Does OSEM achieve the lowest variance?

Cloquet, Christophe; Defrise, Michel

doi:10.1109/nssmic.2011.6153880

Cited by 7 publications

(6 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such limit probably depends on sinogram bin size, image bin size, and phantom size: for this phantom, it corresponds to a few true counts per sinogram bin, after randoms and scatter correction, and accounting only for the region occupied by the phantom. Other authors observed a similar threshold in order to obtain an unbiased image in MLEM reconstruction [20] or achieve the lowest theoretical variance [21]. Our work seems to indicate that, from the point of view of reproducibility and accuracy of the measurement, the rule of thumb of few counts per sinogram bin is also needed (for MLEM and OSEM).…”

Section: Discussionsupporting

confidence: 61%

See 1 more Smart Citation

Comparison of TOF and non-TOF iterative reconstruction at low statistics

Westerwoudt

Conti

Eriksson

2012

2012 IEEE Nuclear Science Symposium and Medical Imaging Conference Record (NSS/MIC)

View full text Add to dashboard Cite

In this paper, Monte Carlo generated data are used to simulated a time-of-flight (TOF) positron emission tomograph y (PET) scanner with improving time resolution, down to 200-ps time resolution. TOF and non-TOF ordered subsets expectation maximization (OSEM) reconstructions were compared in terms of contrast recover y and signal-to-noise ratio, with the purpose of characterizing the differences of TOF and non-TOF at ver y lowcounts. This stud y shows that there is a lower limit to the number of counts in a PET scan, in order to obtain a "reliable" PET image. Below such limit, the reliabilit y of the measurement is ver y poor. The use of TOF information has the beneficial effect of lowering this statistical limit, allowing for shorter reliable PET scans. Another interesting finding derives from the comparison with TOF filtered back projection (FBP) images: if time resolution can be improved, TOF FBP images become ver y competitive with TOF MLEM and OSEM algorithms, not onl y in terms of contrast recover y but also in terms of reduced noise level.

show abstract

Section: Discussionsupporting

confidence: 61%

“…The behavior of MLEM (and OSEM) variance at low statistics has been found to be anomalous by others [21]: below a threshold, again in the range 1-100 counts per bin (depending on the object and sampling), the empirical variance of OSEM starts being higher then the theoretical lower bound and then grows dramatically [21].…”

Section: Introductionmentioning

confidence: 99%

Comparison of TOF and non-TOF iterative reconstruction at low statistics

Westerwoudt

Conti

Eriksson

2012

2012 IEEE Nuclear Science Symposium and Medical Imaging Conference Record (NSS/MIC)

View full text Add to dashboard Cite

show abstract

“…bias, standard deviation, Co V) and the number of iterations. RT A causes high levels of overestimation and underestimation in cold and hot regions, respectively presumably due to non negativity constraints [28] and interpolation errors resulting in resolution degradation. It is found that for small iterations (e.g.…”

Section: Discussionmentioning

confidence: 99%

Statistical evaluation of PET motion correction methods using MR derived motion fields

Polycarpou

Tsoumpas

Marsden

2011

2011 IEEE Nuclear Science Symposium Conference Record

View full text Add to dashboard Cite

Although there have been various proposed methods for motion correction in PET, there is not sufficient evidence to answer which method is better in terms of image quality and quantification. This study aims to characterize the behavior of the two main motion correction methods in terms of convergence and image properties. During the first method, reconstruct transform-average (RT A), independent reconstructions of each gate are transformed to a reference gate and averaged. In the second method, motion-compensated image reconstruction (MCIR), all data are used with the motion information within a single reconstruction. The two methods are studied based on the OS EM algorithm. In this investigation motion fields were obtained from real dynamic MR acquisitions and concurrent PET data were simulated from the dynamic MR images. The two motion correction approaches were assessed statistically. Results indicate that MCIR is successful in the recovery of the true values of all regions, whereas RT A has high bias due to the non negativity constraints and interpolation errors during transformation. It has been demonstrated that whilst at low iterations (i.e. 46 sub-iterations) the two methods have similar noise characteristics; at high iterations MCIR has increased noise. Finally, for low iterations the two methods have marginally similar coefficient of variation (CoV) whereas for large number of iterations RTA has clearly better CoY than MCIR. This studyindicates that MCIR may provide superior performance overall to RT A providing the noise can be minimized.

show abstract

“…However, recent studies [3] clearly indicate that the popular iterative OS-EM (OS-EM) may not give the lowest variance for low count levels. In the work presented here, realistic distributions of common radiopharmaceuticals, a broad range of count levels and an adaptive application of post-smoothing are used.…”

Section: Theorymentioning

confidence: 99%

Impact of tracer distribution, count level, iterations and post-smoothing on PET quantification using a variously weighted least squares algorithm

Letourneau¹,

Verhaeghe²,

Reader³

2012

2012 IEEE Nuclear Science Symposium and Medical Imaging Conference Record (NSS/MIC)

View full text Add to dashboard Cite

This work is a task-oriented quantitative analysis to assess the best range of reconstruction algorithms parameters which are most suited for quantitative PET imaging of the brain. Starting from a general iterative weighted-least squares objective function, well-known methods such as the Maximum Likelihood Expectation Maximization (MLEM), the Image Space Reconstruction Algorithm (lSRA) and related techniques are obtained. A detailed analysis was done for single-frame imaging as well as a comprehensive study on the impact of post-smoothing by applying a Gaussian convolution kernel at each and every possible post-reconstruction iteration number in order to achieve the lowest mean absolute error (MAE). It has been clearly demonstrated that an appropriate choice of iteration number and post-smoothing level makes most algorithms deliver similar MAE within an interval of 5%, with a considerable reduction of the MAE in comparison to non-optimized reconstructions. Also, reconstruction method performance was influenced by the total number of counts and the activity distribution, meaning that each radiopharmaceutical is not always best reconstructed by the same method. Broadly speaking, MLEM with a point spread function (PSF), [SRA PSF with smoothed expected data as weighting factors and filtered backprojection (FBP), when applied with a suitable level of post smoothing, were the methods offering the best quantitative images and from these three methods, FBP offered the most robust performance for a broad range of post-smoothing levels. An appropriate choice of iteration number and post-smoothing level for a task-oriented analysis significantly lowered the MAE in comparison with the standard practice of using default fixed numbers of iterations and/or post-smoothing.

show abstract

Does OSEM achieve the lowest variance?

Cited by 7 publications

References 16 publications

Comparison of TOF and non-TOF iterative reconstruction at low statistics

Comparison of TOF and non-TOF iterative reconstruction at low statistics

Statistical evaluation of PET motion correction methods using MR derived motion fields

Impact of tracer distribution, count level, iterations and post-smoothing on PET quantification using a variously weighted least squares algorithm

Contact Info

Product

Resources

About