One of the challenges of PET/MR scanners in generating accurate attenuation maps, which is necessary for PET image reconstruction, is bone tissue identification [1]. A new bone imaging technique, based on proton density-weighted, single-echo zero-echo time (ZTE) acquisition, has been recently developed for PET/MR attenuation correction (AC) [2]. Here we have compared head atlas-based attenuation correction with ZTE-based attenuation correction in an ¹⁵O-water brain study. We compared the tracer activity in voxels close to the skull and those in deep brain to study the effect of bone tissue identification on standardized uptake value (SUV) of the tracer.The study was performed in compliance with regulations of the local Institutional Review Board and all subjects were consented prior to the study. Six subjects were injected with 925 MBq of ¹⁵O-water and scanned on Signa PET/MR scanner (GE-Healthcare, Waukesha, WI, USA) in order to compare CBF measurement by arterial spin labeling (ASL) and PET [3]. In addition to the scanner default Dixon-based fat/water imaging (18 s scan-time) used for attenuation correction, a whole brain 3D ZTE scan was acquired using the following parameters: FOV 24cm, ST 1mm, 338 slices, FA 1°, frequency 192, 2 NEX, BW 62.5 kHz and 2:37 scan-time. The scanner’s default PET image reconstruction software uses the Dixon fat/water images to register a head atlas that estimates the skull position for the AC map. A new attenuation map was generated using the ZTE images by direct segmentation of bony structures [4]. PET images were reconstructed with both AC methods and SUV values were compared.Figure 1 shows typical PET and MR images from the ¹⁵O-water study for slices at the level of the lateral ventricles and at the skull base. Clear gray–white matter contrast is observed for this healthy volunteer. Figure 2 shows the AC maps obtained by head atlas-based technique (a, b) and ZTE-based technique (c, d) for these two levels. The comparison shows that ZTE-based AC map is more consistent with known bony anatomy at the skull base. Figure 3 shows the comparison of brain PET images reconstructed with head atlas-based AC maps (a, b) and ZTE-based AC maps (c, d). The difference between the reconstructed PET images is multiplied by a factor of 5 and shown in (e) and (f), respectively. The brain PET images reconstructed from each AC map are very similar above the skull base (a, c), while there is a visible difference at the lower level (b, d). Because the ZTE-based AC map is more consistent with known bony anatomy at the skull base (Fig 2d), the PET images reconstructed with ZTE-based AC maps (Figs. 3c, 3d) are likely to be more accurate compared to those reconstructed with head atlas-based AC method (Figs. 3a, 3b). A zero-TE sequence was used for bone/soft tissue segmentation in the head and MRAC maps were generated and compared with head atlas-based MRAC maps. While CT scans were not available in these subjects for a gold-standard comparison, the ZTE-based MRAC maps identified known bony anatomy at the skull base better, such as the petrous temporal bone, and as a result the PET images in these regions are likely to be more accurate. NEX=2 and ST=1mm was used for the ZTE sequence in this study, and some errors were observed in bone tissue identification around ear canals and nasal cavities. Increasing signal to noise ratio with more averages and/or thicker slices (e.g. NEX=4, ST=2mm) may minimize these errors.