Reliability of dynamic contrast-enhanced magnetic resonance imaging data in primary brain tumours: a comparison of Tofts and shutter speed models
The purpose of this study is to investigate the robustness of pharmacokinetic modelling of DCE-MRI brain tumour data and to ascertain reliable perfusion parameters through a model selection process and a stability test.
DCE-MRI data of 14 patients with primary brain tumours were analysed using the Tofts model (TM), the extended Tofts model (ETM), the shutter speed model (SSM) and the extended shutter speed model (ESSM). A no-effect model (NEM) was implemented to assess overfitting of data by the other models. For each lesion, the Akaike Information Criteria (AIC) was used to build a 3D model selection map. The variability of each pharmacokinetic parameter extracted from this map was assessed with a noise propagation procedure, resulting in voxel-wise distributions of the coefficient of variation (CV).
The model selection map over all patients showed NEM had the best fit in 35.5% of voxels, followed by ETM (32%), TM (28.2%), SSM (4.3%) and ESSM (< 0.1%). In analysing the reliability of Ktrans, when considering regions with a CV < 20%, ≈ 25% of voxels were found to be stable across all patients. The remaining 75% of voxels were considered unreliable.
The majority of studies quantifying DCE-MRI data in brain tumours only consider a single model and whole tumour statistics for the output parameters. Appropriate model selection, considering tissue biology and its effects on blood brain barrier permeability and exchange conditions, together with an analysis on the reliability and stability of the calculated parameters, is critical in processing robust brain tumour DCE-MRI data.
KeywordsGlioma DCE-MRI Shutter speed model Tofts model Primary brain tumour
Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a non-invasive methodology that allows tissue perfusion and permeability to be quantified through analysis of T1-weighted MR images acquired before, during and after an intravenous (IV) injection of a gadolinium-based contrast agent (CA).
Nowadays, DCE-MRI finds application in a wide range of oncological studies. In brain tumours, the Tofts model (TM), also known as the standard model, together with its extended version (EMT), are regularly applied [1, 2, 3, 4, 5, 6]. There are a few examples of the application of shutter speed model (SSM) in brain tumours in the literature [7, 8] while it has been mainly implemented in the study of breast cancer , prostate cancer  and hepatocellular carcinoma . The extended shutter speed model (ESSM), also called second generation shutter speed model, or BALDERO (blood agent level dependent and extravasation relaxation overview), has previously found applications only in hepatocellular carcinoma  and simulated data . These four models assume that the CA passes readily between the intravascular compartment and the tissue interstitium. However, this assumption is not valid in the presence of the intact blood-brain barrier (BBB), where there is negligible leakage. In this case the CA will only affect the intravascular T1 value on first pass of the bolus. In addition, the vascular component of a brain tumour is heterogeneous and certain regions may not be perfused (e.g. necrotic regions). We therefore also implemented a no-effect model (NEM) which assumes negligible effect of the CA on the voxel T1 properties.
In the last decade, the number of pharmacokinetic studies of DCE-MRI brain data has increased as this quantification technique has been found to be indicative of the malignant grade of brain tumours . However, two main issues need to be taken into account when analysing DCE-MRI data. Firstly, pharmacokinetic studies often show different and discordant results, thus bringing the reliability of this quantification technique into question [15, 16]. In fact, DCE-MRI data analysis is affected by (a) the acquisition protocol (trade-off between spatial and temporal resolution)  and (b) the quantification procedure. Furthermore, most studies present only results from application of a single pharmacokinetic model and the consequent statistical analysis of averaged values evaluated over the whole tumour volume [18, 19, 20]. This approach completely ignores the particularly heterogeneous nature of brain tumour vasculature and vascular permeability.
In this prospective study, we investigate the robustness of pharmacokinetic modelling of DCE-MRI brain data. We propose a method to identify reliable DCE-MRI data based on a model selection procedure, building on previous work by Bagher-Ebadian et al. , and a stability test: five different pharmacokinetic models (TM, ETM, SSM, ESSM and NEM) are assessed and the Akaike information criteria index is used for model selection; we then evaluate the stability of each parameter extracted from the model of choice, in terms of coefficients of variations (CVs), through a noise propagation procedure.
