Gaussian Filter for Brain SPECT Imaging

Background. The presence of a noise component on 3D images single-photon emission computed tomo­graphy (SPECT) brain significantly distorts the probability distribution function (PD) radioactive count rate in images. and further filtering data, based subjective assessment image quality, have significant impact calculation volumetric cerebral blood flow values uptake asymmetry radiopharmaceutical brain. Objective. We are aimed to develop method for optimal SPECT with lipophilic radiopharmaceuticals, Gaussian filter (GF), subsequent segmentation by threshold method. Methods. water phantom patients 99mTc-HMPAO were used. developed technique artificial addition speckle conditionally flawless data order determine parameters smoothing SPECT, GF. quantitative criterion was standard deviation between PD smoothed ideal one. Results. It shown that maximum has an extremum changing GF range 0.3–0.4 pixels. greater image, more quasi-linearly corresponding changes. This dependence allows determining parameters. application restoring (distribution histogram) accuracy up 5–10%. provides possibility standardize Conclusions. research results work solve specific applied problem: restoration histogram radiopharmaceuticals correct regional flow. In contrast well-known publications filtration takes into account initial tomographic 3D, rather than 2D slices, contain not only uniform random noise, but also pronounced component.