Rapid construction of pinhole SPECT imaging system matrices by Gaussian interpolation method combined with geometric parameter estimations

In pinhole SPECT applied to small-animal studies, it is essential to have an accurate imaging system matrix, called H matrix, for high-spatial-resolution image reconstructions. Generally, an H matrix can be obtained by various methods, such as empirical measurements, simulations or some combinations of both methods. In this study, Gaussian Interpolation Method combined with Geometric Parameter Estimations (GIMGPE) was proposed. It utilizes a simplified grid-scan experiment on selected voxels and parameterizes the measured images into 2D Gaussians. The image responses of missing voxels were interpolated by the relations between the Gaussian coefficients and the geometric parameters of the imaging system. A full H matrix is constructed by combining the measured and interpolated image responses of all voxels. OSEM reconstructed images of a resolution phantom and the detectability of a SKE/BKE synthetic spherical test-object demonstrated that the constructed H matrix provided comparable resolution and detectability respectively as the H matrix acquired by a full 3D grid-scan experiment. The reduction in the acquisition time of the grid-scan experiment was about 3.6 times on the preliminary trial. Finer-grid interpolations by the proposed method are currently under investigation.