Brain tumor perfusion: Comparison of dynamic contrast enhanced magnetic resonance imaging using T₁, T₂, and $T_2^{*}$ contrast, pulsed arterial spin labeling, and H₂¹⁵O positron emission tomography

Abstract

Objectives

Different techniques for measuring of perfusion are clinically available, but these are usually applied to healthy brain tissue.

Material and methods

Five different techniques were used here in 12 patients with brain tumors to investigate the impact of tumor vascularization on the perfusion signal: three qualitative dynamic contrast-enhanced/susceptibility-contrast magnetic resonance imaging (DCE-MRI/DSC-MRI) techniques exploiting T₁, T₂, $T_2^{*}$ contrast, and two quantitative techniques, pulsed arterial spin labeling (PASL) and H₂¹⁵O positron emission tomography (H₂¹⁵O-PET).

Results

In a first approximation, a linear correlation was found between all five imaging modalities regarding the perfusion signal of both, normal brain tissue and tumor. The estimated values for tumor perfusion differed significantly between the techniques (1 = methodical mean in arbitrary units): PASL: 0.83, H₂¹⁵O-PET: 0.62, T₁-DCE: 1.73, T₂-DCE: 0.69, $T_2^{*}\text{-DSC}$: 0.89.

Conclusions

The tumor perfusion values, determined with different techniques are not comparable. The $T_2^{*}\text{-DSC}$, here applied with contrast agent presaturation of extravascular space, and PASL depict median perfusion most reliably.

Introduction

Perfusion is increasingly being used as a functional parameter for tumor diagnostics and is especially useful for assing the outcome of tumor treatment. The growth of tumors beyond a certain size largely depends on the development of a vascular supply adequate for the metabolic requirements of the neoplastic tissue. Once a growing tumor reaches a critical mass, it induces growth of new blood vessels. Tumor metabolism varies with the function and architecture of these new blood vessels [1], [2], [3]. Various techniques for assessment of tumor perfusion are now in clinical use after having been investigated in healthy brain tissue.

Perfusion and blood volume can be qualitatively determined using dynamic contrast-enhanced/susceptibility-contrast magnetic resonance imaging (DCE-MRI/DSC-MRI) [4]. These parameters determine the shape of signal–time curves of tissue extracted from MR images acquired after the intravenous administration of a short contrast medium (CM) bolus. Both perfusion and blood volume can be used to characterize tumors, in particular to differentiate glioma grades [5], [6], [7], [8], [9] as well as different brain tumors [10], [11], [12], [13].

However, in case of blood–brain barrier disruption, methodical limitations of common perfusion measurement techniques in tumorous tissue become evident. Contrast medium (Gd-DTPA) extravasation modifies the first pass bolus signal used for perfusion measurement. In $T_2/T_2^{*}$ contrast-MRI, the bolus signal is reduced due to the T₁-effect of the extravasated CM [14], [15], [16]. In contrast with T₁-DCE-MRI, perfusion may be overestimated due to amplification of the intravascular signal by the signal from the extravasated CM [17].

Two additional quantitative methods using water as a tracer are in clinical use: H₂¹⁵O positron emission tomography (PET) and pulsed arterial spin labeling (PASL). The main difference between the water tracer methods is the decay time of labeling. While ¹⁵O decays with a half-life of T_1/2 (¹⁵O) = 122.24 s, the magnetic labeling of inflowing blood water spins decays much faster at a field strength of 1.5 T (within approximately T_1B = 1.2 s). Both techniques use the one-compartment model for quantification of perfusion [18]. This model assumes a complete exchange of blood and tissue water. This prerequisite for application of the one-compartment model might not be completely fulfilled and may affect the perfusion determined in tumors [19], [20], [21].

We investigated the perfusion signal of tumors using different common imaging techniques. The techniques were applied to brain tumors with a disrupted blood–brain barrier and evaluated separately for tumor and healthy brain tissue. The methodical differences of the techniques are presented and discussed.

Twelve patients with contrast-enhancing brain tumors (9 high grade gliomas (WHO grades III–IV), 2 meningiomas, 1 cerebral metastasis) were investigated. No patient had undergone any major therapeutic interventional procedures or any other major therapy during the preceding 6 months. The investigations were approved by the local ethics committee. All patients gave their informed consent before the MRI and PET examinations. The patients ranged in age from 17 to 78 years, mean 53 years. All tumor specimens were histologically examined following extirpation or biopsy and classified according to the WHO grading system.

Conventional and dynamic MR images were acquired during the same session to ensure an exact comparison of the results obtained. Imaging was performed on a Magnetom Vision (Siemens, Erlangen, Germany) 1.5 T superconducting whole-body MR imager. Prior to imaging, a 20-gauge catheter was inserted into a large peripheral vein, usually the antecubital vein. A saline drip was used to maintain patency of the vein. In all patients, the contrast agent Gd-DTPA (Magnevist, Schering, Berlin, Germany) was administered with a power injector (Spectris; Medrad, Volkach, Germany). The images were acquired in the following order.

The MRI examination began with the acquisition of anatomic images of the whole brain by using double-echo intermediate-weighted T₂- and T₁-weighted sequences with the following parameters: sections, 21; thickness, 5 mm; field of view, 230 mm; and matrix, 256 × 256. With T₂-weighted sequences, parameters were as follows: repetition time (T_R)/echo times (T_Es), 3800/22, 90 ms, with a 90° flip angle. For T₁-weighted sequences, the parameters were as follows: 735/14 ms, with a 70° flip angle. Tumor regions were identified on the T₂-weighted images, and the central tumor section was chosen for the perfusion studies.