Materials and methods
Fourteen patients (7 male, 7 female; aged 23–73 years, mean 40 years) with primary brain tumours were recruited to this study. Ethical approval was given by the local ethics committee and informed consent was obtained from all patients. Patients had an MRI at diagnosis, prior to receiving any treatment. Following surgery, histopathological data showed three patients with WHO grade IV glioblastoma, two patients with WHO grade III astrocytoma, two patients with WHO grade III oligodendroglioma, three patients with WHO grade II astrocytoma, three patients with WHO grade II oligodendroglioma and one patient with a WHO grade I dysembryoplastic neuroepithelial tumour.
DCE-MRI data acquisition
MR images were acquired on a 3-T Siemens Verio MRI system using a 32 channel head coil; including pre- and post-contrast T1-weighted images, T2-FLAIR images and a DCE sequence with a variable flip angle pre-contrast T1 map acquisition. The DCE-MRI protocol included five pre-contrast spoiled gradient recalled echo (SPGRE) 3D vibe sequences at five different flip angles (2, 8, 12, 15, 20, 26), and a dynamic 3D vibe sequence (TR = 3.34 ms, TE = 0.99 ms, flip angle = 26, FOV 240 × 240 mm, acquisition matrix 128 × 128, slice thickness 5 mm, slice gap 1 mm, 80 volumes). To obtain acquisitions before, during and after the injection of the CA, 0.1 mmol/kg body-weight gadolinium-diethylene triaminepentacetate (Gd-DTPA, Gadovist) was injected using a power injector on the fifth acquisition using a flow rate of 3 ml/s immediately followed by 20 ml saline solution. The 3D acquisition allowed us to cover the entire brain; 80 time points were acquired with an average temporal resolution of 2.89 s and a total acquisition time of ≈ 4 min.
DCE-MRI data analysis
The longitudinal relaxation rate is determined in order to calculate the concentration of the CA. This is done through a calibration between the concentration of CA [CA] and the measured H2O MR signal. This can be modelled by either a linear or nonlinear relationship as described below.
Tofts model (TM)
Shutter speed model (SSM)
Both the TM and its extended version (2.3.3) embed the implicit assumption that equilibrium transcytolemmal water exchange (between the intracellular space and extracellular extravascular space) is infinitely fast, or that the system is in, what is called, the fast exchange limit . Water exchange between the intracellular space and the extracellular extravascular space effects the degree of T1 shortening caused by CA. To account for this effect on the brain MRI signal amplitude, we applied the shutter speed model (SSM), which introduces a new pharmacokinetic parameter, the mean intracellular water molecule lifetime, τi .
Extended Tofts model (ETM)
The ETM was fitted by substituting Eq. 13 in Eq. 6 using the MATLAB functions fminsearch and fminsearchbnd, similarly to the TM and SSM. The initial estimates for ETM Ktrans and ve were again taken from the output of the TM and the initial estimate for vp was set at 0.01 . The final Ktrans, kep, ve and vp maps were obtained from the fitting procedure with the best fit (lowest AIC) as described in Fig. 1.
Extended shutter speed model (ESSM)
I is the identity matrix and S0 (= (Sb0, So0, Si0)) is the signal at baseline.
The model was fitted by substituting Eq. 18 into Eq. 17 using the MATLAB function fminsearch function. The outputs of the SSM model were used as the initial estimates for Ktrans and vo and one third of the measured signal was used as the initial estimate for Sb0, So0 and Si0. Furthermore the initial estimates for kbo, kob, koi and kio were taken from literature defined values as 1.2, 1.5, 1.1 and 1.2, respectively . The following parameters were also derived: τb = 1/kbo, τi = 1/kio, vb = (kob − koi − ((kio/vo)fw) − kio)/((kbo + kio)/vo) where fw is tissue volume fraction accessible to mobile aqueous solutes (assumed to be a constant and set to 0.8) and vi = 1 − (vb + vo) .
No-effect model (NEM)
The no-effect model describes the case where [CA] within a brain voxel is so low that there is no permeability or vascular filling. In this situation, the MR signal is assumed to be unperturbed by the injection of the gadolinium-based CA and, as a consequence, the longitudinal relaxation rate does not change from its baseline value (R10). The system is therefore described by this value at all times such that the data is fitted by the constant R10.
