The signal intensity that is measured in MRA studies of flowing blood depends on blood flow velocities (both magnitude and direction) and the variation of those quantities (e.g., from cardiac pulsatility) over the time that data are acquired. The design and implementation of imaging strategies and the choice of imaging parameters therefore require careful consideration of the anticipated hemodynamics in the vessels, in both the normal and the diseased conditions. Setting aside the problem of gross motion for the moment, MRA methods provide accurate images of the lumen when the image resolution is sufficiently high, to provide several pixels across the diameter of the lumen, where flow is sufficiently high that blood traveling through the imaging volume receives only a small number of radiofrequency (RF) excitations, and where the flow patterns remain regular and reproducible throughout the cardiac cycle. Deviations from these flow conditions can occur in both healthy and diseased vessels (1,2 and 3).

Variations in flow conditions can result from large differences in systolic and diastolic flows. This can be particularly pronounced; for example, in the arteries of the lower extremities where a pronounced interval of rapid antegrade flow is followed by a short, smaller retrograde component, which is, in turn, followed by an extended diastolic interval where the flow velocities are extremely low. In vivo flow patterns are also profoundly impacted by variations in geometry. Even in healthy individuals, blood vessels are curved with numerous branches and bifurcations. These geometric variations can substantially impact the velocity fields resulting in features, such as flattened profiles across the vascular lumen at the entrance to branch vessels or helical flow patterns in curved vessels, such as the aortic arch. However, in the absence of disease, flow is generally laminar and reproducible from cycle to cycle. In regions of atherosclerosis the vascular wall can be irregular, with a residual lumen that is highly asymmetric and that has rapid changes in diameter. These geometric variations can create regions of disturbed flow, such as velocity jets emanating from a tight stenosis accompanied by the shedding of flow vortices representing turbulent flow that varies from one cycle to the next.

All MR angiographic techniques aim to create high contrast between spins that are moving and those that are stationary (4,5,6,7,8 and 9). MR imaging methods are capable of measuring both the magnitude of the transverse magnetization and the orientation of that magnetization in space (the phase). Methods have therefore been devised that are designed to create large differences in either the magnitude or the phase of the magnetization between spins that are stationary and spins that are moving (5,7). MR sequences that rely on blood flow to transport fully magnetized blood into the imaging volume and thereby create a substantial difference between the magnetization of flowing and stationary spins are generally referred to as time-of-flight (TOF) methods, which display the magnitude of the transverse magnetization (4). Sequences that rely on the presence of contrast agents injected into the bloodstream to enhance vascular signal are referred to as contrast-enhanced MRA (CE-MRA), and they create images that display the magnitude of the transverse magnetization (10). Images that display the phase of the magnetization are referred to as phase contrast (PC) images (11). These methods rely on the motion of spins, with respect to the imaging gradients for vessel-to-stationary tissue contrast.

Accurate spatial localization, providing images with a high signal-to-noise ratio (SNR), is accomplished by ensuring that in the center of the readout interval, an echo is formed where the orientation of the transverse magnetization of all excited material is the same (rephased). In general, when imaging gradients are applied, spins will accumulate a phase shift. Conventional imaging gradients that are designed to generate a signal echo from stationary material do not account for the motion of flowing spins and, at the center of the echo, moving spins accumulate a phase that depends on their motion between the RF excitation and the time the echo is collected. In a voxel containing blood spins moving with a spread of velocities, various phases occur. The mean magnetization in the voxel is the sum of the individual vectors, and because of the spread of phases, a drop in signal strength, referred to as intravoxel phase dispersion, takes place. For many MR angiographic sequences it is important to implement gradient waveforms such that the magnetization of moving blood is also rephased at the center of the echo (12,13). These are referred to as motion compensation gradients (14). Whereas conventional gradient echo (GRE) sequences use gradient waveforms consisting of two lobes of gradient field strength applied with opposite polarity, flow compensation gradients require additional lobes of applied gradients. In principle, flow compensation could be pursued to include all orders of motion (15). In practice, however, the added gradient lobes needed for higher order motion compensation require additional time to play out, thus lengthening the echo time. If motion compensation is used, the most effective method for improving signal retention from moving spins, even in disturbed flow that contains highorder motion terms, is to rephase constant velocity terms only (16,17,18 and 19).

Most MRA sequences incorporate velocity compensation gradients along the directions of the slice selection and
frequency encoding gradients. The phase encoding direction is not compensated; the short duration of the phase encoding gradients minimizes the effects of motion during the application of those gradients.

The contrast obtained in an MRA study is closely related to the strength of the longitudinal magnetization in flowing blood relative to that in stationary tissue (6,20,21). In a GRE sequence (flip angle < 90 degrees), the strength of the longitudinal magnetization of spins subjected to a series of RF pulses decreases with each excitation. The longitudinal magnetization of spins will continue to decrease with increasing number of RF excitations received, a process referred to as saturation, until it reaches a steady-state value that is determined by the flip angle, the repetition time (TR), and the T1 relaxation time for that tissue. Stationary material remains in the imaging volume throughout data acquisition and, therefore, the magnetization strength of stationary spins decreases to the steady-state value. In blood vessels, the longitudinal magnetization of spins at any given location depends on how many RF excitations those spins have received between entering the imaging volume and reaching that location. Fast moving blood may receive only a few RF pulses, thus retaining substantial magnetization strength. Slowly moving spins may receive a large number of pulses, and for those spins, their longitudinal magnetization, such as stationary spins, may also drop to the steady-state value. In magnitude images, spins that have received many RF pulses and are strongly saturated will appear dark, whereas spins that retain substantial magnetization strength will appear bright.

In many cases, evaluating an arterial structure is confusing when it is obscured by an overlying vein and vice versa. To eliminate the signal from these vessels, a presaturation band with a flip angle >90 degrees can be placed adjacent to the imaging slice or volume so that the blood is saturated before it enters the slice (Fig. 2.3) (25,26). For example, to remove IVC signal from a renal artery origin study, one could place a transverse presaturation band across the abdomen inferior to the kidneys. Despite their usefulness, the application of presaturation pulses needs careful attention because their existence in a given acquisition is often not directly noted in the resultant images, if they are placed outside the imaging volume. Care must be taken to ensure that the vessel of interest is not similarly presaturated.

In generating an MR angiogram, parameter choices are often made that optimize one feature of the angiogram at the expense of others. In choosing optimum pulse sequence parameters, one needs to take into account the expected flow velocity, the thickness of the imaging volume that must
be traversed, and the acceptable level of contrast between flowing and stationary material.

The selection of the most suitable repetition time is a compromise. The shorter the TR the more saturated the signal from stationary material. However, as TR decreases, blood will receive more RF excitations in traversing the same distance than it would with a larger TR. Figure 2.4 illustrates this, which is a transverse 3D acquisition through the carotid bifurcation from a normal volunteer. The most suitable TR will depend on the rate at which blood transits a slice and will, therefore, vary from vessel to vessel and, for a given vessel, be somewhat patient dependent. One should also
note that an increased TR will dictate a larger total study time and hence, an increased possibility of patient motions, which can substantially degrade image quality.

Although significant progress has been made using conventional TOF methods, a series of innovative strategies have been pursued to further improve MRA. Specialized sequences have been proposed to reduce the well-known loss of signal noted at sites of disturbed flow. Other approaches have been pursued to enhance the contrast between flowing spins and stationary spins either by more effectively suppressing the signal from stationary spins or by retaining flow signal in the more distal portions of the vessels (as in the TONE approach noted above).