The Emergence of Time-Resolved Contrast-Enhanced MR Imaging for Intracranial Angiography

Time-resolved contrast-enhanced MR angiography (MRA) (1), which was developed primarily for use in the diagnosis of abdominal and peripheral occlusive vascular (2), is emerging as a tool for use in the diagnosis of intracranial vascular disease. In recent years, due to advances in gradient hardware and the development of new pulse sequences, contrast-enhanced MRA can produce the high frames rates needed for intracranial applications.

In the initial intracranial application of thick-slab multiphase T1-weighted projection angiography, Wang et al (3) acquired 0.9 × 1.25-mm voxels with a thick-section (5.0-cm) two-dimensional (2D) imaging technique. They showed the feasibility of imaging the intracranial circulation with the simultaneous acquisition of two projection angles after the bolus injection of a gadolinium-based contrast agent with a rate of one frame 2.2 s. Hennig et al (4) adopted these ideas to image the pulmonary arteries with subsecond frames rates. More recently, Klisch and co-workers (5) were able to track the progress of intravascular embolization of arteriovenous malformations (AVMs) and dural arteriovenous fistulae. They obtained serial measurements of the cerebral circulation in response to endovascular treatment with a rapid 2D projection angiographic technique with an in-plane spatial resolution of 0.98 × 0.98-mm and a rate of less than one frame per second. In this issue of the American Journal of Neuroradiology, Coley and coworkers (6) report on the detection and classification of dural AVMs in three patients.

The use of MR techniques has several advantages compared with x-ray DSA; primarily, MR techniques are much less invasive than x-ray DSA. In addition to the fact that MR techniques produce images without the use of ionizing radiation, MR contrast agents are not nephrotoxic and have lower complication rates than those associated with x-ray contrast materials. The spatial localization possible with MR techniques also allow for the selection of a limited through-plane field of view so that targeted acquisition of the relevant vasculature is possible. Furthermore, since MR contrast agents are infused by means of a venous access site, such as that in an antecubital vein, MR angiography does not have the risk and complications associated with arterial punctures or the placement of a catheter.

Compared with nonenhanced MR imaging, such as phase-contrast MR imaging or time-of-flight imaging, contrast-enhanced MR techniques are less susceptible to artifactual signal loss associated with the saturation of slow-flowing blood. The rapid image acquisitions that are essential for contrast-enhanced MRA reduce examination time. Typically, localizers and angiograms can be acquired in less than 15 min. Time-resolved MR images can depict complex flow patterns that may not be seen on non–time resolved MR images. This ability allows visualization of complex flow patterns or delayed arrival of contrast material through collateral vessels that form as the result of occlusions. We have seen the utility of this in the aforementioned work in the depiction of AVMs. The production of a time series of images also opens the door for innovative techniques for vessel segmentation based on temporal characteristics (5, 7). These have been shown to increase the signal-to-noise ratio (SNR) of images and allow the physician to isolate specific blood vessels from the surrounding background and venous structures.

An obvious improvement with intracranial MRA is the acquisition of three-dimensional (3D) volume images. 2D acquisitions are restricted to a few centimeters of through plane coverage by signal loss caused by intravoxel spin dephasing. In addition to the greater coverage afforded by 3D volume acquisitions, 3D images have the added benefit of SNRs higher than those of 2D acquisitions. With the acquisition of 3D volumes with isotropic spatial resolution, the radiologist has the ability to reproject the relevant features in the image for optimal display or to manually track blood vessels by using individual source images.

One means of increasing frame rates in 3D MR digital subtraction angiography (DSA) is to acquire the critical central phase-encoding values more frequently than the phase-encoding values that correspond to high spatial frequencies. This is the basis of the 3D time-resolved imaging of contrast kinetics (TRICKS) pulse sequence (1). TRICKS has been used successfully to increase the frame rate of 3D multiphase examinations by factors of 3 or 4. These rapid frame rates have been particularly useful in acquiring high-quality angiograms of the carotid bifurcation (8, 9) where rapid venous opacification has proven to be problematic. Some authors have had excellent results in acquiring 3D volumes with exceedingly high frame rates without the use of TRICKS by using partial acquisition of k-space volumes (10). However, to increase the temporal sampling rate of the imaging sequence, fewer k-space lines are acquired, and these are asymmetric about the origin of k space. These acquisitions have high SNRs, but they lack the isotropic spatial resolution required for reprojection. 2D implementation of TRICKS has been used to achieve subsecond frame rates for MR catheter tracking (11), and more recently, they have been used for intracranial multislab 2D angiography (12).