Arterial input function (AIF)
Stability of pharmacokinetic parameters
Once the model selection had been selected, the stability of each parameter within the selected models was evaluated in a simulation environment. Tissue curves were generated back from the extracted pharmacokinetic parameters and signal intensity curves were calculated with the inverse formula of Eq. 4. White Gaussian noise was added to the signal intensity curves using a signal to noise ratio (SNR) of 20. The SNR value for the simulated data was set by evaluating the SNR of the acquired data from the second and third phase of the dynamic acquisition sequence using the subtraction method . The simulated noisy signal intensity curves were reconverted to noisy tissue concentration curves, and fitted to the selected pharmacokinetic model. This procedure was repeated 500 times for every kinetic parameter and the variability of each parameter was expressed in terms of coefficient of variation (CV): the percentage ratio between the standard deviation and the mean.
Model selection and parameter variability
Ktrans stability test results
CV ≤ 10%
CV ≤ 20%
CV ≤ 50%
TOT n. voxel
Whole tumour Ktrans mean
% n. voxel
% n. voxel
% n. voxel
Dysembryoplastic neuroepithelial tumour
In this study, we showed that in over one third of the brain tumours voxels (35.5%), standard model fitting of DCE-MRI data was inconclusive and therefore fitting these models to the data would lead to incorrect perfusion parameters. Considering a CV of 20%, only ≈ 25% of remaining voxels were found to be reliable. The reproducibility of this technique and, as a consequence, its reliability can be improved by improving the main sources of variability in quantitative DCE-MRI (the acquisition method and the quantification process) . Nonetheless, in most studies, the intrinsic heterogeneity of the lesion is ignored by quantifying the perfusion with one single pharmacokinetic model and, more importantly, carrying out statistical analyses on one single whole tumour statistic (usually the average). In this study, we investigated the reliability of DCE-MRI focusing on the quantification analysis. We took into consideration the particularly heterogeneous nature of brain tumour vascular permeability due to the presence of the BBB, as well as existence of necrotic regions and provided a method to identify robust DCE-MRI data based on a model selection procedure and a stability test.
MR scanners usually employ post processing perfusion tools which fit DCE data with the TM. This model (together with its extended version ) considers the system in a fast exchange limit [36, 37], assuming an infinitely fast transcytolemmal water exchange between the EES and the intracellular space, which does not affect the overall signal decrease [36, 37]. Therefore, many studies on the cell membrane water permeability coefficient have shown FXL to be physiologically unreasonable and inconsistent . The shutter speed model was introduced to reflect a more realistic tissue environment. The model accounts for the intercompartmental water exchange effect, modelling this non-infinitely-fast exchange with the mean intracellular water molecule lifetime τi. In 2005, Li et al. introduced a second generation of the shutter speed model which considers also a non-infinitely fast equilibrium transendothelial water exchange.
Model comparison and stability
The heterogeneity that exists in brain tumours means that one model is insufficient in explaining the different biologies that exist in different tumour regions. Multiple pharmacokinetic models are required for a complete description of the tissue. This variability is testified by the model selection procedure which showed how, in a single slice of one tumour, multiple models perform better. This result confirms the study of Bagher-Ebadian where they implemented a selection method based on nested models . They found that in the necrotic core of the tumour, models describing vascular filling with no microvascular leakage (similar to the TM) and leakage without vascular reabsorption were selected because of the lack of blood flow. They also hypothesised that the model describing leakage with reabsorption (similar to the ETM) would be selected in the fast growing rims of the lesion. Our results show that there are a number of regions in the tumour where the CA exudation is prevented by the BBB and where the concentration of CA is so low that the evidence of perfusion is missing. In this case, the use of the NEM is recommended as the use of different models could result only in overfitting the data. In fact, our results showed that no leakage of the CA into the interstitium and the lack of flow of the CA through the tissue made the NEM the model of choice for the majority of regions, particularly in the non-enhancing lesions (37.5% of voxels). The result is very close to the ETM (32%), which was the model of choice in the enhancing lesions (54.8%). This suggests that, in areas where there is enhancement, a model with three parameters performs better and that the choice is dependent on the underlying state of the tissue. In fact, both the ETM and SSM are fitted by three parameters but the third parameter is very different between the two models (vp for the ETM describing a vascular component in the tissue, and τi for the SSM describing the transcytolemmal water exchange). Furthermore, with the implementation of the ESSM, we saw that the transendothelial water exchange did not have any impact on the signal (compared with parameters derived by the fitting of simpler models). It is necessary to consider that the ESSM required nine parameters to be fitted and that the cost of fitting extra parameters is often contrary to the principle of parsimony. In fact, in fitting data to a noise-limited dataset, the estimation could be very poor and dependent on the optimization procedure itself (the initial conditions, for example) . We compared the AIC values from the different fitting procedures to check whether a model with more parameters is more appropriate than a simple one. The ESSM was selected as the model of choice by < 0.1% of voxels, indicating that a model with three parameters performed better in the description of brain tumours and further confirming the poor quality of fit observed for the ESSM model. Our outcome agreed with the results of Duan et al. . Using representative in silico and clinical (cervical cancer) DCE-MRI data, they demonstrated the sensitivity of complicated models (parameters > 3) to noise and their decreasing probability of being selected in low signal-to-noise data .
