A new component-based normalization method for PET cameras

1793 Objectives: In positron emission tomography (PET), in order to generate artifact-free images, the nonuniformity of the sensitivity of the lines of response (LORs) must be corrected with appropriate normalization coefficients (NCs). There are two commonly used methods for estimating the normalization coefficients in 3D PET: direct method and component-based method. In direct method the NC is proportional to the inverse of the counts acquired for a given LOR when every LOR is illuminated by the same level of activity. The direct method requires long acquisition time (many hours) for good statistic in each LOR. In component-based technique the NC is generally expressed as the product of crystals efficiencies and geometric factors. In this work, we present a new and fast component-based normalization method for 3D PET data which also accounts for count-rate effects in NCs. Methods: In this new method the geometric correction factor is expressed as the product of two components: the first component is only a function of radial position and second term depends on location of the crystal in the block and in the module. The first geometric component which is count-rate independent is calculated from the geometry of the crystals in camera’s ring and does not change unless camera configuration changes. The second geometric component, in combination with the detector efficiency, can be obtained from data acquired using a uniformly distributed activity. In this method, the number of unknown parameters is reduced from the extremely large number of LORs to the number of crystals in the camera. This makes it possible to collect data with reasonable statistics in short time (e.g. 10 minutes). Results: We have evaluated the new normalization method for a dedicated small animal PET camera operating in 3D mode over a wide range of count-rates and compared the results with direct method using a 2 inch uniform normalization phantom for several different count rates. Currently we are also implementing this method for a whole-body PET camera and initial results are very promising. Conclusions: The new component-based normalization method for 3D PET data allows a fast acquisition of normalization data for different levels of count rates and improves the image quality compared to the much slower direct method.