Quantitative Accuracy Assessment and Optimization of SPECT Geometrical Calibration

Accurate geometrical calibration is critical to obtaining high resolution and artifact free reconstructed images for modern animal single photon emission computed tomography (SPECT) systems. Although there have been many published works on the calibration of various SPECT systems, few studies have been done to evaluate the efficacy of the proposed calibration methods in a quantitative manner. This paper presents a numerical method to assess both the uniqueness and the quantitative accuracy of SPECT calibration, which is based on analyzing the singular value decomposition (SVD) components of the Jacobian matrix from a least-square cost function of the calibration. The proposed method is firstly validated by applying it to the calibration of a single pinhole SPECT system and comparing the results with those derived using a published method, and is then used to optimize the calibration setup for a slit-slat SPECT system. With the proposed method, a minimum required number of point source projections to achieve the desired calibration accuracy can be estimated and used as figure-of-merit to evaluate the goodness of a calibration setup. An inverse-square relationship between the calibration accuracy and the number of sampled projections is revealed. Optimal calibration setup is determined through an exhaustive search among all the possibilities of point source arrangements under certain conditions. We demonstrate that for the studied system, the best calibration accuracy is achieved by arranging the point source over the edge of FOV with evenly-spaced angular positions. Point source experiments were conducted to validate the proposed method.