Measurement of Blood-Brain Barrier Permeability with T₁-Weighted Dynamic Contrast-Enhanced MRI in Brain Tumors: A Comparative Study with Two Different Algorithms

Abstract

The purpose of this study was to assess the feasibility of measuring different permeability parameters with T₁-weighted dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) in order to investigate the blood brain-barrier permeability associated with different brain tumors. The Patlak algorithm and the extended Tofts-Kety model were used to this aim. Twenty-five adult patients with tumors of different histological grades were enrolled in this study. MRI examinations were performed at 1.5 T. Multiflip angle, fast low-angle shot, and axial 3D T₁-weighted images were acquired to calculate T₁ maps, followed by a DCE acquisition. A region of interest was placed within the tumor of each patient to calculate the mean value of different permeability parameters. Differences in permeability measurements were found between different tumor grades, with higher histological grades characterized by higher permeability values. A significant difference in transfer constant (𝐾^trans) values was found between the two methods on high-grade tumors; however, both techniques revealed a significant correlation between the histological grade of tumors and their 𝐾^trans values. Our results suggest that DCE acquisition is feasible in patients with brain tumors and that 𝐾^trans maps can be easily obtained by these two algorithms, even if the theoretical model adopted could affect the final results.

1. Introduction

The blood-brain barrier (BBB) is formed by specialized endothelial cells lining capillaries in the central nervous system (CNS), and it prevents or slows the passage of some drugs and other chemical compounds, radioactive ions, and disease-causing organisms, such as viruses, from the blood into the CNS. BBB breakdown is associated with many CNS-related pathologies, including inflammatory diseases such as multiple sclerosis [1] and chronic and acute cerebrovascular pathology [2, 3]. Pathological modifications of the BBB have also been well described in degenerative diseases such as Alzheimer disease [4]; in addition, it has been shown that in brain tumors the BBB is structurally and functionally abnormal [5].

Quantitative investigation of BBB permeability is possible using Magnetic Resonance Imaging (MRI) [5], and it has been applied to the study of brain tumors [6, 7]. In particular, experimental and clinical studies have demonstrated that dynamic contrast-enhanced (DCE) MRI, with a macromolecular contrast agent (CA), can be used to quantify microvascular permeability in tumors [8] and that permeability increases with increasing histological tumor grade [7, 9].

Different theoretical models have been proposed for the analysis of DCE-MRI data in order to find a more accurate approach for the tumor vascular bed and discriminate blood flow and vascular permeability. These models include the standard [10] and the extended Tofts-Kety (ETK) models [11, 12], the adiabatic tissue homogeneity (AATH) model [13, 14], the two-compartment exchange model (2CXM) [15–17], the distributed capillary adiabatic tissue homogeneity (DCATH) model [18], and the gamma capillary transit time model (GCTT) [19]. The absolute quantification of permeability parameters may substantially differ on the basis of the specific model adopted to fit the DCE experimental data [11].

The purpose of this study was to quantify permeability parameters in patients with different histological types of brain tumors by using two different theoretical algorithms. To this aim, we compared the volume transfer constant between blood plasma and the extravascular extracellular space (EES) (𝐾^trans), the volume of EES per unit volume of tissue (𝑣_𝑒) (with 0≤𝑣_𝑒≤1), the vascular volume fraction within the tissue (𝑣_𝑝), and the washout rate from the EES back into the blood plasma (𝑘_𝑒𝑝). These parameters were calculated using the Patlak algorithm [20], which assumes that 𝑘_𝑒𝑝 is small, and therefore negligible, and the ETK model. Additionally, the results were evaluated in order to understand if the permeability metrics obtained with the two methods correlated with histological grade.

2. Materials and Methods

2.1. Patients

Twenty-five adult patients (eight females and seventeen males; mean age ± standard deviation = 54.6 ± 10.0 years; age range = 37–75 years) with different brain tumors were enrolled in this study. The histological grade of each lesion was determined by using the World Health Organization (WHO) classification of brain tumors [21].

