Abstract:Given the noise nature of positron emission tomography (PET) measurements, it is critical to know the image quality and reliability as well as expected radioactivity map (mean image) for both qualitative interpretation and quantitative analysis. While existing efforts have often been devoted to providing only the reconstructed mean image, we present a unified framework for joint estimation of the mean and corresponding variance of the radioactivity map based on an efficient optimal min-max criterion. The proposed framework formulates the PET image reconstruction problem to be a transformation from system uncertainties to estimation errors, where the minimax criterion is adopted to minimize the estimation errors with possibly maximized system uncertainties. The estimation errors, in the form of a covariance matrix, express the measurement uncertainties in a complete way. The framework is then optimized by infinity-norm optimization and solved with the corresponding H-infinity filter. Unlike conventional statistical reconstruction algorithms, that rely on the statistical modeling methods of the measurement data or noise, the proposed joint estimation stands from the point of view of signal energies and can handle from imperfect statistical assumptions to even no a priori statistical assumptions. The performance and accuracy of reconstructed mean and variance images are validated using Monte Carlo simulations. Experiments on phantom scans with a small animal PET scanner and real patient scans are also conducted for assessment of clinical potential.

PET image reconstruction: a robust state space approach.

State Space Strategies for Estimation of Activity Map in PET Imaging

State-Space Reconstruction of Pet Parametric Maps

A robust state-space kinetics-guided framework for dynamic PET image reconstruction.

Robust Framework for PET Image Reconstruction Incorporating System and Measurement Uncertainties.

Robust Reconstruction Of Physiological Parameters From Dynamic Pet Data

Limited View PET Reconstruction of Tissue Radioactivity Maps.

PEt image reconstruction based on particle filter framework

Uncertainty Penalized Weighted Least Squares Framework For Pet Reconstruction Under Uncertain System Models

Robust adaptive Kalman filter for PET image reconstruction

Sparse/Low Rank Constrained Reconstruction for Dynamic PET Imaging.

PET Image Reconstruction: Mean, Variance, and Optimal Minimax Criterion

Pet Image Reconstruction: GPU-accelerated Particle Filter Framework

A Robust Least Squares Framework for PET Image Reconstruction

Joint Reconstruction of Dynamic PET Activity and Kinetic Parametric Images Using Total Variation Constrained Dictionary Sparse Coding

Graphics processing unit (GPU)-accelerated particle filter framework for positron emission tomography image reconstruction.

State Space Approach for Joint Estimation of Activity and Attenuation Map from Pet Emission Sinograms

Simultaneous Reconstruction and Segmentation for Dynamic PET: A Low Rank Framework.

Deep Reconstruction Model For Dynamic Pet Images

A Novel Static PET Image Reconstruction Method

EMISTA-Based Quantitative PET Reconstruction.