Visual Exploration of Simulated and Measured Blood Flow - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 293

25.3.2 Noise and Artifacts
Measurements come with imperfections which complicate the meaningful quanti-
tative analysis of the flow data. In particular, measures derived from the data are
sensitive to relatively small errors in the flowmeasurements. Inaccuracies are caused
by a combination of many factors, associated with the MRI hardware, imaging
sequences and their parametrization, and patient movement. For sensitive cardiac
applications, the generally accepted objective is to acquire flow data with less than
10% error [ 14 ].
The parametrization of the imaging sequence has a large influence on the accu-
racy of flow measurements. The parametrization directly influences the spatial and
temporal resolution [ 15 ]. In particular quantitative analysis of small vessels (e.g., in
the brain) becomes cumbersome at low resolution [ 2 ].
Besides user parametrization, motion is an important cause of imaging artifacts.
There are three major causes of tissue displacement due to patient movement: motion
artifacts by peristaltic motion, artifacts caused by contraction of the heart muscle, and
respiration. Contraction of the heart muscle artifacts can be considerably reduced dur-
ing acquisition. The impact of the respiratory motion can also be largely suppressed,
by exploiting the relatively motionless period after exhalation.
In addition, flowmeasurements are subject to general MRI artifacts, largely due to
hardware imperfections common in all MRI scans [ 4 ]. A relevant artifact for flow is
due to the fast gradient switching which induces eddy currents in the electromagnetic
field. This causes background phase errors in the image, which manifest as slowly
varying image gradients in both the spatial and temporal domains. These effects are
difficult to predict and therefore challenging to correct [ 14 , 38 ]. The conventional
MRI noise follows a Rician distribution. For flow imaging, it can be shown that
the noise in flow regions depends on the velocity encoding speed and is inversely
proportional to the SNR of the corresponding magnitude image [ 26 ]. Hence, the
VENC parameter should be chosen as small as possible, while capturing the full
dynamic range of the actual flow. The decision about the VENC value is often not
easy to make.
There are additional artifacts that are specific to flow data. For instance, aliasing,
or phase wrapping, erroneously introduces regions with opposite flow directions.
Whenever the actual blood flow speed transcends the VENC value, a phase wrap
occurs. Several methods have been devised to correct these artifacts caused by a
single phase wrap through postprocessing [ 22 , 54 ]. Another flow-specific artifact is
misregistration where blood flow regions are shifted from the stationary tissue. This
is due to the time between the phase encoding and frequency encoding gradients.
These artifacts can be corrected by adding a bipolar gradient to each phase encoding
gradient [ 51 ].
The flow imaging sequences are based on the assumption that the blood flow
velocities are constant at the time of measurement. Hence, measurements of acceler-
ated flows are less accurate and can cause undesirable artifacts. Accelerated flows can
be found in pulsatile flows, stenotic flows, or jets that can cause so-called flow voids