Time-resolved 3D quantitative flow MRI of the major intracranial vessels: initial experience and comparative evaluation at 1.5T and 3.0T in combination with parallel imaging

Abstract

Exact knowledge of blood flow characteristics in the major cerebral vessels is of great relevance for diagnosing cerebrovascular abnormalities. This involves the assessment of hemodynamically critical areas as well as the derivation of biomechanical parameters such as wall shear stress and pressure gradients. A time-resolved, 3D phase-contrast (PC) MRI method using parallel imaging was implemented to measure blood flow in three dimensions at multiple instances over the cardiac cycle. The 4D velocity data obtained from 14 healthy volunteers were used to investigate dynamic blood flow with the use of multiplanar reformatting, 3D streamlines, and 4D particle tracing. In addition, the effects of magnetic field strength, parallel imaging, and temporal resolution on the data were investigated in a comparative evaluation at 1.5T and 3T using three different parallel imaging reduction factors and three different temporal resolutions in eight of the 14 subjects. Studies were consistently performed faster at 3T than at 1.5T because of better parallel imaging performance. A high temporal resolution (65 ms) was required to follow dynamic processes in the intracranial vessels. The 4D flow measurements provided a high degree of vascular conspicuity. Time-resolved streamline analysis provided features that have not been reported previously for the intracranial vasculature.

FIG. 1

Schematics of 4D flow pulse sequence with retrospective gating. The four flow-encoding schemes (FE1…FE4) represent the innermost loop structure (i.e., the flow-encoding unit (shaded bars and insert)) and determine the minimum possible temporal resolution (i.e., 4 × TR). After the four flow-encoded echoes are acquired, the phase-encode gradient along the slice dimension is altered and the acquisition of another flow-encoding quadruple is performed. In this example, four (N_U) different phase-encode steps (k_z1, …, k_z4), are played along the slice dimension, leading to an intrinsic temporal resolution of ΔT = N_U × 4 × TR. This set of N_U phase- and flow-encode steps is repeated over the entire R-R interval. The acquisition of other section-direction phase encodes (e.g., k_z5, …, k_z8) is analogous to this scheme. After completion of all required section-direction phase-encode steps, N_kz, the sequence changes the in-plane phase-encode gradient and the acquisition cycle starts all over again. Depending on the instance when the cardiac trigger occurs, the acquired k-space data will be assigned to different cardiac cycles.

FIG. 2

Streamline visualization of blood flow. Blood flow at four different cardiac phases is shown through the carotid siphon (curved arrows) and cavernous segment (asterisk) of the right ICA and through the bifurcation of the ICA into the MCA (arrows) and the ACA (open arrow). Top row: systolic phase; bottom row: end-diastolic phase. Red and yellow streamline colors reflect high velocities, and green and blue colors reflect slower velocities. During the systolic phase, the tailing edge of the pulse wave can be very well appreciated as it travels distally toward the MCA and ACA. Both proximal and distal to the ACA/MCA bifurcation the velocity also remains increased in the end-diastolic phase due to the vessel caliber reduction. In all phases the twisting of the streamlines can be appreciated. No retrograde flow is apparent.

FIG. 3

Flow pattern in the BA and the following distal vessels. The streamlines running into the left PCA and SCA are shown in yellow and red, respectively. The streamlines branching into the right PCA and SCA are shown in green and blue, respectively. As in the ICA, the blood flow within the BA demonstrates a helical pattern as it moves distally. Blood entering the right cerebellar artery (blue) branches off at a more acute angle than that entering the left cerebellar artery.

FIG. 4

Incidental finding of a type 15 variation (De Almeida classification) of the ACAs and AcomA in one volunteer. The left A1 segment of the ACA has a smaller caliber than the corresponding contralateral vessel. a: Streamline visualization of the A1/A2 segments of both ACAs and the AcomA (arrow) reveals reduced velocities in the A1 segment of the left ACA, and suggests that the left A2 segment (asterisk) is predominantly fed through the right ACA via the AcomA. This can be much better appreciated by the color-coded streamlines coming from the left side in blue and from the right side in red. b and c: Streamlines in the end-diastolic phase obtained from two different view angles. While there is significant blood flow through the right A2 segment (open arrow), only mild flow can be seen in the left A2 segment. The increased perfusion pressure in the systolic phase (d and e) allows blood flowing through the AcomA and the left A1 to supply the left A2 segment.

