A Fourier rebinning algorithm incorporating spectral transfer efficiency for 3D PET.

This paper presents a Fourier rebinning algorithm for three-dimensional image reconstruction in PET that incorporates the concept of spectral transfer function. It suggests the need for discarding low-frequency components in the rebinning. It also includes the correction for rebinning efficiency which was evaluated by simulations as a function of oblique angle of projections. The performance was optimized by high-pass filters and axial smoothing. The algorithm yields satisfactory images with negligible axial cross-talk for a maximum oblique angle up to 26.6 degrees. The statistical noise was evaluated in terms of 'noise equivalent number of oblique angles', and reasonable results were obtained in view of the theoretical expectation. Ring artefacts due to noise are negligibly small.