Purpose

Time resolved phase contrast MRI (Time-resolved PCMRI)¹ has been extensively used to evaluate vascular function such as 4D flow². The potential of this technique includes providing dynamic physiological information and morphology of the imaged vessels. While Time-resolved PCMRI has been successful in monitoring dynamics of cardiovascular system, its application to cerebral vessels as Time-resolved PCMRA and Pulse wave velocity (PWV) imaging is still challenging due to the requirement of high temporal and spatial resolution and the inherited long scan time. Although ultrasound also provides 2D illustration of PWV³, the imaging region is limited to the anatomy where ultrasonic wave can go through. Henceforth, an MR-based alternative should be favorable when deeper brain regions are of interest. The present work aims at developing an acceleration strategy for Time-resolved PCMRA covering from carotid arteries to its downstreams for PWV evaluation. Current implementation involves the two strategies, MUSE⁴ and hybrid SENSE⁵ in a hope to retain morphological and dynamic information at good quality within short scan time.

Methods

Four healthy subjects were recruited and scanned following informed consent, using an IRB-approved protocol on a 3.0T scanner (GE Discovery MR750). After scout scans to locate the orientation of carotid arteries, the proposed fast Time-Resolved PC MRI was performed with eight 7-mm slices (FOV = 26cm within a 256x256 matrix) to encompass carotid arteries and its downstream branches. The other imaging parameters are as follows: TR=5.0ms, TE=2.8ms , View per segment=1, and Venc=180cm/s along three orthogonal directions leading to temporal resolution equal to 20ms. A variable density sampling function (Fig. 1) covering the central quarter of k-space at two-fold acceleration and the rest at four-fold acceleration was used. The total scan time for a single slice required 80 cardiac cycles. Despite heart rate variability for each subject and different heart rates among subjects, the scans were all done within 10 minutes. Photoplethysmogram(PPG) was used to monitor pulses for cardiac gating and for gradient encoding synchronization to achieve fully gating reconstruction. The data without velocity encoding were firstly averaged to remove artifacts for coil sensitivity estimation. The acquired data were processed by a two pass reconstruction to improve SNR and to retain temporal accuracy as suggested in the reference algorithms^4-6. The first-pass reconstruction was performed using SENSE on all the acquired data with prospective cardiac gating. A spatial low-pass filter was applied after the reconstruction to improve SNR and to serve as the regularization information in the second-pass reconstruction. Subsequently, the original undersampled data were retrospectively regridded in terms of cardiac phase and processed by the algorithm expressed by:$$ \widehat{\rho}_{y,f,\overrightarrow{v}}=argmin_{\rho_{y,f,\overrightarrow{v}}}||s_{k,t,\overrightarrow{v}}-E_{k,t} \rho_{y,f,\overrightarrow{v}}||^2_2 - \lambda||M^{-1}_{y,f,\overrightarrow{v}}\rho_{y,f,\overrightarrow{v}}||^2_2 $$, where $$$E_{k,t}$$$ represents the encoding matrix including Fourier transformation from k-t space, the sampling trajectory, the density compensation and the coil sensitivity, $$$M^{-1}_{y,f,\overrightarrow{v}}$$$ stands for the regularization information from the filtered first-pass reconstruction with its weight $$$\lambda$$$ to minimize noise amplification, $$$\rho_{y,f,\overrightarrow{v}}$$$ denotes the final reconstruction and $$$s_{k,t,\overrightarrow{v}}$$$ is the input rawdata. The number of cardiac phases in retrospectively-gating reconstruction was set to be 80. For the visualization purpose and the analysis of PWV, $$$\rho_{y,f,\overrightarrow{v}}$$$ is then transformed back to time domain to calculate flow speed and direction multiplied by the magnitude mask to enhance temporal vascular signals.

Results

Fig.2 shows the temporal flow speed changes from subject 1. Enhancement can be found starting from the common carotid arteries (CCA) followed by the internal carotid arteries (ICA) then to the middle cerebral arteries (MCA). The animated version of the full time series is shown in Fig. 3 on the on-line version. Figure 4a shows a spatial-temporal Maximum Intensity Projection (MIP) result using data from subject 2. Temporal signals from the three ROIs, CCA, ICA and MCA, are shown in Figure 4b. Significant delay among different vascular segments can be noticed and due to high temporal resolution, delay time from the adjacent ROIs of ICA and MCA could still be resolved.

Discussions and Conclusion

The results suggest that to resolve small vessels and the latency of pulsatile flow in the cerebral arteries required high spatial and temporal resolution. If no acceleration strategy is involved in the sequence, it will take 256 cardiac cycles to achieve similar spatial and temporal resolution for scanning one slices, roughly around 4 minutes. The volumetric scan using proposed method has brought the scan time down to 10 minute. According to Figs 2 and 4, not only carotid arteries and its downstreams but also the vertebral arteries and the basilar arteries can be enhanced. Thus the present strategy could have the potential to help monitor cerebral PWV as a risk factor of cerebral vascular stiffness⁷.

Acknowledgements

Acknowledgements: Support is acknowledged from MOST grants:105-2314-B-006 -044 –MY2, 105-2314-B-007-003, and 105-2221-E-035 -049 -MY2 and Mind Research and Imaging Center in National Cheng-Kung University.