Component based normalization for PET systems with depth of interaction measurement capability

The number of lines of response (LOR) in a system with small crystals and depth of interaction (DOI) measurement capability can require prohibitive amounts of memory and computational resources. The number of independent LORs in the small animal imaging system evaluated in this study is ∼2 × 10⁹. We propose a method to adapt the component-based normalization model to include (1) DOI and (2) large field of view (FOV) to system inner bore volume percentage (near 100%) or percentage of view (POV) considerations without an order of magnitude increase in computational resources. This method can also compensate for bin location errors made during the LOR calculation. Included in the study were radial, axial and interference geometric efficiency and crystal absorption factors. We have observed a large effect on sensitivity from axial geometric factors, distance between positioned coincident photon interactions, and amount of crystal surrounding any particular crystal in our system. Assessing direct plane sinograms of direct normalization data (infinitely thin cylindrical annulus), we have observed that there are LORs that received zero counts due to inter-module gaps. Component based normalization pre-correction weights significantly improved the coefficient of variation for the reconstructed 1.5 mm and 1.75 mm spheres in the row nearest the center line in the axial image slices by 17%, 32%, and 51% in the z = 0 cm, z = 1.5 cm, and z = 3.0 cm axial planes, respectively.