The Carotid Artery Through a Keyhole

Universal acceptance of an MR angiographic method for imaging the carotid arteries has been elusive. Despite numerous improvements in data acquisition and processing for 2D and 3D time-of-flight (TOF) techniques, the limitations imposed by spin dephasing caused by disordered flow, spin saturation, or both adversely affect the accurate measurement of stenosis on source and maximum intensity projection (MIP) images in many patients. Because of these flow effects, MR angiography is sometimes relegated to the role of “road map” in the noninvasive examination of neurologically symptomatic patients considered for carotid endarterectomy, with Doppler sonography (or even CT angiography) providing the stenosis measurement.

To overcome the limitations of TOF studies of the carotids, some radiologists have turned to ultrafast 3D gadolinium-enhanced methods. These methods, however, require strategies to ensure that the central lines of k-space are acquired during the first passage of the intravenous bolus of gadolinium through the carotid arteries, prior to gadolinium arrival in the jugular veins. Protocols that involve a timed injection (mechanical or manual) and a linear phase-encoding profile for filling k-space are commonly used. These protocols often incorporate a “test dose” or automatic bolus detection technique with scan triggering, if available, to determine the optimal time delay between contrast infusion and 3D data acquisition. Unfortunately, such protocols can be time-intensive for the radiologist, and may not temporally resolve carotid and jugular enhancement. In those cases, carotid stenosis can be obscured on reformatted source or targeted maximum intensity projection (MIP) images by a closely adjacent jugular vein.

To achieve better temporal resolution in a series of images without markedly compromising spatial resolution, two approaches can be identified. The first is to accelerate acquisition of the full space matrix with ultra-short echo and repetition (TE/TR) times that require more powerful gradient sets. The second is to restrict the size of the acquired k-space data and to use one of several methods of “constrained reconstruction” to generate the images in the time series. In this issue of the American Journal of Neuroradiology (page 263), Melhem et al have used one of the earliest and simplest of the constrained reconstruction methods, keyhole imaging. In their study, the acquisition of a complete (102 phase x 256 read) k-space reference matrix was followed by reduced acquisition of only the central 46 phase-encoding steps. While this decreased imaging time by a factor of 102/46, intrinsic spatial resolution of the update images along the phase-encoding direction would also be decreased by the same factor. With the keyhole method, the missing outer k-space rows of the update matrices are filled with the corresponding rows of the reference matrix prior to image reconstruction. Optimally, the result is a series of images with apparent high resolution.

Melhem et al observed that MIP projections generated from subtracted images had superior vessel-to-background contrast compared to MIP projections from unsubtracted images. The subtracted images, however, show only the changes between the updated and reference data sets, and thus have low spatial resolution along the phase-encoding direction (1). This limitation, with the relatively large partition thickness (5 mm), conspires to undermine the spatial resolution of their gadolinium-enhanced images, making them less impressive compared to standard TOF images. An alternative method, the so-called 3D-TRICKS (3D time-resolved imaging of contrast kinetics) proposed by Korosec and colleagues (2) produces images with slightly lower temporal resolution (4.5 sec/3D volume versus 3.6 sec/3D volume reported by Melhem et al) yet much higher spatial resolution. The 3D-TRICKS method includes intermittent acquisition of full k-space update images, shared data among update images, and temporal interpolation between acquired data sets. These capabilities improve the high spatial frequency information in the gadolinium-enhanced time series compared to the basic keyhole method.

Despite the claim by Melhem et al that their keyhole method can be implemented on clinical scanners with average gradient performance, the probability that most manufacturers will be ready and willing to do this is low. It is more likely that a sophisticated constrained reconstruction method, with minor variations among manufacturers, will become available commercially once the efficacy of gadolinium-enhanced carotid MR angiography, which can be performed by a technologist, has been demonstrated in clinical trials. Many questions remain, though. What will be the role of unenhanced TOF angiography? Which MR angiographic method will be used to examine the carotid siphon and common carotid origin for possible tandem stenosis? How well can an x-ray angiographic “string sign” of the cervical carotid be detected? Until these and related questions are answered, we will not know whether time-resolved, contrast-enhanced MR angiography will become the universally accepted MR method for assessing carotid disease.