2.2. MRI Protocol and Image Evaluation

MRI examinations were performed on a 1.5-T clinical system (Siemens Magnetom Avanto, Erlangen, Germany) with an 8-channel head coil, and the following imaging sequences were acquired: T₁-weighted (TR/TE = 500/10 ms; acquisition matrix = 320 × 384; NEX = 1; field of view = 250 × 300 mm; slice thickness = 5 mm, flip angle = 80°), T₂-weighted (TR/TE = 3800/95 ms; acquisition matrix = 320 × 384; NEX = 2; field of view = 250 × 300 mm; slice thickness = 5 mm, flip angle = 150°), and fluid attenuated inversion recovery (FLAIR) (TR/TE = 9000/119 ms; acquisition matrix = 320 × 384; NEX = 1; field of view = 250 × 300 mm; slice thickness = 5 mm, flip angle = 150°). Before intravenous injection of the CA, fast low-angle shot (FLASH) axial 3D T₁-weighted images (TR/TE = 6.7/1.0 ms; acquisition matrix = 320 × 384; NEX = 1; field of view = 250 × 300 mm; slice thickness = 5 mm) were acquired with multiple flip angles (5°, 10°, 15°, 20°, and 30°).

Acquisition of a DCE-MRI sequence was started immediately after intravenous administration of a gadolinium-based CA (gadobutrol 0.1 ml/kg, Gadovist) by a power injector (Spectris Solaris EP Medrad) at a rate of 5 mL/s. Dynamic axial 3D T₁-weighted TurboFLASH images (TR/TE = 6.7/1.0 ms; acquisition matrix = 320 × 384; NEX = 1; field of view = 250 × 300 mm; slice thickness = 5 mm; flip angle = 30°) were acquired for 30 time points.

Permeability maps were created as described in the Theory and Data Analysis section, and the mean of 𝐾^trans, 𝑣_𝑒, and 𝑣_𝑝 was obtained for each patient in a region of interest (ROI) within the tumor.

2.3. The Extended Tofts-Kety Model

The Patlak algorithm and the ETK model were used in order to calculate different permeability parameters.

The tracer concentration in tissue, 𝐶_𝑡, and the tracer concentration in arterial blood plasma, 𝐶_𝑝, are related by the differential equation 𝑑𝐶_𝑡𝑑𝑡=𝐾^trans𝐶_𝑝−𝑘_𝑒𝑝𝐶_𝑡, (1) where 𝑘_𝑒𝑝=𝐾^trans/𝑣_𝑒 is the rate constant [10].

The solution of (1), using the initial conditions 𝐶_𝑝=𝐶_𝑡=0 at 𝑡=0, is 𝐶_𝑡(𝑡)=𝐾^trans𝐶_𝑝(𝜏)exp−𝑘_𝑒𝑝(𝑡−𝜏)𝑑𝜏, (2) where 𝑡 represents the current time step, 𝜏 is the variable of integration, 𝐶_𝑡(𝑡) is the time course of the CA concentration in the tissue compartment, and 𝐶_𝑝(𝑡) is the time course of the CA concentration in the plasma (AIF) [22, 23].

Equation (2) is the standard Tofts-Kety model and is acceptable in tumors with no large increase in blood volume; however, it is not valid in other contexts, for instance when blood volume can increase markedly in neoplasms [24]. Models of additional sophistication are required to adequately describe these cases. It is possible to extend (2) to include the concentration of CA in the blood plasma: 𝐶_𝑡(𝑡)=𝐾^trans𝐶_𝑝(𝜏)exp−𝑘_𝑒𝑝(𝑡−𝜏)𝑑𝜏+𝑣_𝑝𝐶_𝑝(𝑡). (3) Equation (3) is the ETK model, where 𝑣_𝑝 is the vascular volume fraction within the tissue. Equations (2) and (3) are the basis of the most of DCE-MRI experiments currently being reported in the literature. The general approach is to measure 𝐶_𝑝 and 𝐶_𝑡 time courses and perform a nonlinear least squares fit of these equations to such data. By varying the parameters in those equations, it is possible to obtain estimates on 𝐾^trans, 𝑣_𝑒 and 𝑣_𝑝. For this study, postprocessing analysis for the ETK model was performed using in-house software.

2.4. Patlak Analysis

This analysis assumes that the rate constant between EES and blood plasma (𝑘_𝑒𝑝) in (3) can be ignored due to low permeability and short measuring time. In these cases, (3) can be reduced to 𝐶_𝑡(𝑡)=𝑣_𝑝𝐶_𝑝(𝜏)+𝐾^trans𝐶_𝑝(𝜏)𝑑𝜏. (4) Using the Patlak analysis, it is possible to linearize (4) to obtain a graph of the ratio 𝐶_𝑡(𝑡)/𝐶_𝑝(𝑡) versus ∫𝐶_𝑝(𝜏)𝑑𝜏/𝐶_𝑝(𝑡) in order to calculate the values of 𝐾^trans and 𝑣_𝑝. It is useful to keep in mind that this model does not take into account the backflow and, therefore, the results could have some limitations [25]. For this work, postprocessing analysis for the Patlak algorithm was performed using TOPPCAT (T-One weighted Perfusion imaging Parameter Calculation Toolkit) [https://dblab.duhs.duke.edu/] [26], which is a free permeability analysis software available on Internet.

