Locally adaptive filtering for edge preserving noise reduction on images with low SNR in PET

As well known, the signal-to-noise ratio (SNR) of PET images can be considerably low. This is especially true for whole-body examinations of heavy patients, for respiratory gated studies, and dynamic studies with short frames. In these cases moving average filters (MAF) such as a Gaussian filter are applied in order to achieve an acceptable SNR. Image resolution is, however, considerably reduced by these MAFs. This affects detectability and quantification of small structures. Interesting alternatives to MAFs are non-linear, locally adaptive filters (NLF), which enable noise reduction while preserving sharp edges. It was the aim of this study to investigate the performance of a special NLF (bilateral filter, BF) when applied to PET images with a low SNR. In three phantom studies using a cylinder phantom with sphere inserts different signal-to-background ratios have been investigated. In addition, images with different noise levels were generated. Finally, respiratory-gated whole-body studies were analyzed. All data were filtered, both with BF and MAF. The results were analyzed regarding noise level, image resolution and relative signal recovery. In the phantom studies BF is able to preserve the spatial resolution near the edges of the spheres while improving the noise characteristics equally to MAF. Signal recovery even of small spheres is not significantly reduced. Using MAF compromises the spatial resolution and leads to unacceptable reduction of signal recovery. The positive properties of BF were also apparent when applying it to single gates of respiratory-gated studies, which otherwise were not suitable for visual inspection. NLF is a powerful alternative to MAF commonly used in PET. For studies with high noise and high signal-to-background ratios using NLF represents a suitable filter for edge preserving image enhancement. Its performance, however, is critically dependent on a sensible choice of its intensity and spatial dependent part.