Naslov
2D PET Image Reconstruction Using Robust L1 Estimation of the Gaussian Mixture Model

Autori
Tafro, Azra ; Seršić, Damir ; Sović Kržić Ana

Izvornik
Informatica (0868-4952) (2022); 1-17

Vrsta, podvrsta i kategorija rada
Radovi u časopisima, članak, znanstveni

Ključne riječi
Gaussian mixture models ; positron emission tomography ; Expectation - Maximization (EM) algorithm ; image segmentation

Sažetak
An image or volume of interest in positron emission tomography (PET) is reconstructed from gamma rays emitted from a radioactive tracer, which are then captured and used to estimate the tracer’s location. The image or volume of interest is reconstructed by estimating the pixel or voxel values on a grid determined by the scanner. Such an approach is usually associated with limited resolution of the reconstruction, high computational complexity due to slow convergence and noisy results. This paper presents a novel method of PET image reconstruction using the underlying assumption that the originals of interest can be modelled using Gaussian mixture models. Parameters are estimated from one-dimensional projections using an iterative algorithm resembling the expectation- maximization algorithm. This presents a complex computational problem which is resolved by a novel approach that utilizes L1 minimization.

Izvorni jezik
Engleski

Znanstvena područja
Matematika, Elektrotehnika, Računarstvo

POVEZANOST RADA