Faster PET reconstruction with non-smooth priors by randomization and preconditioning

Matthias Joachim Ehrhardt, Pawel J Markiewicz, Carola-Bibiane Schönlieb

Research output: Contribution to journalArticle

Abstract

Uncompressed clinical data from modern positron emission tomography (PET) scanners are very large, exceeding 350 million data points (projection bins). The last decades have seen tremendous advancements in mathematical imaging tools many of which lead to non-smooth (i.e. non-differentiable) optimization problems which are much harder to solve than smooth optimization problems. Most of these tools have not been translated to clinical PET data, as the state-of-the-art algorithms for non- smooth problems do not scale well to large data. In this work, inspired by big data machine learning applications, we use advanced randomized optimization algorithms to solve the PET reconstruction problem for a very large class of non-smooth priors which includes for example total variation, total generalized variation, directional total variation and various different physical constraints. The proposed algorithm randomly uses subsets of the data and only updates the variables associated with these. While this idea often leads to divergent algorithms, we show that the proposed algorithm does indeed converge for any proper subset selection. Numerically, we show on real PET data (FDG and florbetapir) from a Siemens Biograph mMR that about ten projections and backprojections are sufficient to solve the MAP optimisation problem related to many popular non-smooth priors; thus showing that the proposed algorithm is fast enough to bring these models into routine clinical practice.

Original languageEnglish
JournalPhysics in Medicine and Biology
Early online date20 Aug 2019
DOIs
Publication statusE-pub ahead of print - 20 Aug 2019

Cite this

Faster PET reconstruction with non-smooth priors by randomization and preconditioning. / Ehrhardt, Matthias Joachim; Markiewicz, Pawel J; Schönlieb, Carola-Bibiane.

In: Physics in Medicine and Biology, 20.08.2019.

Research output: Contribution to journalArticle

@article{b6884f16de224308b3d50701ef04473e,
title = "Faster PET reconstruction with non-smooth priors by randomization and preconditioning",
abstract = "Uncompressed clinical data from modern positron emission tomography (PET) scanners are very large, exceeding 350 million data points (projection bins). The last decades have seen tremendous advancements in mathematical imaging tools many of which lead to non-smooth (i.e. non-differentiable) optimization problems which are much harder to solve than smooth optimization problems. Most of these tools have not been translated to clinical PET data, as the state-of-the-art algorithms for non- smooth problems do not scale well to large data. In this work, inspired by big data machine learning applications, we use advanced randomized optimization algorithms to solve the PET reconstruction problem for a very large class of non-smooth priors which includes for example total variation, total generalized variation, directional total variation and various different physical constraints. The proposed algorithm randomly uses subsets of the data and only updates the variables associated with these. While this idea often leads to divergent algorithms, we show that the proposed algorithm does indeed converge for any proper subset selection. Numerically, we show on real PET data (FDG and florbetapir) from a Siemens Biograph mMR that about ten projections and backprojections are sufficient to solve the MAP optimisation problem related to many popular non-smooth priors; thus showing that the proposed algorithm is fast enough to bring these models into routine clinical practice.",
author = "Ehrhardt, {Matthias Joachim} and Markiewicz, {Pawel J} and Carola-Bibiane Sch{\"o}nlieb",
note = "Creative Commons Attribution license.",
year = "2019",
month = "8",
day = "20",
doi = "10.1088/1361-6560/ab3d07",
language = "English",
journal = "Physics in Medicine and Biology",
issn = "0031-9155",
publisher = "IOP Publishing",

}

TY - JOUR

T1 - Faster PET reconstruction with non-smooth priors by randomization and preconditioning

AU - Ehrhardt, Matthias Joachim

AU - Markiewicz, Pawel J

AU - Schönlieb, Carola-Bibiane

N1 - Creative Commons Attribution license.

PY - 2019/8/20

Y1 - 2019/8/20

N2 - Uncompressed clinical data from modern positron emission tomography (PET) scanners are very large, exceeding 350 million data points (projection bins). The last decades have seen tremendous advancements in mathematical imaging tools many of which lead to non-smooth (i.e. non-differentiable) optimization problems which are much harder to solve than smooth optimization problems. Most of these tools have not been translated to clinical PET data, as the state-of-the-art algorithms for non- smooth problems do not scale well to large data. In this work, inspired by big data machine learning applications, we use advanced randomized optimization algorithms to solve the PET reconstruction problem for a very large class of non-smooth priors which includes for example total variation, total generalized variation, directional total variation and various different physical constraints. The proposed algorithm randomly uses subsets of the data and only updates the variables associated with these. While this idea often leads to divergent algorithms, we show that the proposed algorithm does indeed converge for any proper subset selection. Numerically, we show on real PET data (FDG and florbetapir) from a Siemens Biograph mMR that about ten projections and backprojections are sufficient to solve the MAP optimisation problem related to many popular non-smooth priors; thus showing that the proposed algorithm is fast enough to bring these models into routine clinical practice.

AB - Uncompressed clinical data from modern positron emission tomography (PET) scanners are very large, exceeding 350 million data points (projection bins). The last decades have seen tremendous advancements in mathematical imaging tools many of which lead to non-smooth (i.e. non-differentiable) optimization problems which are much harder to solve than smooth optimization problems. Most of these tools have not been translated to clinical PET data, as the state-of-the-art algorithms for non- smooth problems do not scale well to large data. In this work, inspired by big data machine learning applications, we use advanced randomized optimization algorithms to solve the PET reconstruction problem for a very large class of non-smooth priors which includes for example total variation, total generalized variation, directional total variation and various different physical constraints. The proposed algorithm randomly uses subsets of the data and only updates the variables associated with these. While this idea often leads to divergent algorithms, we show that the proposed algorithm does indeed converge for any proper subset selection. Numerically, we show on real PET data (FDG and florbetapir) from a Siemens Biograph mMR that about ten projections and backprojections are sufficient to solve the MAP optimisation problem related to many popular non-smooth priors; thus showing that the proposed algorithm is fast enough to bring these models into routine clinical practice.

U2 - 10.1088/1361-6560/ab3d07

DO - 10.1088/1361-6560/ab3d07

M3 - Article

JO - Physics in Medicine and Biology

JF - Physics in Medicine and Biology

SN - 0031-9155

ER -