Activity outside the field-of-view (FOV) degrades the count rate performance of 3D PET and consequently reduces signal-to-noise ratios (SNRs) of reconstructed images. The aim of this study was to evaluate a neck-shield installed in a 3D PET scanner for reducing the effect of the outside FOV activity. Specifically, we compared brain PET scans ((15)O(2) and H(2)(15)O) with and without the use of the neck-shield. Image SNRs were directly estimated by a sinogram bootstrap method. The bootstrap analysis showed that the use of the neck-shield improved the SNR by 8% and 19% for H(2)(15)O and (15)O(2), respectively. The SNR improvements were predominantly due to the reduction of the random count rates. Noise equivalent count rate (NECR) analysis provided SNR estimates that were very similar with the bootstrap-based results for H(2)(15)O, but not for (15)O(2). This discrepancy may be due to the fundamental difference between the two methods: the bootstrap method directly calculates the local SNR of reconstructed images, whereas the NECR calculation is based on the whole-gantry count rates, indicating a limitation of the conventional NECR-based method as a tool for assessing the image SNR. Although quantitative parameters, e.g. cerebral blood flow, did not differ when examined with and without the neck-shield, the use of the shield for brain (15)O study is recommended in terms of the image SNR.