4D Flow (Time-resolved 3D phase-contrast MRI) has demonstrated potential for assessment of hemodynamic-related cardiovascular pathologies in the entire chest [1-3] and possesses the advantage of reformatable volumetric imaging of flow in all directions compared to the conventional 2D technique. However, its whole volume acquisition saturates blood signal and generates dark blood images. Typically, contrast agent is needed for reliable flow measurements, creating workflow difficulties with other post-contrast sequences (e.g. MRA, perfusion and LGE) and making the sequence inaccessible to patients with renal insufficiency. Also, due to its free-breathing acquisition, 4D flow is sensitive to residual respiratory motion artifacts. In this work, we investigated the feasibility of a multiple slab acquisition scheme for bright-blood motion-robust 4D Flow without contrast agents.

A conventional Cartesian 4D Flow sequence [4] was modified as shown in Fig. 1 for this study.

1. Multiple thin slab (MSLAB) [5] acquisition: whole-chest imaging is divided into acquisition of multiple thin axial slabs. The thin slabs are acquired sequentially from the most inferior to the most superior with small overlaps (shaded regions) between adjacent slabs for volume merging. Though blood flow in the chest wall region is not unidirectional, the overall flow is primarily along superior-inferior (SI) in the blood pool and vessels of interest (e.g. intra-ventricular flow and flow along aorta and main pulmonary arteries). Thus, each thin axial slab is constantly replenished with unsaturated blood during acquisition and the resulting bright blood feature should provide higher SNR and reduced sensitivity to residual motion artifacts.

2. Variable density (VD) radial sampling: repeated acquisition is used to suppress respiratory motion and k-t sampling is used for spatiotemporal correlation-based reconstruction. For scan efficiency, a VD sampling scheme is used with linearly decreasing sampling rate and NEX (number of excitation) from the highest at center k-space toward outer k-space. The k-space is acquired with a radial vieworder in ky-kz using golden angle increments. Such a vieworder revisits central k-space repeatedly during the acquisition of each slab. Therefore, the effect of any heart rate variation or respiratory drift is spread in the entire k-space of each slab and propagates smoothly across slabs, which should yield continuous anatomy and flow in the whole volume without slab misregistraton of a conventional linear vieworder.

In reconstruction, first, each slab is processed separately using kat ARC – a k-t auto-calibrating parallel imaging method with cardiac motion adaptive temporal window selection [6]. Then, the reconstructed slabs are merged into a whole-chest volume and the overlapping slices are averaged with weights based on their off slab-center distance.

3 healthy volunteers were scanned without contrast agent on GE 3T (MR750, Waukesha, WI) using both the MSLAB and the conventional 1-slab (1SLAB) sequences. Typical imaging parameters were: 380×260 mm2 FOV, 2.2×2.2mm2 resolution, 2.4mm slice thickness, VPS of 3, 8× acceleration and NEX of 4 on 10% central k-space data. The entire chest is covered by a single slab with 68 slices in ~7min for the conventional sequence and by 4 thin slabs (20 slices in ~2.5min / slab) with 4 overlapping slices for the modified sequence. The reconstructed images are processed on Arterys (San Francisco, CA).

MSLAB provides smooth anatomy and flow on all volunteers without slab misregistration. As shown in Fig.1, conventional 4D Flow generates dark blood with apparent motion artifacts at aortic (AO) root and pulmonary artery (PA) even at systole with peak flow, while MSLAB provides bright blood anatomy without visible motion artifacts. The velocity maps reveal that 1SLAB suffers from low SNR, motion artifacts (solid arrows) and missing vessel depiction (dotted arrows). In comparison, MSLAB provides more complete vessel depiction with higher quality flow. Slab boundaries introduce visible shading across slabs on reformatted MSLAB anatomy, but do not impact phase-difference-based flow maps. Table 1 shows the blood enhancement of MSLAB over 1SLAB at aortic arch (AO_a) and descending aorta (AO_d), PA and left ventricle (LV). The highest blood enhancement is obtained at cardiac phases with peak flow, while significant enhancement is also observed at low flow.This work demonstrated a new non-contrast 4D flow method. Our preliminary results show that MSLAB can take advantage of the dominant SI flow and generate high quality whole-chest 4D Flow with improved SNR and reduced sensitivity to residual motion. The proposed VD radial acquisition provides high acquisition efficiency and robustness to cardiac and respiratory variations.