2.5. The Voxel Relation Rate

The calculation of the permeability parameters by using (2), (3), or (4) requires the values of 𝑆₀, the equilibrium longitudinal magnetization, and 𝑇₁ pregadolinium mapping for each voxel. Precontrast 𝑇₁ mapping of tissue can be obtained with different approaches; a common method employs multiple 3D gradient echo acquisitions at multiple variable flip angles. For this purpose, pulse sequence and flip angle must be chosen to maximize the sampling rate within the constraint of an adequate signal-to-noise ratio (SNR). Another frequently used method to calculate the 𝑇₁ baseline values utilizes 2D T₁-weighted inversion recovery scans at various inversion times [27, 28]. This method is considered more accurate than the variable flip angle method, but it is usually more time consuming.

In our study, we chose to use acquisitions at multiple variable flip angles to obtain 𝑆₀ and 𝑇₁ values. In FLASH MRI with a complete spoiling of the transverse magnetization, the steady-state signal is given by 𝑆_𝛼=𝑆₀1−exp−TR/𝑇₁sin𝛼1−cos𝛼⋅exp−TR/𝑇₁, (5) where TR is the repetition time and α is the flip angle for TR >> T2* [29]. To construct the 𝑇₁ and 𝑆₀ maps, it is necessary to fit the data from FLASH sequence by a linearization of (5). The 𝑇₁ and 𝑆₀ values are then utilized to estimate the voxel 𝑅₁ time courses, or relaxation rates, from the acquired signal intensity time courses, 𝑅₁(𝑡)1=−𝑆(𝑡)(0)TR×ln1−−𝑆𝑆₀+sin𝛼1−𝑚()×𝑆(𝑡)(0)1−𝑚⋅cos𝛼1−cos𝛼−𝑆𝑆₀+sin𝛼1−𝑚1−(𝑚⋅cos𝛼)⁻¹, (6) where 𝑚=exp[−TR/𝑇₁], α is the flip angle of the DCE-MR sequence, and 𝑆(0) and 𝑆(𝑡) are the signal intensities at time 𝑡=0 and time t, respectively. The relaxation rate 𝑅₁(𝑡) is related to the tracer concentration 𝐶_𝑡(𝑡) by 𝑅₁(𝑡)=𝑅₁₀+𝑟₁𝐶(𝑡), (7) where 𝑅₁₀ is the relaxation rate before tracer injection and 𝑟₁ is the relaxivity of the CA, which varies with the molecule’s contrast type (in our case 𝑟₁=4.3 mM⁻¹s⁻¹) [30].

2.6. Statistical Analysis

The Kolmogorov-Smirnov test of normality was used to determine whether the distribution of values was normal, and comparisons between these two theoretical models were performed using the student’s t-test. A P value lower than 0.05 was considered to indicate statistical significance for all comparisons. The relationship between 𝐾^trans and tumor grade was assessed by using Pearson correlation coefficients.

3. Results

The study group consisted of twenty-five patients, whose histological diagnoses were as follows seventeen glioblastoma multiforme (WHO grade IV), two oligoastrocytoma (WHO grade III), two anaplastic astrocytomas (WHO grade III), one pleomorphic xanthoastrocytoma (WHO grade III), and three astrocytomas (WHO grade II).

Differences in permeability measurements were found between different tumor grades, with higher histological grades characterized by higher permeability values. In particular, for grade IV tumors, the mean 𝐾^trans value calculated by using the Patlak algorithm was 0.039 ± 0.016 min⁻¹, and the mean 𝐾^trans value calculated by using the ETK model was 0.051 ± 0.015 min⁻¹ (Figure 1). For grade III tumors, the mean 𝐾^trans value was 0.032 ± 0.011 min⁻¹ for the Patlak algorithm and 0.043 ± 0.022 min⁻¹ for the ETK model. In grade II tumors, we obtained a mean 𝐾^trans value of 0.010 ± 0.006 min⁻¹ with the Patlak algorithm and a mean 𝐾^trans value of 0.011 ± 0.007 min⁻¹ with the ETK model. Statistical analysis showed significant differences in the mean 𝐾^trans values obtained with the Patlak algorithm versus those obtained with the ETK model for grade IV tumors (𝑃=0.049,𝑡=2.04) and for high grade tumors (i.e., WHO IV and WHO III grouped together) (𝑃=0.031,𝑡=2.02). No significant differences were found between mean 𝐾^trans values obtained with the two algorithms for grade II tumors. In addition, we did not find significant differences in mean 𝑣_𝑝 values between the two methods in both high-grade tumors and low-grade tumors.

