The De-noisingMethodsforSPECT Reconstruction withNon-uniform Attenuation Based on Wavelet

In single photonemission computedtomography (SPECT), thenon-stationary Possion noise intheprojection data (sinogram) isamajorcausetocompromise thequality ofthe reconstructed images. Toimprove thequality, wemustremove thePossion noise inthesinogram before reconstruction. However, theconventional spaceorfrequency domainde-noising methods possibly removetheedgeinformation thatisveryimportant for theaccurate reconstruction, especially forSPECTreconstruction withnon-uniform attenuation. Asatime-frequency analysis tool, wavelet transform hasbeenwidely usedinthesignal andimage processing fields, demonstrated itspowerful functions inthe application ofde-noising. Inthispaper, wetrytofindoutthe de-noising abilities ofthewavelet basedde-noising methods for SPECTreconstruction withnon-uniform attenuation, andthe effect oftheAnscombe transform inthewavelet basedde-noising. Fivemosteffective de-noising methodswereselected, andthe non-uniform attenuation reconstruction algorithm wasapplied tothede-noised projection data. Fromthereconstructed results, itisclear thattheRevised BivaShrink withcomplex wavelet is the bestwaveletbasedde-noising methodforSPECT reconstruction withnon-uniform attenuation, andtheeffect of Anscombetransform, whichconverts thePossion noiseinto Gaussian noise, isnotsignificant inthewavelet basedde-noising.