The periodically rotated overlapping parallel lines with enhanced reconstruction (Propeller) technique is a useful acquisition scheme and reconstruction method to reduce motion artifact (1). The main strategy of Propeller acquisition scheme is to cover k-space with different rotating blades, and each blade consists of a set of parallel k-space lines. It can be combined with different imaging methods, such as fast-spin-echo (1) or gradient-echo imaging (2), to fulfill different clinical applications. Recent studies showed that Propeller technique can substantially reduce the motion problem in both brain and body acquisitions. However, the higher specific absorption rate (SAR) of RF pulse at high-field magnetic strength can limit the number of multiple slices for a given TR, which in turn reduces the efficiency of acquisition, especially for Propeller-FSE sequence. A possible solution is to reduce the echo-train-length with under-sampling of data of each blade. Afterward, the parallel imaging, such as sensitivity encoding (SENSE), can be used to solve the aliased image before Propeller reconstruction (3). The main issue of parallel imaging is undesired noise amplification during matrix inversion, especially with high acceleration factor (e.g., R = 4). Projection onto convex sets based multiplexed sensitivity encoding (POCSMUSE) reconstruction method has been shown to be useful in reduction of motion-related artifact in multi-shot acquisition with less noise amplification compared to conventional SENSE reconstruction (4). In this study, we propose to apply POCSMUSE instead of SENSE reconstruction and Propeller reconstruction, to reconstruct the image from all under-sampled blade data with reduced noise amplification. The data acquired from brain and liver using Propeller-GRE will be used to test purposed POCSMUSE algorithm.POCSMUSE reconstruction is a general algorithm that can be applied to either Cartesian or non-Cartesian k-space trajectory (4). For the under-sampled Propeller data, the SENSE reconstruction can be applied to each blade data to solve the aliasing prior to Propeller reconstruction. Figure 1 shows a 28 x 288 single blade image with high acceleration factor R = 4 (7 of 28 central k-lines acquired) obtained from SENSE reconstruction that can be clearly observed the undesired noise amplification at the center of image. To reduce the undesired noise amplification, the POCSMUSE algorithm method has been modified to the framework shown in Figure 2a to accommodate multiple blade data. First, the motion-induced and rotation-induced phase variation is estimated by using POCSENSE method for each blade data, and then follows a low-pass filtering. Afterward, an initial guess of source image (P0) was generated from direction weighted-average of all under-sampled blade data with demodulation of both phase variation and coil sensitivity (Fig. 3). Figure 2b shows a feasibility of bulk motion correction incorporated into POCSMUSE framework for Propeller data set. The bulk rotation and translation of each blade can be firstly estimated from POCSENSE produced single blade image. The proposed reconstruction method was evaluated with brain and liver Propeller-GRE data obtain from a 1.5T MRI scanner (GE Healthcare): FOV = 24cm/40cm (for brain/liver), blade size = 28*288, TE = 40ms/20ms (for brain/liver), TR = 50ms/25ms (for brain/liver), acceleration factor R = 4/2 (for brain/liver), rotating angle = 12°, and 16 blades for 180° k-space coverage.Figure 4 shows the brain images reconstructed from fully-sampled data with Propeller reconstruction (Fig. 4a), under-sampled data with SENSE and Propeller reconstruction (Fig. 4b), and under-sampled data with POCSMUSE reconstruction (Fig. 4c). Figures 4a-4c show the POCSMUSE reconstructed liver images acquired under free-breathing, breathing-hold, and respiratory-triggering. The acquisition time of each reconstructed image are also showed at below of image.Our study demonstrates that the POCSMUSE can be used to reconstruct under-sampled Propeller data set with reduced noise amplification compared with conventional SENSE reconstruction prior to Propeller reconstruction. Although only Propeller-GRE data are presented in this study, the reconstruction of under-sampled Propeller-FSE data should be comparable. It is particularly useful for implementing Propeller-FSE at high magnetic field because the under-sampling of each blade data can substantially reduce SAR accumulation, thereby improving the acquisition efficiency. In conclusion, POCSMUSE can successfully reconstruct under-sampled Propeller data set with ability of motion correction and reduced noise amplification.