Data reduction for lesion detection in dynamic positron-emission tomography (PET)

The reconstruction of dynamic positron emission tomograph (PET) images is time consuming. Previous work to speed up the dynamic image reconstruction was based on low-order approximation using Karhunen-Loeve (KL) transform. Those KL transform basis vectors corresponding to the largest K eigenvalues were retained and then used in the reconstruction process. However, with this approach the kinetic characteristic (time activity curves) of weak signals, such as non-palpable tumors, may be lost or significantly changed in the reconstructed images. In this paper, a metric is found to select a subset of vectors from the KL transform basis, which relates directly to maximizing lesion-to-background ratio (signal-to-noise ratio). The results using computer simulation PET data show that by this approach the contrast of small lesion to the background can be objectively enhanced. The characteristic of the time activity curve in the lesions can also be approximately kept