A novel model of the geometric and detector response for limited angular sampling pinhole SPECT

Reconstruction methodologies for data sets with reduced angular sampling (RAS) are essential for efficient dynamic or static preclinical animal imaging research using single photon emission computed tomography (SPECT). Modern iterative reconstruction methods can obtain 3D radiotracer distributions of the highest possible quality and resolution. Essential to these algorithms is an accurate model of the physical imaging process. We developed a new point-spread function (PSF) model for the pinhole geometry and compared it to a Gaussian model in a RAS setting. The new model incorporates the geometric response of the pinhole and the detector response of the camera by simulating the system PSF using the error function. Reconstruction of simulated data was done with OS-EM and COS-EM: a new convergent OS-EM based algorithm. The reconstruction of projection data of a simulated point source using the novel method showed improved FWHM values compared to a standard Gaussian method. COS-EM delivers improved results for RAS data, although it converges slower than OS-EM. The reconstruction of Monte Carlo simulated projection data from a resolution phantom shows that as few as 40 projections are sufficient to reconstruct an image with a resolution of approximately 4 mm. The new pinhole model applied to iterative reconstruction methods can reduce imaging time in small animal experiments by a factor of three or reduce the number of cameras needed to perform dynamic SPECT.