Learned Primal Dual Reconstruction for PET.

Alessandro Guazzo,Massimiliano Colarieti-Tosti

doi:10.3390/jimaging7120248

Alessandro Guazzo, Massimiliano Colarieti-Tosti

Open Access

https://doi.org/10.3390/jimaging7120248

Copy DOI

Abstract

We have adapted, implemented and trained the Learned Primal Dual algorithm suggested by Adler and Öktem and evaluated its performance in reconstructing projection data from our PET scanner. Learned Primal Dual reconstructions are compared to Maximum Likelihood Expectation Maximisation (MLEM) reconstructions. Different strategies for training are also compared. Whenever the noise level of the data to reconstruct is sufficiently represented in the training set, the Learned Primal Dual algorithm performs well on the recovery of the activity concentrations and on noise reduction as compared to MLEM. The algorithm is also shown to be robust against the appearance of artefacts, even when the images that are to be reconstructed present features were not present in the training set. Once trained, the algorithm reconstructs images in few seconds or less.

Highlights

Positron Emission Tomography (PET) is an inherently three-dimensional molecular imaging technique that is able to image the distribution of an injected radioactive tracer in vivo
The PET imagereconstruction problem can be seen as the task of estimating the unknown 3D activity map, f, from data, g = A( f ) + ν, where ν is generated by a random variable and represents observation error and quantum noise
This paper focuses on the results obtained with the Learned Primal Dual (LPD)

Summary

Introduction

Positron Emission Tomography (PET) is an inherently three-dimensional molecular imaging technique that is able to image the distribution of an injected radioactive tracer in vivo (for an introduction see, for example, [1,2]). PET image reconstruction deals with the tomographic inverse problem of finding an image, f ∈ X , given a finite number of noisy projections, g ∈ Y. Measured PET projection data, g, can be seen as a noisy realisation of a PET forward-projection operator, A : X → Y , that acts on the 3D activity map, f. The PET imagereconstruction problem can be seen as the task of estimating the unknown 3D activity map, f , from data, g = A( f ) + ν, where ν is generated by a random variable and represents observation error and quantum noise. Ionising-radiation based medical imaging faces a compelling trade-off between image quality and dose, and this, together with the relatively low detection efficiency of PET, pushes data acquisition towards a low-count regime, where Poisson noise dominates

Methods

Results

Discussion

Conclusion