PET imaging combined with MRI has emerged as a new and promising molecular imaging platform. While PET/MR benefits from the rich soft tissue contrast of MRI, it remains challenging to obtain a reliable photon attenuation correction map necessary for accurate PET quantitation due to MRI’s limitations in imaging bone (which has high photon attenuation relative to soft tissues). Many recent approaches for MR-based attenuation correction (MRAC) have proposed ultrashort echo time (UTE) imaging schemes due to its capability to resolve objects with short T2* decay (i.e., bone). Unfortunately, UTE techniques are not typically rapid (e.g., several minutes of acquisition time, particularly for multi-echo approaches), and are thus likely to impede PET/MR workflow (especially for whole body PET/MR where only 3-5 minutes may be available for MRI at each bed position). In this study, we explored a new scheme for MRAC that achieves rapid acquisition (<35 seconds) with improved short T2* signal estimation. Conventional frequency encoded UTE (FE-UTE) and ramped hybrid encoding (RHE)¹ with G_RF=0% were explored as imaging schemes for attenuation correction. In FE-UTE, k-space is measured using a center-out radial frequency encoding gradient that is applied after RF coil deadtime as shown in Figure 1-a. In RHE, the readout gradient is applied before RF coil deadtime as seen in Figure 1-b, and the k-space data missing during RF coil deadtime is measured by single point imaging (SPI). In this study, each acquisition is extended with additional gradient echoes to secure multiple images at later TEs. Note that two images before and after each gradient echo can be obtained in RHE owing to the SPI-encoded region. Both acquisitions utilized a gradient measurement technique similar to that proposed in Reference 1. Data was acquired in a human volunteer (IRB approved study) using a 3T PET/MR system (GE Signa PET/MR) with the following scan parameters: 8µs hard RF pulse (flip angle=2 degree), TR=3.6ms, Gmax or Greadout=14.85mT/m, and G_RF=0%. In RHE, the number of frequency and SPI encodings were 5,000 and 4,145 (diameter of SPI encoding=21), respectively, and the total scan time was 33sec. In FE-UTE, the number of frequency encodings was 5,000 or 8,878 with scan times=18sec or 32sec. In FE-UTE four images were obtained at TE=80, 834, 1500, and 2200µs, while in RHE seven images were obtained at TE=80, 728, 778, 1420, 1470, 2112, and 2162µs. All images were reconstructed with a spatial resolution of 3x3x3mm using convolution gridding and iterative density compensation. Using reconstructed images from later TEs excluding the UTE image, IDEAL² was performed to obtain fat and water separated images, and a short T2* image was synthesized by subtracting fat and water images from the UTE image.Figure 2 shows fat image, water image, fat fraction, and synthesized short T2* image obtained using RHE, FE-UTE with 5,000 radial spokes, and FE-UTE with 8,978 radial spokes. In the brain, fat fraction was measured as 2.2 ± 1.1%, 11.8 ± 2.1%, and 12.6 ± 3.7% respectively for RHE, FE-UTE with 5,000 spokes, and FE-UTE with 8,878 spokes. Note that RHE allows better fat and water separation compared with FE-UTE as shown in the fat fraction map where imperfect fat and water separation is seen as a speckled artifact, which is also propagated into short T2* images. Figure 3 shows a resultant pseudo CT images generated using fat, water, and short T2* images. Both RHE and FE-UTE show reliable image segmentation, while mis-segmented, small specks are still shown in the computed pseudo CT images in FE-UTE. Rapid multi-echo UTE imaging was demonstrated with clinically feasible scan times for whole-body PET/MR (18~33 sec). The ability of IDEAL to compensate for intravoxel partial volume effects enabled fat and water separation at lower than conventional spatial resolutions for clinical MRI (3x3x3mm). Furthermore, the use of a subtractive UTE-IDEAL signal model enabled computation of a short T2* image via subtraction of long T2* water and fat components inherently corrected for the intravoxel partial volume effects of each component (short T2*, water, and fat). Note that MRAC is typically applied at the native resolution of the PET system, and is generally lower (e.g., 4.5x4.5x4.5mm) than that used here, thus supporting the use of rapid multi-echo UTE imaging in this scheme. Finally, two UTE methods were compared, where RHE showed better performance at fat and water separation (likely due to the SPI-encoded central k-space region that allows time-constant phase at TE and reduced eddy current effects). The improved performance of RHE-IDEAL was propagated to the computed pseudo CT images.