FIG. 5

a: Comparison of 4D flow imaging using different parallel imaging acceleration factors at 1.5T and 3T. From top to bottom 4D flow measurements are shown with GRAPPA ORFs of 1 (row 1), 2 (row 2), and 3 (row 3). The slice was placed at the level of the Circle of Willis (COW). Parts of the MCAs (M1 segment, arrows) and PCAs (open arrows) can be seen. The two leftmost columns were obtained at 1.5T and show the magnitude image (left) and velocity image (right) for one particular instance within the cardiac cycle. High velocities are encoded in red and yellow, and slow velocities are encoded in green and blue (see legend). The two rightmost columns are corresponding slices from the same subject scanned at 3T. The increased baseline SNR at 3T is clearly apparent. The 3T scans have a substantially higher vessel contrast-to-noise ratio (CNR), which can be seen best on the velocity images for ORF 1 and 2. For ORF = 3 the parallel imaging related noise enhancement at 3T is comparable to that of ORF = 2 at 1.5T. Both the magnitude image and speed image for ORF = 3 at 1.5T demonstrate considerable noise enhancement. In this case the magnitude drops so dramatically that the threshold procedure eliminates a large number of pixel around the sella region, pons, vermis, and cerebellum. Nonzero mean background noise is apparent especially for all of the 1.5T velocity images and the 3T ORF = 2 and 3 images, which is due to the magnitude operations on the phase data when velocity components are computed. b: The slice shows both carotid siphons (asterisks) and the cavernous segments of the ICAs (arrows). Further, the BA (open arrow) and a small segment of the SCA (curved arrow) can be very well delineated across all three reduction factors. In all images there is excellent vessel CNR for the ICAs. With increasing ORFs the smaller vessels become less conspicuous. Similarly to the slice through the COW at 1.5T, the magnitude and velocity maps start to deteriorate when moving from ORF = 2 to ORF = 3. Especially with ORF = 3 at 1.5T, a significant SNR loss can be noticed in the mid-sagittal line (e.g., throughout the cerebellum). An increase in baseline noise in the cerebellar peduncles at 3T with ORF = 3 is very well visualized in the velocity maps and can be appreciated as a change from blue to a more greenish color in this particular region. Note that this noise level is roughly comparable to the ORF = 1 scan at 1.5T. As in image a at 1.5T with ORF = 3, many pixels are eliminated by the thresholding procedure. c: A magnified area of image a shows the left and right carotid canals, the BA, and the right SCA. The two leftmost columns show the 1.5T data, while the two columns on the right show the corresponding 3T data. From top to bottom: ORF = 1, ORF = 2, and ORF = 3. The tissue surrounding the vessels clearly reflects the expected SNR loss associated with increasing ORF. However, the intravascular signal still remains significantly above the noise floor, as can be seen very well in the velocity maps due to the high baseline signal of blood.

FIG. 6

Comparison of blood velocity measurements in different subjects. Velocity measurements in the BA (left) and left ICA (right) performed at 1.5T (top row) and 3T (middle row) are shown for five subjects. For each subject the same color was used for all plots. The velocity difference between 3T and 1.5T measurements is shown at the bottom row. The average difference in velocity measured over the entire cardiac cycle was approximately 8.5 ± 22.5 mm/s. From these waveforms it can be very well appreciated that the intersubject variation of velocity measurements is significantly higher than the intrasubject differences between 1.5T and 3T. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

FIG. 7

Effect of parallel imaging acceleration on the measurement of velocities in the BA over the cardiac cycle observed in three volunteers at 1.5T and 3T. Each row shows the time course of velocity changes over the cardiac cycle in a different subject imaged at 1.5T (left) and 3T (right). Note that there are only small differences between the time course of ORF = 1 (bold black), ORF = 2 (bold gray), and ORF = 3 (thin gray) data at 3T. Somewhat more considerable deviations can be seen at 1.5T for ORF = 2. These deviations become substantial for ORF = 3 data. In the presence of such significant fluctuations, fine details in the time course, such as the dicrotic notch (arrow), which is associated with the end-systolic closure of the aortic valve, are clearly missed. Note that the time course does not start at peak velocity because the pulse wave requires longer time to propagate to the fingers (from which the cardiac trigger is derived) than to the brain.

FIG. 8

Effect of temporal resolution on the measurement of velocities over the cardiac cycle. The time course of velocity changes over the cardiac cycle is plotted for 1.5T (left) and 3T (right) in the same volunteer. The top row shows the time course for the BA, and the bottom row shows the time course for the right ICA. The fully sampled data sets (bold black) are subsampled by a factor of 2 (bold gray) or 3 (thin black) and interpolated to match the temporal resolution of the fully sampled scans. While there is little deviation between the fully sampled and subsampled data for slow signal changes, the lack of temporal resolution makes it difficult to follow rapid changes in velocity, leads to a significant degradation of the velocity time course, and directly affects related parameters, such as the PI and RI. The reduced temporal resolution also makes it difficult to detect the dicrotic notch (arrow).