Joint estimation of the Patlak model and plasma input function within a direct 4D parametric image reconstruction algorithm integrating respiratory motion correction for PET oncology applications

429 Objectives Kinetic modelling techniques such as Patlak analysis usually require the knowledge of the plasma input function theoretically derived from arterial blood sampling which is both invasive and tedious to perform. Image derived input functions (IDIF) are a potential alternative but face additional issues as the plasma input function cannot be directly measured and quantitative corrections are needed in order to limit estimation bias. This study aims at improving kinetic modelling for lung cancer PET studies by combining, within a 5D reconstruction algorithm including respiratory motion correction, the estimation of plasma and Patlak kinetic parameters. Methods Respiratory motion correction is performed using elastic deformation parameters derived from respiratory gated PET images. The derived displacement matrices are subsequently integrated in a 4D list-mode based OSEM reconstruction algorithm incorporating both Patlak and plasma function models in order to alternatively estimate parametric images and input function from a dynamic dataset. Quantitative evaluation was performed using Monte-Carlo simulations of an hour dynamic FDG PET acquisition of the NCAT phantom including respiratory motion. The performance of the proposed direct method was compared with indirect kinetic analysis using IDIFs estimated from the aorta. Results Results indicate a bias reduction for the Patlak and plasma model parameters estimation when they are obtained during the reconstruction. The incorporation of elastic motion correction allowed significant estimation bias reduction (~25%) for the Patlak slope in the simulated tumor region. Conclusions Joint incorporation of the plasma input function and Patlak models in a 4D reconstruction process allows to improve both plasma function and Patlak parameters estimation. Further evaluation of the method on 18F-FDG clinical datasets of patients with non-small cell lung carcinoma will be presented.