Purpose

4D Phase-Contrast MRI (PC-MRI) is mainly performed in vessels with an important diameter, such as the aorta¹. Only a few studies show its potential when used in intracranial vessels² or arteries from the head and neck area, since their study implies a good spatial resolution, resulting in a long acquisition time. The application of an acceleration factor, using Principal Component Analysis (k-t PCA) can help solving this problem. Based on under-sampling the k-space, PCA reduces highly dimensional data into lower dimensions, using correlations within the data³. The main purpose of this study is to assess the influence of k-t PCA on the flow measurement of a phantom composed by branches with small diameters.

Methods

This study is performed with a MRI compatible Plexiglas phantom, composed by 6 branches with diameters similar to those found on the face and neck area, Figure 1. A pump (Masterflex), connected by pipes, provides a pulsatile flow to the phantom with a frequency selected at 100 cycles/min for an average flow rate of 87 ml/min. Before MRI acquisitions, an accurate flow sensor (Transonic System Inc.) is placed before the phantom input to measure the flow curve with high accuracy. The sensor is then placed in the same conditions in the 3T, Achieva dStream, Philips MR system to perform the 4D PC-MRI acquisitions. Sequences are acquired using a 32 head coils channels in the axial plane with the following parameters: FOV=50x50mm², TE/TR=3/8ms, spatial resolution=0.6x0.6x0.6mm³, 1 repetition, SENSE=0, 140 slices, and 16 frames for each cycle. The velocity encoding is selected at 50cm/s in all three flow directions, above the maximum velocity observed in the phantom. The sequence is repeated 4 times using a k-t PCA acceleration factor set at 0, 2, 4 and 8 (factor of the under-sampling of the k-space). A homemade software has been developed to process images from 4D PC-MRI sequences, using a semi-automatic segmentation, based on a Fast Fourier Transform⁴, to reconstruct the acquired volume. A slice is manually placed on the volume reconstructed from the 4D PC-MRI images to rebuild a 2D slice of the segmented branch, Figure 1. 4D flow data analysis included corrections for background phase error and velocity aliasing. For each branch of the phantom (except the input branch, since it is too far from the coil to have enough signal), the flow curve is reconstructed along the pulsatile cycle. This operation is repeated 3 times for 3 different slice positions on the same branch. A mean flow curve is then calculated for each branch. Accuracy of the flow measurement is evaluated by calculating the difference between the expected and the measured flow in each branch. The difference between the values is normalized in percent of the expected flow rate. The amplitude of each flow curve is measured by calculating the difference between the maximum and minimum peaks of the flow curve.

Results

The mean flow curves for each value of acceleration factor are shown in Figure 2 and the evolution of the amplitude obtained in each branch for each value of acceleration factor is shown in Figure 3. Table 1 shows the percent error of mean flow and the amplitude calculated in each branch as function of the acceleration factor.

Discussion

For each acceleration factor the percent error is less than 5%. Therefore, it is possible to divide by 15 the acquisition time and still maintain a good measurement accuracy. However, the results also show a decrease in the amplitude of the flow curves by 25% using the acceleration factor 8. These results are consistent with those found by Greil et al.⁵ in great vessels. This result can be explained by the method used to under-sample the k-space, which mainly samples the high frequencies³: if mean flow is kept constant, when the under-sampling factor increases, the amplitude of the signal decreases. This decrease can be more significant in high frequency signals, such as from arterial blood flow. In fact, this could reflect a possible limitation of the k-t PCA acceleration factor when it is used to calculate the Wall Shear Stress, which uses amplitude as the main parameter to determine its value.

Conclusion

In an experimental setup providing physiological pulsatile flow with frequency and average flow rate that can be found in human beings, it is possible to obtain accurate mean flow measurement in vessels with diameters of one millimeter with an acquisition time clinically acceptable. However, the k-t PCA acceleration factor causes loss of information for peaks detection, resulting in a decrease of the amplitude of the flow curve.

Acknowledgements

No acknowledgement found.