A key difference between x-ray DSA and MRA is the relation between image acquisition time and spatial resolution. In MRA, for a given field of view, the acquisition time is the product of the matrix size, number of sections and TR. This limitation forces spatial and temporal resolution to be traded off, depending on the imaging application (13). TRICKS encoding has proven to be useful in gaining three- and fourfold increases in the frame rate compared with this simple frame rate–spatial resolution relation for 3D image acquisitions. However, 3D intracranial applications require subsecond frame rates and submillimeter voxels. Therefore, time-resolved MRA of the intracranial vessels currently is limited to high–frame rate 2D acquisitions, low–frame rate 3D acquisitions, or 3D acquisitions with highly anisotropic spatial resolution.

Given the requirements of submillimeter isotropic spatial resolution with subsecond frame rates and high SNRs, much work is required to improve time-resolved MRA. The use of conventional Fourier spin-warp imaging is limited in is ability to meet the demands of intracranial MRA. However, novel approaches to MR image acquisitions are beginning to show promise in addressing some of the technical challenges in the development of conventional MRA. Radial projection reconstruction (PR) k-space trajectories were introduced as the first technique for spatial localization in MR imaging. Unlike Fourier encoding in which spatial resolution and imaging time are proportional, imaging time in radial PR acquisitions is proportional to the number of angular samples. However, unlike traditional Fourier phase encoding, in PR, decreasing the number of angular samples does not decrease spatial resolution. Rather, decreasing the number of radial samples results is a low-intensity streak artifact similar to those seen on CT scans. The undersampling artifact has not proven to be problematic in MR imaging, because, in contrast-enhanced MRA, blood vessels are the dominant signal source; in CT, artifact from bone can confound diagnostic findings.

Peters et al (14) have shown that undersampled PR acquisitions allows exceedingly rapid acquisition of high-spatial-resolution images with minimal artifact. In this work, they achieved factors of 4 times fast acquisitions relative to Fourier-encoded acquisition of the same spatial resolution. Vigen et al (15) combined undersampled PR with TRICKS, or PRTRICKS, encoding in the section-select direction to further increase the frame rate in high-resolution (0.5 × 0.76 × 2.4-mm voxels; rate, one frame per 2.56 s) contrast-enhanced MRA imaging of the renal arteries. The first intracranial application of undersampled PR by Barger et al (16) resulted in two- to fourfold decreases in imaging times for non–time resolved phase-contrast flow measurements in the circle of Willis. This technique recently has been extended to time-resolved, fully isotropic 3D projection acquisitions (17). Time-resolved vastly undersampled isotropic projection VIPR acquisition of 256 × 256 × 256 volumes in less than 4 s is possible with an undersampling factor of 100.

An important consideration with high-spatial-resolution time-resolved acquisitions is the dependence of the SNR on the acquisition parameters. Because SNR is proportional to the voxel size and the square root of the imaging time, rapid high-spatial-resolution image acquisitions lose SNR in two ways. High-frame-rate high-resolution image acquisitions that use gradient hardware and pulse sequence improvements ultimately are limited by low SNR. However, we anticipate improvement in SNR and vessel contrast enhancement with the development of high-field-strength 3-T MR imaging units. Furthermore, the approval of high-relaxivity blood-pool contrast agents will augment intravascular signal intensity and overall vessel contrast enhancement, particularly with the short TRs that are required for rapid image acquisition.

In conclusion, time-resolved contrast-enhanced MRA is emerging as means of imaging intracranial circulation with high spatial and temporal resolutions. Work must still be done to allow the acquisition of full 3D image volumes with subsecond frame rates. However, the acquisition of submillimeter isotropic volumes to depict intracranial circulation with subsecond frame rates are on the horizon.

• Copyright © American Society of Neuroradiology