(a)

(b)

(c)

Figure 1 

Image and intensity data from a 69-year old female patient with glioblastoma multiforme (WHO IV). (a) T1-weighted postcontrast image. (b) Relative 𝐾trans map. (c) Signal intensity plot for a region of the tumor and for a portion of healthy brain tissue.

Additionally, we calculated 𝑣_𝑒 only for the ETK model, finding for high-grade tumors (WHO IV and WHO III grouped together) a mean 𝑣_𝑒 of 0.15 ± 0.05, whilst a mean 𝑣_𝑒 of 0.017 ± 0.006 was found for low-grade tumors.

The histological grade of tumors for these patients was found to have a significant correlation with 𝐾^trans values. For the Patlak algorithm, we found a Pearson correlation coefficient 𝑟=0.54(𝑃=0.004), and for the ETK model we found a Pearson correlation coefficient =0.58(𝑃=0.002). Correlations between tumor grade and mean 𝐾^trans values are shown in Figure 2. No significant correlation was found between 𝑣_𝑝 and tumor grade for both methods.

(a)

(b)

Figure 2 

A statistically significant correlation was observed between mean 𝐾trans and tumor grade for both Patlak algorithm ((a): 𝑟=0.54(𝑃=0.004)) and for the ETK model ((b): 𝑟=0.58(𝑃=0.002)).

4. Discussion

Over the past two decades, important progress has been made in the development of a robust method to noninvasively quantify the microvascular permeability of the BBB for clinical use [11]. BBB can be altered in brain tumors neoangiogenesis because of new vessels, which are structurally and functionally abnormal. This abnormality impairs effective delivery of therapeutic agents to all regions of tumors, creates an abnormal microenvironment (e.g., hypoxia) that reduces the effectiveness of radiation and chemotherapy, and selects for more malignant cells [31]. BBB disruption caused by tumors is heterogeneous, thus, the permeability can vary widely in different areas of the same tumor. Creation of parametric maps with a 𝐾^trans value for each voxel can be of practical importance and useful, for instance, to guide the biopsy target [32].

Different theoretical models can be used to fit DCE-MRI experimental data, and the resulting permeability parameters can be influenced by the model used. In this study, 25 patients with different brain tumors underwent DCE MRI. The Patlak algorithm, which assumes that the washout rate from the EES back into the blood plasma is small and therefore negligible, and the ETK model were applied to estimate kinetic parameters. Our results showed that 𝐾^trans mean values in ROIs within high-grade tumors were significantly different between the two methodologies, with higher 𝐾^trans mean values calculated with the ETK model.

In many studies, 𝐾^trans is estimated by applying the TK model or the ETK model. However, the main disadvantage of the TK model is that it overestimates 𝐾^trans in highly vascularized regions, since the contribution of intravascular CA to the signal enhancement is mistaken as tracer that enters the EES and thus appears to reflect permeability [33]. Harrer et al. [33] used the TK and ETK models to estimate permeability on 18 high-grade gliomas, and they found that 𝐾^trans values calculated by using the TK model were considerably higher than the 𝐾^trans values obtained using the ETK technique. On the other hand, Port et al. [34] estimated the difference between noncompartmental, TK and ETK models on 20 patients with recurrent glioblastoma, finding significant differences in the results obtained with the different models.

In our study, we found statistically different 𝐾^trans values for high-grade tumors, but no significant difference for low-grade tumors. This result may be related to the low number of patients with grade II tumors in this study, but biological differences between high-grade tumors and low-grade tumors can also be associated with these findings.

Both techniques reveal a significant correlation between the histological grade of tumors and their 𝐾^trans values. These data are in line with other results that can be found in the literature [35, 36].

5. Conclusion

In conclusion, we demonstrated that DCE acquisition is feasible in patients with brain tumors and that the choice of the postprocessing tool can influence the permeability metrics. In particular, the use of the Patlak algorithm versus the ETK model can lead to statistically significant differences in the 𝐾^trans values. In our small sample, this difference in results did not affect the correlation with histological grade.