Adaptive Speckle Reducing Anisotropic Diffusion Filter for Positron Emission Tomography Images Based on Anatomical Prior

Abstract

Positron Emission Tomography (PET)/Computed Tomography (CT) is the main medical imaging technique which used for diagnosing cancer. PET image is showing the functional activities in the patient while CT imaging presents the anatomical information. The PET raw-projection data (sinogram) contains a very high level of Poisson noise, while The reconstructed image through filtered back-projection algorithm (FBP) is contaminated with unknown noise that is very similar to speckle noise distribution. This noise may lead to increase the doze of radioactive material that given to the patient for imaging PET and to errors in the diagnosis results. Applying a suitable filtering approach can increase the effectiveness of the diagnosing process. Using the high resolution information in the CT, we propose in this work an adaptive post-reconstruction curvature motion filtering technique for PET image. The proposed filter consider computing the diffusivity function (edge stopping function) based on the fused image (PET/CT) to guide the smoothing and the sharpening process in the image. Experiments demonstrate through simulated images that the performance of the proposed method significantly enhance the reconstructed PET using FBP algorithm. Further, it compared with recently published methods, both visually and in terms of statistical measures.