Occlusion of lenticulostriate arteries (LSAs) were reported to lead to lacunar infarction ^[1,2]. Currently, the major technique to image LSAs is digital subtraction angiography (DSA) for its high resolution and good definition of small vessels ^[3]. However, the clinical use of DSA is restricted by its invasiveness and simple image contrast. Recently, ultra-high field (7T) MRI scanner has achieved qualified images using time-of-flight (TOF) Magnetic resonance angiography (MRA) ^[4,5]. But LSAs with diameters range from approximately 0.3 to 0.7 mm ^[4] are difficult to be imaged on clinical MRI systems. Recently, one technique independent of in-flow effect, flow-sensitive black-blood (FSBB), was reported to image LSAs at 3T ^[6]. In this study, we aim to optimize the flow-dephasing gradients (FDG) in FSBB for LSAs imaging at 3T.

In FSBB sequence, one usually applies FDG in all three directions to achieve a greater gradient moment for more efficient blood suppression than in only one direction. However, the net direction of them is not aligned with the main flow direction of LSAs, thus it is not optimized. Since the gradients are applied to all vessels, the net direction should be aligned with the overall flow direction of the vessels interested, which can be estimated by an equation:$$V_i\sim\sum_k(|\sum_jv_{i,j,k}\cdot\triangle l_{j,k}|/\sum_j\triangle l_{j,k}),(i=x,y,z)$$We assume that one vessel can be divided into a group of unit vessels with single flow direction. $$$v_i$$$ denotes the overall flow direction vector projection along $$$i$$$ axis. $$$v_{i,j,k}$$$ denotes the projection of the vessel unit flow direction along $$$i$$$ axis. $$$\triangle l_{j,k}$$$ denotes the vessel unit length. $$$k$$$ denotes the index of vessels and $$$j$$$ denotes the vessel unit position in vessel $$$k$$$. Absolute value is used to prevent cancelation of opposite flow. Here, we assume LSAs structure of adults and their overall projection on x-y-z axes are similar. According to a cursory calculation, flow component of the foot to head (FH) direction is dominant. We apply the FDG only in FH direction in FSBB as an example (Figure 1). By this means, we can save nearly half of the scan time, due to the increased duty cycle when only one direction gradient is used.

In-vivo experiments were performed on a Philips 3.0T Achieva TX MRI scanner (Philips Healthcare, Best, The Netherlands) equipped with a 32-channel head coil. In the original FSBB sequence, the scan parameters were TR/TE=59/13ms, FA=14$$$^\circ$$$, FOV=200×190×50mm³, acquisition voxel size=0.5×0.5×1mm³, BW=144.4Hz/pixel, m1=1107ms²·mT/m (in the resultant gradient direction), scan time=8min50sec with SENSE factor=2. In the modified FSBB sequence, TR/TE=30/13ms, m1=639ms²·mT/m (in FH direction), scan time=8min56sec with SENSE factor=1, other parameters were the same with the original FSBB.

We acquired data in coronal and transverse planes respectively using FSBB and modified FSBB. Minimum intensity projection (mIP) was used for coronal slabs (Figure 2). The white arrows show the LSAs imaged which are more visualized in (a) and (c) than (b) and (d) because of higher SNR. (b) suffers much greater noise than (d) because SENSE was used in FH direction in which the coil sensitivity variation is limited. LSAs seem sharper in (a) than (c) due to its higher resolution after all the images reformatted to coronal plane. To avoid noisy images and keep scan time reasonable, the modified FSBB is proved effective.The results show it is feasible to image the LSAs, which are susceptible to noise, at 3T. A large net gradient moment can be achieved when applying the gradients in all three directions, but the m1 value is comparable between the original FSBB and the modified FSBB if we only consider about the LSAs. Small gradients will keep TR short so that SENSE is not essential. Compared with the original FSBB using SENSE=2, it can be proved that the SNR, which is a key point, is still improved for the modified FSBB although its TR is short. So applying resultant gradient in the overall flow direction of the LSAs will make FSBB more efficient in SNR. However, there are still a few LSAs that can only be seen in the original FSBB (yellow arrows). The reasons may be these vessels are perpendicular to the FH direction or they are almost parallel to the resultant gradient direction in original FSBB with extremely slow flow. A flow dephasing gradients optimization algorithm for targeting imaging vessels is proposed in this work. The in-vivo results prove the optimized flow dephasing gradients can improve the image SNR while maintaining the blood suppression efficiency. This technique makes FSBB more practical and robust in LSAs imaging at 3T for clinical applications.