A fully automated algorithm for estimating attenuation length in 3D from projection data

An algorithm is described for estimating attenuation length in 3D from sinogram or projection data, by which 3D attenuation correction can be performed in brain PET imaging for given attenuation coefficients. The algorithm begins with the estimation of scalp contours from sinogram data, based on optimal thresholding, spline smoothing and interpolation and backprojection. A cubic B-spline surface is then introduced to accurately describe the scalp surface by the least squares fitting to the stacked 2D estimated scalp contours. Due to the use of this surface model, an efficient and accurate technique is developed for estimation of a distance between two intersecting points formed by an arbitrary projection ray and the scalp surface. This fully automated and fast algorithm has been applied to sinogram and projection data collected using PENN-PET scanners. It has the potential to be used in routine clinical PET imaging.