Noise and Edge Artifacts in Maximum-Likelihood Reconstructions for Emission Tomography

Snyder, Donald L.; Miller, Michael I.; Thomas, Lewis J.; Politte, David G.

doi:10.1109/tmi.1987.4307831

Cited by 367 publications

(214 citation statements)

References 11 publications

Supporting

Mentioning

203

Contrasting

Unclassified

Order By: Relevance

“…We have implemented penalized-likelihood algorithms based on Hebert's GEM strategy [16] for both the ML-IA and ML-IB completedata spaces. We have also implemented both ML-IA and ML-IB with sieve constraints [1,8,19]. We found empirically that the penalized ML-IB algorithm converged substantially more rapidly, in terms of both likelihood increase and apparent image contrast.…”

Section: Discussion 9 V Discussionmentioning

confidence: 96%

On complete-data spaces for PET reconstruction algorithms

Fessler

Clinthorne

Rogers

1993

IEEE Trans. Nucl. Sci.

View full text Add to dashboard Cite

As investigators consider more comprehensive measurement models for emission tomography, there will be more choices for the complete-data spaces of the associated expectation-maximization (EM) algorithms for maximum-likelihood (ML) estimation. In this paper, we show that EM algorithms based on smaller complete-data spaces will typically converge faster. We discuss two practical applications of these concepts: (i) the ML-IA and ML-IB image reconstruction algorithms of Politte and Snyder [1] which are based on measurement models that account for attenuation and accidental coincidences in positron-emission tomography (PET), and (ii) the problem of simultaneous estimation of emission and transmission parameters. Although the PET applications may often violate the necessary regularity conditions, our analysis predicts heuristically that the ML-IB algorithm, which has a smaller complete-data space, should converge faster than ML-IA. This is corroborated by the empirical findings in [1].

show abstract

Section: Discussion 9 V Discussionmentioning

confidence: 96%

On complete-data spaces for PET reconstruction algorithms

Fessler

Clinthorne

Rogers

1993

IEEE Trans. Nucl. Sci.

View full text Add to dashboard Cite

show abstract

“…Researchers believe that ringing artifacts occur because the reconstruction problem is ill-posed [1]. The experimental results shown above indicate that the type of artifact objects of the same shape develop is related to the object size.…”

Section: Artifacts and The Ratio Of Object Size Over Fwhmmentioning

confidence: 95%

“…It has been observed that the SPECT images reconstructed by expectationmaximization algorithms, such as ML-EM and OS-EM, are more noisy and have strong edge artifacts when the total number of iterations in the reconstruction is large [1,2].…”

Section: Ringing Artifactsmentioning

confidence: 99%

“…There are several known sources that can introduce distortions into SPECT. For example, Compton scattering and photoelectric absorption can be regarded as physical factors that cause errors in the SPECT projection data, the geometry of the collimators and the nonuniform sensitivity of the gamma camera are measurement factors that cause distortion in data acquisition, and the reconstruction algorithms, such as maximum likelihood-expectation maximization (ML-EM) and ordered subset expectation maximization (OS-EM) can cause high noise and edge artifacts [1].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Nature of Ringing Artifacts in Spect Reconstruction and Their Reduction by the Use of Side Information

Koral

2007

2007 4th IEEE International Symposium on Biomedical Imaging: From Nano to Macro

View full text Add to dashboard Cite

ML-EM iterative reconstruction algorithms tend towards ringing artifacts, and high noise, after a large number of iterations just as filtered backprojection algorithms tend toward streak artifacts. OS-EM is a type of ML-EM algorithm and an edge overshoot has been observed experimentally for uniform-activity spheres of large size but not smaller size. Herein, we shed light on such artifacts by showing that they are a function of the object size relative to the FWHM of the collimator-detector response function. That is, for a given number of iterations, an over-or undershoot can occur at a given spatial location or not depending on the value of the ratio. We further show that a new, regularized, penalized-likelihood algorithm based on ordered subsets and employing side information stabilizes the iterative procedure against the development of artifacts.

show abstract

“…4 and 5 often lead to more noisy activity imagesf [59]. One simple and practical solution for this issue is to use an early stopping rule.…”

Section: Penalized Ml-em Algorithmmentioning

confidence: 99%

The Use of Anatomical Information for Molecular Image Reconstruction Algorithms: Attenuation/Scatter Correction, Motion Compensation, and Noise Reduction

Chun

2016

Nucl Med Mol Imaging

View full text Add to dashboard Cite

PET and SPECT are important tools for providing valuable molecular information about patients to clinicians. Advances in nuclear medicine hardware technologies and statistical image reconstruction algorithms enabled significantly improved image quality. Sequentially or simultaneously acquired anatomical images such as CT and MRI from hybrid scanners are also important ingredients for improving the image quality of PET or SPECT further. High-quality anatomical information has been used and investigated for attenuation and scatter corrections, motion compensation, and noise reduction via post-reconstruction filtering and regularization in inverse problems. In this article, we will review works using anatomical information for molecular image reconstruction algorithms for better image quality by describing mathematical models, discussing sources of anatomical information for different cases, and showing some examples.

show abstract

Noise and Edge Artifacts in Maximum-Likelihood Reconstructions for Emission Tomography

Cited by 367 publications

References 11 publications

On complete-data spaces for PET reconstruction algorithms

On complete-data spaces for PET reconstruction algorithms

Nature of Ringing Artifacts in Spect Reconstruction and Their Reduction by the Use of Side Information

The Use of Anatomical Information for Molecular Image Reconstruction Algorithms: Attenuation/Scatter Correction, Motion Compensation, and Noise Reduction

Contact Info

Product

Resources

About