Fast and accurate PET preclinical data analysis: Segmentation and Partial Volume Effect correction with no anatomical priors

摘要

The Partial Volume Effect hampers the quantification of rodent dynamic PET images. The work proposes a whole process for TAC extraction from the PET images without anatomical information requirement. The PET images were segmented using the Local Means Analysis segmentation method, and the resulting regions were used as spatial domains for a Geometric Transfer Matrix based method. The TAC estimation was assessed on phantom simulations and experimental datasets with ex-vivo measurements, and compared to reference methods: the mean TAC computation method the original Geometric Transfer Matrix method. Our GTM based TAC estimation method performed better than the GTM method and the mean TAC computation in terms of correlation of the estimated TACs with the true ones, of contrast recovery and error.