The reliability of DCE-MRI data is not only based on the goodness of fit of the chosen pharmacokinetic model, but also on the robustness of the extracted parameters. For this reason, we assessed, for each lesion and for each parameter, the coefficient of variation. We worked in a simulation environment where we added Gaussian noise to our signal and we fitted the noisy curves 500 times. This procedure resulted with a heterogeneous distribution of CVs that was not linked to contrast enhancement. In fact, Fig. 6 shows the plot of four different tissue activity curves together with the Ktrans value and its CV in an enhancing and non-enhancing tumour. The curves in Fig. 6b, c belong to the same enhancing lesion and while they correspond to similar Ktrans values of 0.81 and 0.52 [1/min], they varied by 12 and 128%, respectively. On the other hand, the curves in Fig. 6e, f belong to the same non-enhancing lesion and show regions with both a low (4%) and a high (97%) variability.
We set three different thresholds for the CV to evaluate the variability of the Ktrans. Table 1 shows the percentage of voxels and their relative mean Ktrans value at different CVs thresholds (10, 20 and 50%), for each lesion and for each patient. Higher grade glioma tend to have more voxels with a lower CV and also a more stable value of Ktrans while, for some of the other patients, the mean value of Ktrans is highly affected by the portion of voxels taken into consideration (P04, P07). This result confirmed the improper use of one average value in statistical comparisons of brain tumours. Not only because of the heterogeneity of the tissue under investigation but also, and more importantly, because it is affected by the reliability of fit within voxels.
Finally, Fig. 6 gives a graphical representation of this effect showing Ktrans values under the 20% threshold of CV covering only 25% of voxels (an average percentage value evaluated among all patients). This result suggested that only this selection of voxels represents robust values, which can be used in the following statistical analyses, as, more importantly, in clinical evaluations. The selection of the threshold that makes DCE-MRI robust is, however, dependent on the effect size that is being measured and hence will vary across studies.
The main limitation of this study is the small size of the dataset. Furthermore, the sensitivity of DCE-MRI data to water exchange effect was reduced by the 26° flip angle acquisition (exchange-minimized approach) . As a consequence, the precision of the τi parameter extracted might be low.
In conclusion, DCE-MRI methods hold great promise for quantitative in vivo evaluation of permeability and vascular properties under different pathophysiological conditions. It allows us to identify, and quantitatively measure, smaller changes in permeability for pathological conditions effecting the BBB, than would be observed through visuassessment of post-contrast T1-weighted images. Different models yield different pharmacokinetic parameters and, for this reason, a model selection is critical for the appropriate analysis of DCE-MRI time courses based on the regional tissue biology, specifically permeability and vasculature. Future work needs to assess the physiological basis for each model in the reliable selection of DCE-MRI data. The applicability of each model depends on the physiology, anatomy and heterogeneity of the tumour and the tumour microenvironment. In addition, due to the noisy nature of DCE-MRI data, a model selection procedure alone is not enough: pharmacokinetic parameters need to be validated with a stability test in order to give only robust results for statistical analyses and clinical evaluation.
