Accelerated PET kinetic maps estimation by analytic fitting method

• Published in: Computers in biology and medicine Vol. 99; pp. 221 - 235
• Main Authors: Scipioni, Michele, Giorgetti, Assuero, Della Latta, Daniele, Fucci, Sabrina, Positano, Vincenzo, Landini, Luigi, Santarelli, Maria Filomena
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.08.2018; Elsevier Limited
• Subjects: 18F[FDG] 4D PET; Algorithms; Analytic fitting; Computer applications; Computer simulation; Computing time; Convergence; Data processing; Dynamic PET; Kinetic modeling; Medical imaging; Methods; Non-linear least square fitting; Ordinary differential equations; Parameter estimation; Parametric images; Tomography; Non-linear least square fitting; Dynamic PET; Analytic fitting; Kinetic modeling; Parametric images; 18F[FDG] 4D PET
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2018.06.015

In this work, we propose and test a new approach for non-linear kinetic parameters' estimation from dynamic PET data. A technique is discussed, to derive an analytical closed-form expression of the compartmental model used for kinetic parameters' evaluation, using an auxiliary parameter set, with the aim of reducing the computational burden and speeding up the fitting of these complex mathematical expressions to noisy TACs. Two alternative algorithms based on numeric calculations are considered and compared to the new proposal. We perform a simulation study aimed at (i) assessing agreement between the proposed method and other conventional ways of implementing compartmental model fitting, and (ii) quantifying the reduction in computational time required for convergence. It results in a speed-up factor of ∼120 when compared to a fully numeric version, or ∼38, with respect to a more conventional implementation, while converging to very similar values for the estimated model parameters. The proposed method is also tested on dynamic 3D PET clinical data of four control subjects. The results obtained supported those of the simulation study, and provided input and promising perspectives for the application of the proposed technique in clinical practice. [Display omitted] •Voxel-wise fitting of pharmacokinetic model for map estimation is highly time consuming.•A novel closed-form analytic implementation of compartmental models is described.•Speed-up obtained: ∼120x vs a fully numeric version and ∼38x vs a conventional one.•Speed-up does not affect parameter estimation, both on simulated and real data.•It could make PET 4D direct reconstruction methods feasible in clinical practice.