve as parameters under a noncontinuous scanning mode. Further studies with a standardized scanning protocol are needed to ensure that results from various centers can be communicated conveniently and reliably.

Acknowledgments

The authors would like to express their gratitude for the technical support and assistance from Zhenyu Zhou, PhD, and Dandan Zheng, PhD, of MR Research GE Healthcare China.

REFERENCES

1. El Khouli RH, Macura KJ, Jacobs MA, et al. Dynamic contrast-enhanced MRI of the breast: quantitative method for kinetic curve type assessment. AJR Am J Roentgenol 2009; 193:W295–300.
2. Engelbrecht MR, Huisman HJ, Laheij RJ, et al. Discrimination of prostate cancer from normal peripheral zone and central gland tissue by using dynamic contrast-enhanced MR imaging. Radiology 2003; 229:248–254.
3. Jackson A, O’Connor JP, Parker GJ, et al. Imaging tumor vascular heterogeneity and angiogenesis using dynamic contrast-enhanced magnetic resonance imaging. Clin Cancer Res 2007; 13:3449–3459.
4. Oostendorp M, Post MJ, Backes WH. Vessel growth and function: depiction with contrast-enhanced MR imaging. Radiology 2009; 251:317–335.
5. Bali MA, Metens T, Denolin V, et al. Tumoral and nontumoral pancreas: correlation between quantitative dynamic contrast-enhanced MR imaging and histopathologic parameters. Radiology 2011; 261:456–466.
6. Nguyen TB, Cron GO, Mercier JF, et al. Diagnostic accuracy of dynamic contrast-enhanced MR imaging using a phase-derived vascular input function in the preoperative grading of gliomas. AJNR Am J Neuroradiol 2012; 33:1539–1545.
7. Hsu CY, Shen YC, Yu CW, et al. Dynamic contrast-enhanced magnetic resonance imaging biomarkers predict survival and response in hepatocellular carcinoma patients treated with sorafenib and metronomic tegafur/uracil. J Hepatol 2011; 55:858–865.
8. Akisik MF, Sandrasegaran K, Bu G, et al. Pancreatic cancer: utility of dynamic contrast-enhanced MR imaging in assessment of antiangiogenic therapy. Radiology 2010; 256:441–449.
9. Sun MRM, Ngo L, Genega EM, et al. Renal cell carcinoma: dynamic contrast-enhanced MR imaging for differentiation of tumor subtypes—correlation with pathologic findings. Radiology 2009; 250:793–802.
10. Hahn OM, Yang C, Medved M, et al. Dynamic contrast-enhanced magnetic resonance imaging pharmacodynamic biomarker study of sorafenib in metastatic renal carcinoma. J Clin Oncol 2008; 26:4572–4578.
11. Rosen MA, Schnall MD. Dynamic contrast-enhanced magnetic resonance imaging for assessing tumor vascularity and vascular effects of targeted therapies in renal cell carcinoma. Clin Cancer Res 2007; 13 (2 pt 2):770s–776s.
12. Orton MR, Miyazaki K, Koh DM, et al. Optimizing functional parameter accuracy for breath-hold DCE-MRI of liver tumours. Phys Med Biol 2009; 54:2197–2215.
13. Ruthotto L, Hodneland E, Modersitzki J. Dawant B, Christensen G, Fitzpatrick JM, Rueckert D. Registration of dynamic contrast enhanced MRI with local rigidity constraint. Biomedical Image Registration. Vol. 7359. Heidelberg: Springer Berlin; 2012. 190–198.
14. Zollner FG, Sance R, Rogelj P, et al. Assessment of 3D DCE-MRI of the kidneys using non-rigid image registration and segmentation of voxel time courses. Comput Med Imaging Graph 2009; 33:171–181.
15. Klein A, Andersson J, Ardekani BA, et al. Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration. Neuroimage 2009; 46:786–802.
16. Rueckert D, Sonoda LI, Hayes C, et al. Nonrigid registration using free-form deformations: application to breast MR images. IEEE Trans Med Imaging 1999; 18:712–721.
17. Pluim JP, Maintz JB, Viergever MA. Mutual-information-based registration of medical images: a survey. IEEE Trans Med Imaging 2003; 22:986–1004.
18. Tofts PS, Kermode AG. Measurement of the blood-brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts. Magn Reson Med 1991; 17:357–367.
19. Calamante F. Arterial input function in perfusion MRI: a comprehensive review. Prog Nucl Magn Reson Spectrosc 2013; 74:1–32.
20. Fritz-Hansen T, Rostrup E, Larsson HB, et al. Measurement of the arterial concentration of Gd-DTPA using MRI: a step toward quantitative perfusion imaging. Magn Reson Med 1996; 36:225–231.
21. Parker GJ, Suckling J, Tanner SF, et al. Probing tumor microvascularity by measurement, analysis and display of contrast agent uptake kinetics. J Magn Reson Imaging 1997; 7:564–574.
22. Jackson A, Jayson GC, Li KL, et al. Reproducibility of quantitative dynamic contrast-enhanced MRI in newly presenting glioma. Br J Radiol 2003; 76:153–162.
23. Roberts C, Issa B, Stone A, et al. Comparative study into the robustness of compartmental modeling and model-free analysis in DCE-MRI studies. J Mag Reson Imaging 2006; 23:554–563.