Arterial spin labeling imaging was performed by using a quantitative version of arterial spin labeling imaging. The technique uses a single subtraction with addition of thin-section periodic saturation after inversion and a time delay (Q2TIPS) tagging scheme, which is a pulsed arterial spin labeling method that enables acquisition of multiple slices [22]. In the Q2TIPS technique, the time delay after inversion is called τ. After the slice-selective and non-selective inversion pulse and a time delay τ, the trailing edge of the tagged bolus is cut off by a train of regional saturation pulses producing a sharply defined blood bolus of temporal length τ. If the imaging slice is acquired after T_I = τ + Δt, where Δt is the maximum transit time within the slice, the resulting difference image is a quantitative perfusion image.

Slice-selective and non-selective inversion was accomplished by applying hyperbolic secant pulses of variable bandwidth and shape [23]. For the 35-mm inversion slab that was used in most cases, a 10,240-μs pulse of shape B₁(t) = sec h(βt) and a frequency offset of Δω(t) = −μβ tanh(βt), where Δω(t) is expressed in rads per second, with β = 1200 and μ = 18, were chosen. Multi-slice perfusion imaging was performed by using a single-shot echo-planar imaging readout. Imaging parameters were as follows: T_R/T_E = 3700, 29 ms (asymmetric echo), T_Is 1300, 1430, and 1560 ms (for slices 1–3, respectively); three slices; slice thickness, 8 mm; inter-slice gap, 3 mm; t, 1200 ms; Q2TIPS saturation length, 100 ms; saturation slab thickness, 40 mm; matrix, 128 × 128, interpolated to 256 × 256; field of view, 230 mm; bandwidth, 1250 Hz/pixel; signals acquired, 50; and total acquisition time, 6 min 10 s. For quantification, single-shot echo-planar MR images of all slices were acquired without prior inversion at the beginning of the sequence (M₀ images).

Dynamic imaging was performed with a series of 60 rapid flow-compensated inversion recovery spoiled gradient echo TurboFLASH brain images (T_R/T_E/T_I = 11.0, 4.2, 300 ms, α = 25°, slice thickness = 5 mm) taken in one slice over 60 s. This sequence was repeated 12 times with one-second gaps for technical reasons resulting in a total dynamic imaging time of 12 min. The matrix size was 64 × 128 with a nominal voxel size of 1.80 mm × 3.59 mm × 5.0 mm (FOV = 230 mm). After data acquisition the matrix size was zero filled to 128 × 128. A total dose of 0.1 mmol Gd-DTPA/kg body weight was administered intravenously by constant rate infusion followed by saline injection with the same flow. In all patients Gd-DTPA was injected within 4 s and the saline within 6 s with the flow rate adjusted to the administered dose. Infusion of the paramagnetic CM was started simultaneously with the acquisition of the fifth TurboFLASH image of the first series.

First-pass bolus MR imaging was performed next by using the same slice orientations and thicknesses as were used for arterial spin labeling. At least one slice was positioned at the level of the middle cerebral artery to obtain the arterial input function; one or more slices were added if none of the slices used in arterial spin labeling contained the middle cerebral artery. The extravascular extracellular space was presaturated with CM administered for T₁-DCE-MRI, reducing T₁-shortening effects in case of blood–brain barrier disruption while applying the $T_2/T_2^{*}\text{-technique}$. Dynamic MR imaging was then performed by using a single-shot double-echo echo-planar sequence. After a 90° excitation pulse, a GRE echo-planar MR image was acquired first. Following a 180° slice-selective refocusing pulse, a spin-echo echo-planar MR image of the same slice was recorded. The T_E for the GRE and spin-echo MR images was 35 and 105 ms, respectively. The other imaging parameters were as follows: T_R, 1000 ms; matrix, 80 × 128, interpolated to 256 × 256; field of view, 350 mm; slice thickness, 8 mm; inter-slice gap, 3 mm. Three slices each containing 80 images were acquired with a repetition time of 1 s. Five seconds after starting the acquisition, 0.2 mmol/kg of CM was administered at a flow rate of 5 mL/s.

Transverse and coronal T₁-weighted MR images were acquired with the same parameters as were used to obtain the anatomic images. In patients in whom a subsequent radiation therapy was planned, a high-spatial-resolution three-dimensional turbo fast low-angle shot sequence (magnetization-prepared rapid acquisition GRE sequence) was acquired instead, and transverse images were reconstructed from that data set.

The PET scans were performed using a whole-body scanner (Siemens/CTI, ECAT-Exact/921) within one week before or after the MR scan. The scanner acquires 47 simultaneous slices of approximately 3.4 mm full width half maximum spatial resolution (FWHM) over an axial distance of 16.2 cm. It has a best-case reconstructed spatial resolution of 5.5 mm FWHM in both axial and transaxial directions. Reconstructed spatial resolution under imaging conditions is approximately 8 mm FWHM for low energy positron emission. A 10 min transmission scan was acquired with a ⁶⁸Ge/⁶⁸Ga-ring source for attenuation correction. For each scan, 1.0 GBq H₂¹⁵O in 6–8 ml physiological saline solution was administered as an intravenous bolus followed by 20 ml 0.9% saline solution at a flow rate of 2 ml/s. Forty-six frames were collected over an 11-min period (35 × 2 s, 8 × 30 s, 3 × 120 s) beginning with the start of the injection.