This work was funded in part by the National Institute for Health Research (NIHR) Imperial Biomedical Research Centre (BRC), the Brain Tumour Charity and the Brain Tumour Research Campaign.
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
The study was reviewed and approved by the London–Fulham Research Ethics Committee.
Written and informed consent was obtained from all participants before recruitment to the study and all data was anonymised in accordance with the EU General Data Protection Regulation.
- 4.Ferl GZ, Xu L, Friesenhahn M, Bernstein LJ, Barboriak DP, Port RE (2010) An automated method for nonparametric kinetic analysis of clinical DCE-MRI data: application to glioblastoma treated with bevacizumab. Magn Reson Med 63(5):1366–1375. https://doi.org/10.1002/mrm.22335 CrossRefPubMedGoogle Scholar
- 8.Li X, Rooney WD, Varallyay CG, Gahramanov S, Muldoon LL, Goodman JA, Tagge IJ, Selzer AH, Pike MM, Neuwelt EA, Springer CS Jr (2010) Dynamic-contrast-enhanced-MRI with extravasating contrast reagent: rat cerebral glioma blood volume determination. J Magn Reson 206(2):190–199. https://doi.org/10.1016/j.jmr.2010.07.004 CrossRefPubMedPubMedCentralGoogle Scholar
- 9.Inglese M, Cavaliere C, Monti S, Forte E, Incoronato M, Nicolai E, Salvatore M, Aiello M (2019) A multi-parametric PET/MRI study of breast cancer: evaluation of DCE-MRI pharmacokinetic models and correlation with diffusion and functional parameters. NMR Biomed 32(1):e4026. https://doi.org/10.1002/nbm.4026 CrossRefPubMedGoogle Scholar
- 16.Kim H (2018) Variability in quantitative DCE-MRI: sources and solutions. J Nat Sci 4(1)Google Scholar
- 17.Guo Y, Zhu Y, Goud LS, Lebel RM, Shiroishi M (2015) High-resolution whole-brain dynamic contrast-enhanced MRI using compressed sensing. In: Biomedical optics & medical imaging. SPIE Newsroom, BellinghamGoogle Scholar
- 18.Fahlstrom M, Fransson S, Blomquist E, Nyholm T, Larsson EM (2018) Dynamic contrast-enhanced magnetic resonance imaging may act as a biomarker for vascular damage in normal appearing brain tissue after radiotherapy in patients with glioblastoma. Acta Radiol Open 7(11). https://doi.org/10.1177/2058460118808811 CrossRefGoogle Scholar
- 19.Little RA, Barjat H, Hare JI, Jenner M, Watson Y, Cheung S, Holliday K, Zhang W, O'Connor JPB, Barry ST, Puri S, Parker GJM, Waterton JC (2018) Evaluation of dynamic contrast-enhanced MRI biomarkers for stratified cancer medicine: how do permeability and perfusion vary between human tumours? Magn Reson Imaging 46:98–105. https://doi.org/10.1016/j.mri.2017.11.008 CrossRefPubMedGoogle Scholar
- 21.Yuan J, Chow SK, Yeung DK, Ahuja AT, King AD (2012) Quantitative evaluation of dual-flip-angle T1 mapping on DCE-MRI kinetic parameter estimation in head and neck. Quant Imaging Med Surg 2(4):245–253. https://doi.org/10.3978/j.issn.2223-4292.2012.11.04 CrossRefPubMedPubMedCentralGoogle Scholar
- 27.D’Errico J (2012) fminsearchbnd. 220.127.116.11 edn., MathWorks file exchange. https://in.mathworks.com/matlabcentral/fileexchange/8277-fminsearchbnd-fminsearchcon. Accessed 10 May 2019
- 32.Tofts PS, Brix G, Buckley DL, Evelhoch JL, Henderson E, Knopp MV, Larsson HB, Lee TY, Mayr NA, Parker GJ, Port RE, Taylor J, Weisskoff RM (1999) Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 10(3):223–232CrossRefGoogle Scholar
- 35.McRobbie DW, Moore EA, Graves MJ, Prince MR (2006) MRI from picture to proton, 2nd edn. Cambridge University Press, CambridgeGoogle Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.