24. Heye T, Davenport MS, Horvath JJ, et al. Reproducibility of dynamic contrast-enhanced MR imaging. Part I. Perfusion characteristics in the female pelvis by using multiple computer-aided diagnosis perfusion analysis solutions. Radiology 2013; 266:801–811.
25. Padhani AR, Hayes C, Landau S, et al. Reproducibility of quantitative dynamic MRI of normal human tissues. NMR Biomed 2002; 15:143–153.
26. Galbraith SM, Lodge MA, Taylor NJ, et al. Reproducibility of dynamic contrast-enhanced MRI in human muscle and tumours: comparison of quantitative and semi-quantitative analysis. NMR Biomed 2002; 15:132–142.
27. Ferl GZ, Xu L, Friesenhahn M, et al. An automated method for nonparametric kinetic analysis of clinical DCE-MRI data: application to glioblastoma treated with bevacizumab. Magn Reson Med 2010; 63:1366–1375.
28. Lankester KJ, Taylor JN, Stirling JJ, et al. Dynamic MRI for imaging tumor microvasculature: comparison of susceptibility and relaxivity techniques in pelvic tumors. J Magn Reson Imaging 2007; 25:796–805.
29. Gaens ME, Backes WH, Rozel S, et al. Dynamic contrast-enhanced MR imaging of carotid atherosclerotic plaque: model selection, reproducibility, and validation. Radiology 2013; 266:271–279.
30. Beresford MJ, Padhani AR, Taylor NJ, et al. Inter- and intraobserver variability in the evaluation of dynamic breast cancer MRI. J Magn Reson Imaging 2006; 24:1316–1325.
31. Davenport MS, Heye T, Dale BM, et al. Inter- and intra-rater reproducibility of quantitative dynamic contrast enhanced MRI using TWIST perfusion data in a uterine fibroid model. J Magn Reson Imaging 2013; 38:329–335.
32. Braunagel M, Radler E, Ingrisch M, et al. Dynamic contrast-enhanced magnetic resonance imaging measurements in renal cell carcinoma: effect of region of interest size and positioning on interobserver and intraobserver variability. Invest Radiol 2015; 50:57–66.
33. Heye T, Merkle EM, Reiner CS, et al. Reproducibility of dynamic contrast-enhanced MR imaging. Part II. Comparison of intra- and interobserver variability with manual region of interest placement versus semiautomatic lesion segmentation and histogram analysis. Radiology 2013; 266:812–821.
34. Tofts PS, Brix G, Buckley DL, et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999; 10:223–232.
35. Chandarana H, Amarosa A, Huang WC, et al. High temporal resolution 3D gadolinium-enhanced dynamic MR imaging of renal tumors with pharmacokinetic modeling: preliminary observations. J Magn Reson Imaging 2013; 38:802–808.
36. Michaely HJ, Sourbron SP, Buettner C, et al. Temporal constraints in renal perfusion imaging with a 2-compartment model. Invest Radiol 2008; 43:120–128.
37. Lavini C. Simulating the effect of input errors on the accuracy of Tofts’ pharmacokinetic model parameters. Magn Reson Imaging 2015; 33:222–235.
38. Cutajar M, Mendichovszky IA, Tofts PS, et al. The importance of AIF ROI selection in DCE-MRI renography: reproducibility and variability of renal perfusion and filtration. Eur J Radiol 2010; 74:e154–e160.
39. Wang Y, Huang W, Panicek DM, et al. Feasibility of using limited-population-based arterial input function for pharmacokinetic modeling of osteosarcoma dynamic contrast-enhanced MRI data. Magn Reson Med 2008; 59:1183–1189.
40. Parker GJ, Roberts C, Macdonald A, et al. Experimentally-derived functional form for a population-averaged high-temporal-resolution arterial input function for dynamic contrast-enhanced MRI. Magn Reson Med 2006; 56:993–1000.
41. Li X, Welch EB, Arlinghaus LR, et al. A novel AIF tracking method and comparison of DCE-MRI parameters using individual and population-based AIFs in human breast cancer. Phys Med Biol 2011; 56:5753–5769.
42. Lee J, Platt S, Kent M, et al. An analysis of the pharmacokinetic parameter ratios in DCE-MRI using the reference region model. Magn Reson Imaging 2012; 30:26–35.
43. Yang C, Karczmar GS, Medved M, et al. Estimating the arterial input function using two reference tissues in dynamic contrast-enhanced MRI studies: fundamental concepts and simulations. Magn Reson Med 2004; 52:1110–1117.
44. Yang C, Karczmar GS, Medved M, et al. Reproducibility assessment of a multiple reference tissue method for quantitative DCE-MRI analysis. Magn Reson Med 2009; 61:851–859.
45. Winter KS, Helck AD, Ingrisch M, et al. Dynamic contrast-enhanced magnetic resonance imaging assessment of kidney function and renal masses: single slice versus whole organ/tumor. Invest Radiol 2014; 49:720–727.

Supplemental Digital Content

Copyright © 2015 The Authors. Published by Wolters Kluwer Health, Inc. Health, Inc. All rights reserved.