| Abstrakt: |
A medical image filtering algorithm based on spatial autocorrelation is proposed. The Moran's Z calculation is used to adapt the average filter. The noisy PET image is reconstructed using MLEM iteration with a low-pass filter applied in image quality improvement post-reconstruction. In practice, a Gaussian filter is often performed for post-filtering. Although this low-pass filter removes the noise in reconstructed PET images, it is associated with significant quantitative information loss. Two phantoms, Huffman and Utah, are used in this work to simulate a Siemens PET scanner using the GATE algorithm. The sinogram was reconstructed using the OSEM algorithm on Matlab. The image noise level with different numbers of iteration were then calculated and filtered using an average filter with or without Moran's Z adaptation. Image qualities were estimated using PSNR calculation. The PSNR of images using the proposed scheme are higher than those using the non-adaptive average filter. The algorithm visual processing is good but more forward research is required. [ABSTRACT FROM AUTHOR] |
