Multiband (MB) Multislab (MS) 3D and 2D TOF MRA can reduce acquisition time with high inflow contrast ⁽¹⁾. However to unaliase overlapped slices in MB imaging, coil sensitivity profiles are commonly used that can cause g-factor penalty. In this abstract we explore utilizing the sparse nature of angiographic data to reconstruct blood vessels in 3D and 2D angiography without using standard parallel imaging techniques. To do this we explore the idea of using CAIPIRINHA⁽⁴⁾ to shift each slice/slab differently along the phase encoding direction and then reshift the slab to its original position to form a contiguous vessel tree that can be reconstructed using a clustering algorithm.Data from four healthy subjects, after informed consent, were acquired on 3T Tim Trio systems (Siemens Healthcare, Erlangen Germany) equipped with 32 channel head coil. The 3D MS TOF sequence with 3 slabs, 67 slices per slab, TR=20 ms, TE=3.59 ms, FA=18(degrees), and matrix 384x288. 2D and 3D MB images were simulated offline by adding the slices/slabs together depending upon the multiband factor using MATLAB (MATLAB 7.12).CAIPIRINHA is used to shift each slab by different amount before adding them together to avoid complete overlap of vessels. To reconstruct a continuous vessel from the CAIPI shifted MB data, slabs are concatenated along the z direction, according to the MB factor, and then CAIPI reshifting is used. In the CAIPI reshifting process, each concatenated slab is shifted back to ensure that the slice in that position is in its correct place (see Figure 1b).Subsequently, the largest contiguous vessel trees in this re-concatenated overlapped space are searched, for example that connecting red, green and blue in figure 1. In 3D MS MB TOF this idea was implemented with (MB=3), (Nex=1), CAIPI shifts (0,Fov/2, Fov/3) to each slab as shown in (Figure 1).In 2D TOF, multiple averages are acquired to achieve adequate SNR, and this feature can be utilized to employ different CAIPI shifts for each measurement. Combining all the averages with different shifts results in a reinforced image of the true vessel tree as shown in (Figure 2(b)).2D MB TOF MRA was simulated using (MB=8), (Nex=4), CAIPI(Fov/2, Fov/3, Fov/8, Fov/12) as shown in (Figure 3) with single average.Concatenated and reshifted data from 2D and 3D TOF MRA were filtered using a Frangi filter. Frangi filter is a vessel enhancement filtering process that search for geometrical structures and provides a measure of ‘vesselness’ (V) based on the eigenvalues of the Hessian matrix⁽²⁾. Breadth first search algorithm was used to reconstruct blood vessels from Frangi filtered data. In graph theory, breath first search⁽³⁾ is used for exploring a graph such that in a graph G it start at node s, it proceed by exploring all its neighbors to find all the nodes in a graph G for which there is a path from s. It can be seen from (Figure 1) that slab concatenation with CAIPI reshifting helps in identifying the largest continuous vessel tree. Vessel visibility is enhanced and background tissue is suppressed using Frangi filter as shown in (Figure 2). The reconstructed vessels from Frangi filtered data using breath first search is shown in (Figure 2, (b)). As seen from the (Figure 2) reconstruction of blood vessels in 2D and 3D is comparable to single band standard 3D TOF MRA. In 3D MB results showed that better separation of vessels is reached if high CAIPI shifts are given to the initial slab and low CAIPI shifts to the last slab that mostly contains the branching vessels. In 2D highest MB factor used is (MB=8) with unique CAIPI shifts applied to each average. Applying a unique CAIPI shift to each acquisition results in a number of shifted slices. Reshifting each acquisition and averaging them reinforces the vessel structure (Figure 3).In MB MS 3D TOF angiography saturation of stationary spins due magnetization effects can decrease sensitivity variation in slice direction that can lead to g-noise enhancement.Vessel reconstruction in MB angiography without using coil sensitivity profile can lead to high MB factors reducing the aquistion time and high sensitivity in detecting small vessels.

We have demonstarted the feasibility of vessel reconstrcution in MS MB TOF MRA witout using standard parallel imaging methods.It can be seen from (Figure 2) that vessels reconstructed from Single band standard 3D MS TOF without MB and CAIPI shifts is comparable to 2D MB (MB=8) and 3D MB (MB=3). Further work would be to assess robustness in vessel detectability on high multiband factors.