The principal considerations when designing a flow acquisition sequence are the speed of the fluid and the diameter of the vessel. Slowly flowing fluids (e.g. CSF or venous blood) are more difficult to image versus faster fluids (arterial blood). Smaller diameter structures (e.g. posterior communicating artery, aqueduct of sylvus) are more difficult to image versus larger structures (e.g. internal carotid artery). Secondary considerations that should be considered include tortuosity of the structures (straight versus not) nature of the flow (pulsatile versus not), and type of image desired (qualitative versus quantitative).
Phase contrast sequences are designed to encode from the velocity in a voxel to the phase of the image. This parameter (VENC) defines the maximum positive and negative velocities (encoded to positive and negative π). Faster velocities are easier to measure and slower velocities require stronger gradients on a MR scanner.
If a low VENC is selected, and the fluid flows faster than the VENC, then the image will ‘alias,’ and the true velocity will not be measured. Aliasing typically happens in the center of a vessel, where flow is typically the fastest. If aliasing happens, then it can sometimes be fixed during analysis by adjusting the encoded velocity.
If a high VENC is selected, and the fluid flows slower than the VENC, then the image will waste the range of measurements degrading the precision of the measurement.
To reconstruct the flow in the vessel of interest the image resolution needs to be adequate. The in-plane resolution (in mm) defines the size of the voxel. Flow in a vessel can either be a plug (turbulent) or laminar (laminar) flow profile. When assuming parabolic flow profiles, then an minimum of 3 voxels must be acquired across the lumen of the vessel. Higher resolutions (more, and smaller voxels) improve the estimates of flow. The smaller voxels will decrease the partial volume effect that occurs when averaging static tissue outside the vessel with moving fluid in the vessel. The smaller voxels also increase the number of datapoints that are used to estimate flow. If too high of a resolution is attempted, then the signal to noise ratio of the image will degrade the quality of the imaging.
When flows are through a straight and uniform structure, then the slice thickness can be increased. This has the advantage of increasing the volume in each voxel, which in turn increases the signal to noise ratio of the image. However, if the structure is not straight, or the flow through it is not uniform then the voxel will either average static tissue, or flows oriented in different directions.
When flow is pulsatile (e.g. arterial blood flow, aqueductal CSF flow) phase contrast imaging can assess the flow at multiple timepoints throughout the pulsatile cycle. In some cases, the temporal pattern of flow might contain useful information (e.g. forward versus backward flow through the aqueduct). In other cases, the information might not be useful and a single (static) image is preferable.
When these temporal patterns are needed, then the image sequence should capture multiple dynamic images. Typically, these are defined as dynamics per each heartbeat or each breath but other periodic signals might be relevant. Acquiring more dynamics will generally make the overall image longer to acquire. This is because the imaging will typically acquire only 1 dynamic image each period. Thus, if you desired 10 different dynamic images throughout a single heartbeat, then the scanner would need to wait for 10 heart beats to elapse. Note: this also means that if the temporal flow pattern changes during the acquisition, then the resulting images might be misleading.
A minimum (Nyquist-Shannon sampling theorem) of 2 samples per period must be acquired to reconstruct the basic shape of the temporal pattern. However, 10 or more per period are recommended. Carefully consider if other temporal aspects are important (e.g. flow but only during systole) and adjust the number of samples for that specific period. Alternatively, if you do not need to measure the temporal pattern, either because the flow is relatively constant, or because you wish to ignore the pattern, then no dynamic images need to be acquired.
Averaging allows data from multiple single acquisitions to be combined into one higher-quality image. One can use averaging with or without dynamic imaging.
Most of the focus in this website is on quantitative flow. For the best quantitative flow with phase contrast MRI, the images should be acquired perpendicular to the vessels of interest. This allows each voxel to measure the through-plane velocity, and thus the user to compute total flow. However, phase contrast images can be oriented so the vessel(s) of interest are in-plane. This makes quantification less accurate, because the in-plane measurements will often partially volume other (static) tissues outside the vessel. However, this approach might be useful for qualitative applications.